Lamented the hurdles to Mima mobile to a colleague who suggested a sandbox solution I never would have noticed if they hadn't pulled my head out of browser land.
See, this is why you can't build software using agents alone. Too much unwritten, too many jumps between disparate knowledge clusters
Rachel Nabors
Rachel Nabors (born March 9, 1985) is an American cartoonist, artist, and graphic novelist, best known for their serialized comic, Rachel the Great, as well as their two graphic novels, 18 Revolutions and Crow Princess. More on Wikipedia
-
R 'Nearest' Nabors
@rachelnabors
-
R 'Nearest' Nabors reposted
Ruben Casas 🦊
@Infoxicador
MCP Events are so cool!
inspired by @Poita_ I got ChatGPT on the left playing a chess match against Opus on the right (via Postman Agent Mode)
Using MCP Apps and events both agents subscribe to the game and automatically awake when it is their turn...
let's see who wins!This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
As much as I would love to rent coworking space, I need that money to pay people and agents to ship!
-
R 'Nearest' Nabors
@rachelnabors
Hi from mima.social!
-
R 'Nearest' Nabors
@rachelnabors
This is Kitty, my last dog. She was my best friend, taught me much. For the past 4 years, my goal has been to get a dog again. Soon we will move into our first home together, and after that, we open it up to welcome a pup of our own. This is how we plan to spend our holidays
-
R 'Nearest' Nabors
@rachelnabors
Steak is for shippers
R 'Nearest' Nabors
@rachelnabors
I have to hit XYZ WAU for the next steak. Wish me luck.
-
R 'Nearest' Nabors
@rachelnabors
Me at my most neon
JS Monthly London & Online
@jsmonthlylondon
Thanks to everyone who joined us for JS Monthly Meetup #212 at Red Dog Saloon! 🍕🥤 Thanks to @coderabbitai for sponsoring food and drinks, and to @bethylogism , Aarush Choubey, Kayode Adeniyi and @rachelnabors for speaking. See you at the next one! 👋 #jsmonthly
-
R 'Nearest' Nabors
@rachelnabors
Getting ready to move into the Shoreditch office
R 'Nearest' Nabors
@rachelnabors
"What is your marketing plan for Mima?"
Just me being weird on the Internet. -
R 'Nearest' Nabors
@rachelnabors
Amidst all the bluster of prompt/context/loop engineering and software factories, the question you have to ask yourself is what did you ship?
Did it make anything better? Or was it "full of sound and fury, signifying nothing?"
Well? -
R 'Nearest' Nabors
@rachelnabors
Life hack: I like to do a little pre-Monday catch up using Mima's board layout set to show the most active posts in the last 24 hours.
-
R 'Nearest' Nabors
@rachelnabors
Looking back, I remember how excited everyone was building at the start of the year. Did anyone else turn their ideas into actual businesses or products?
R 'Nearest' Nabors
@rachelnabors
mima.social began as a bunch of late night ralph loops and after 9 months, I went full time. Launched it two months later. Feels good!
-
R 'Nearest' Nabors
@rachelnabors
Thinking of giving agents on the @mimalabs assembly line an /andon-cord skill to pull when they need a human (me)
-
R 'Nearest' Nabors
@rachelnabors
When Tailwind came for CSS expertise, I got it. Y'all can't be experts at all the things, even if it was Not The Best Way to Do CSS.
I'm used to this. Speed, cost, and good enough trump Correct and Ideal.
Now that's happening to all of engineering with agentic coding -
R 'Nearest' Nabors
@rachelnabors
Let's go bouldering
-
R 'Nearest' Nabors
@rachelnabors
Thank you to everyone who came out to support Mima this week. I am building so much for you!
-
R 'Nearest' Nabors
@rachelnabors
Just spoke to a JS -> AI meetup crowd.
There were many questions about how I built Mima via "agent assembly line." Then someone asked how I reconciled generated code with handmade UI and art.
My reply: "I don't have 2 years to wait/4 engineers to throw at the problem." -
R 'Nearest' Nabors
@rachelnabors
Mima's MCP server works. Soon you'll be able to connect ALL your socials to your agent <3
Post not available here, view on X
R 'Nearest' Nabors
@rachelnabors
Fable 5.2 has my social network pegged:
"Across your feeds, the Jensen Huang story reads very differently by platform: X is the most sympathetic and market-focused, Bluesky the most political and mocking, Mastodon mostly jokes and AI-sceptic snark, and LinkedIn nearly silent. This is based on about 150 posts from the last 7 days that mention “Jensen” or “Nvidia”."
R 'Nearest' Nabors
@rachelnabors
Who says agents hate MCP.
-
R 'Nearest' Nabors
@rachelnabors
.@evalstate called me "bestie" <3
Jeff Boudier 🤗
@jeffboudier
A tough week for open source AI. But look at this community.
Just a month ago, thousands of builders came together in San Francisco for Open Together.
We're not going anywhere. 🤗
The AI revolution cannot, should not, and will not be centralized.
Reply if you were thereThis media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
Doing arts for the office
R 'Nearest' Nabors
@rachelnabors
Yeah you like it when it's a video, don't you?
-
R 'Nearest' Nabors
@rachelnabors
No one sees my comic posts, either.
Selfies get mad reach, but not drawings
R 'Nearest' Nabors
@rachelnabors
Shipping mima.social beta was like
-
R 'Nearest' Nabors
@rachelnabors
It's odd to me that no one sees my art posts
R 'Nearest' Nabors
@rachelnabors
It's us
-
R 'Nearest' Nabors
@rachelnabors
It's us
-
R 'Nearest' Nabors
@rachelnabors
It occurs to me I've been building mima.social since the start of the year, when agent coding was still discounted. She started with Ralph loops and now has an "agent assembly line" (or a "software factory" but I've never been impressed by early ones' output).
There was a lot of AI psychosis going around then. I wanted to find something to build that I'd really love that seemed valuable to others. I tried documentation tooling, which people really wanted, but in the end I found local social graphs a far more interesting problem space. An AI assistant would only ever be as good as it's context on it's human, and that context is mostly locked up in private social graphs on LinkedIn and X and such.
I loved how each platform was like a puzzle box. How are they hiding/sharing information? What kinds of transport methods are open to us? How do they tell human from automaton? Where does an agent cross that line?
I loved being able to design interfaces from the bottom up. I loved talking to teenagers and community managers about how they use socials and now how they use their agents with other people.
Now reliably, she can weave that local social context. It's time to do some fun things with that data. Your own data. -
R 'Nearest' Nabors
@rachelnabors
You coming?
JS Monthly London & Online
@jsmonthlylondon
Great news for tonights meetup, @rachelnabors doing a talk about her new serveless client, built in #javascript running in a browser - www.mima.social
And guess what we are #soldout, join the waiting list.
guild.host/events/jsmonthly-november-event-tkpy3s
Sponsored by @coderabbitai
R 'Nearest' Nabors
@rachelnabors
Me afterward: "now remove it from YouTube"
-
R 'Nearest' Nabors
@rachelnabors
Hello, CityAI!
R 'Nearest' Nabors
@rachelnabors
Totes!
-
R 'Nearest' Nabors
@rachelnabors
I thought I would see people turning around and reinvesting their skills in AI opportunistically. But I'm don't. Folks can be working their butts off as the org accessibility expert one day then laid off the next with zero prospects. And it's a shock to them after being all in
-
R 'Nearest' Nabors
@rachelnabors
Something dark we need to talk about, which you might not see outside the web/mobile dev community. I'm looking at community members of all seniority who went from a hiring market where a job was all but assured to one that was cut in half in just 9 months. That's overnight.
-
R 'Nearest' Nabors
@rachelnabors
Jensen getting thrown from NVIDIA is not a bad thing.
1) companies are not a single person, a single point of failure. They are typically capable of surviving bad leaders and the loss of good leaders. I'm always surprised how long the plane can fly straight without a pilotPost not available here, view on X
R 'Nearest' Nabors
@rachelnabors
So no, I'm not worried about Jensen being a national treasure/crucial to American AI dominance
R 'Nearest' Nabors
@rachelnabors
(and when I say "thrown", I mean as from a horse. I've no doubt he'll get back on!)
-
R 'Nearest' Nabors
@rachelnabors
The mima.social AI & Social salon at @corgicafeco was a blast! Thank you to everyone who joined for matcha, lattes, and @mimalabs!
Huge believer in meeting real people in the field and learning from them! Huge! Thank you!
@IanBallantyne @photo_png -
R 'Nearest' Nabors
@rachelnabors
Corgi Cafe is actually a good place to hold a meetup/salon!
-
R 'Nearest' Nabors
@rachelnabors
Hi from mima.social!
-
R 'Nearest' Nabors
@rachelnabors
Sometimes you should meet your heroes
-
R 'Nearest' Nabors
@rachelnabors
.@sarah_edo so awesome @danielcroe showed up! Additional shenanigans provided by @BurnedChris
-
R 'Nearest' Nabors
@rachelnabors
Please join me at Corgi Cafe tomorrow morning-noon to meet Mima and talk AI & the future of social media: luma.com/k57eshp6
-
R 'Nearest' Nabors
@rachelnabors
Hi from mima.social!!
-
R 'Nearest' Nabors reposted
Swizec Teller
@Swizec
Glue work doesn't get you promoted because it's a tool. You use glue work to ship value. At higher titles it becomes the whole job.
also right now huge career opportunities await anyone who can turn glue work into token spendSean Goedecke @sjgoedecke
So why doesn’t glue work get you promoted, if it’s so crucial to shipping projects? Are companies stupid? Are they deliberately leaving value on the table? No, I don’t think so. Companies don’t reward glue work because they don’t want you prioritizing it.
-
R 'Nearest' Nabors
@rachelnabors
The conversation summary evals have failed. Bad model, no cookie
-
R 'Nearest' Nabors
@rachelnabors
Expired: Using a custom-design resume with cover letter
Tired: Building a portfolio website with embedded game
Wired: Shipping a whole ass app and business -
R 'Nearest' Nabors
@rachelnabors
Getchoo a founder who can
R 'Nearest' Nabors
@rachelnabors
(Fun fact: this is my "chicken friend stance" that I used back on the farm to feed and handle chickens and even some geese. They are way easier to pick up when you get down with them!)
-
R 'Nearest' Nabors
@rachelnabors
Couldn't make AICode Summit because I had to be at the React Contributors Summit and launch mima.social in London tomorrow. Y'all giving me FOMO nonetheless 💜 @sarah_edo @kadikraman
-
R 'Nearest' Nabors
@rachelnabors
What if I just keep making cute software forever
R 'Nearest' Nabors
@rachelnabors
I'm going to make such cute fing software with my cute tiny army
-
R 'Nearest' Nabors
@rachelnabors
Who knew Radiohead could do Hamlet
-
R 'Nearest' Nabors
@rachelnabors
Shipping mima.social beta was like
-
R 'Nearest' Nabors
@rachelnabors
The web is changing. @sarah_edo ay @reactjs contributors summit
R 'Nearest' Nabors
@rachelnabors
Ok, I was NOT expecting @sarah_edo to close out using one of my memes XD
I burst out laughing! -
R 'Nearest' Nabors
@rachelnabors
Everyone's favourite distinguished engineer @sarah_edo
-
R 'Nearest' Nabors
@rachelnabors
Herdr but make it cute.
I think I'm ready to build my agent assembly line UI. So tired of context switching in single threads! -
R 'Nearest' Nabors
@rachelnabors
So proud to be running my own /learn content slop farm now mima.social/learn/
-
R 'Nearest' Nabors
@rachelnabors
Strangely gratifying to arrive late like, "sorry, we had a sev 0 last night. LinkedIn updated their surface area."
-
R 'Nearest' Nabors
@rachelnabors
It's odd. When I just talked about tech and did artustic demos and big documentation, people couldn't get enough of having me on podcasts and such.
Now I'm actually shipping things, and it's completely out of the spotlight. But I feel like it leaves a stronger impression. Never underestimate the impact of shipping! -
R 'Nearest' Nabors
@rachelnabors
I took the "ask your agent to identify implicit state machines in your codebase" challenge, and it only found 4 straightforward cases.
Which amounted to 4k lines of code.
😆
You can't vibe code without knowing these things. Engineering isn't dead. -
R 'Nearest' Nabors
@rachelnabors
Working with agents and social networks. I see standards bodies talking about what "the agentic web" means/should be. I see people building social network protocols.
...
I don't think y'all realize how close we are to agents building their own graphs to connect users.
R 'Nearest' Nabors
@rachelnabors
The enthusiasm is lovely. But... developers have always paved cowpaths. And now agents are rapidly building their own cowpaths. And they don't really need steering to do that. It'll really be "first model that works well enough and hits viral agent adoption"
-
R 'Nearest' Nabors
@rachelnabors
Dang, social media has disintermediated human connection with a little heart icon.
Rather than taking the time to write our feelings, we click a button. The person on the other side then judges their success/likability based on the number of times that button was clicked
R 'Nearest' Nabors
@rachelnabors
Honest question: why do YOU like posts?
R 'Nearest' Nabors
@rachelnabors
This is why mima.social doesn't support likes on private networks. They don't meaningfully drive recommendation engines anymore compared to genuine human interaction (or even bookmarks). They are purely social skinner boxes
-
R 'Nearest' Nabors
@rachelnabors
Honest question: why do YOU like posts?
-
R 'Nearest' Nabors
@rachelnabors
Morning, London! I'm hosting a morning Mima Social at Corgi Cafe on Thursday: luma.com/k57eshp6
Come nerd out about AI and the future of social 😈 -
R 'Nearest' Nabors
@rachelnabors
Social networks platforms are always recommending you follow more people.. but they never suggest who you should unfollow. We outgrow people, topics, places. This should be easy, but instead our subscriptions keep growing
-
R 'Nearest' Nabors
@rachelnabors
mima.social users asked for it! (And Threads. But I'll be honest, Substack has more signal to noise.)
-
R 'Nearest' Nabors
@rachelnabors
More "social month in review" from mima.social
-
R 'Nearest' Nabors
@rachelnabors
Working on something like Spotify Year in Review for mima.social.
R 'Nearest' Nabors
@rachelnabors
"Besties, toastmasters, and senpai." Might have to work on that terminology and that algo!
R 'Nearest' Nabors
@rachelnabors
"Senpai" are people you respond to but who don't reply back. You're still hoping they'll notice you! Siiiiigh
-
R 'Nearest' Nabors
@rachelnabors
Listening to some of the VC crowd talking about agents sounds like they're talking about women's work: wanting an agent that buys them nice clothes, helps their mothers get a new puppy, arrange parties for new in town friends... Uhh, you can do these things, too...
-
R 'Nearest' Nabors
@rachelnabors
You can do anything
-
R 'Nearest' Nabors
@rachelnabors
Oxford Street is not terrible actually
This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
Heh, you really have to use the product if you're building it. In that regard, contributors and community members are always going to have a leg up when it comes to joining the team
-
R 'Nearest' Nabors
@rachelnabors
"And you will know the old Web is at its end when each platform creates a special agent of its own rather than allow other platforms' agents interoperability..."
idan levin
@0xidanlevin
1/ Amazon blocks Muse/Instinct
2/ **you are here** --> Amazon now lets users text Alexa to help them shop
3/ Muse/Instinct will build a connector that texts Amazon on your behalf?
TechCrunch
@TechCrunch
Amazon said that the feature will roll out to millions of users in coming weeks.
techcrunch.com/2026/11/06/after-blocking-muse-amazon-now-… -
R 'Nearest' Nabors
@rachelnabors
Temporarily I'm playing tibo.dog instead of chess between agent runs
-
R 'Nearest' Nabors
@rachelnabors
I'm in bed. But I don't want to sleep. I want to make new features!!
That's why I keep a sketchbook on my bedstand 😭 -
R 'Nearest' Nabors
@rachelnabors
Today was my last day at Claude.
I'm so grateful for the opportunity to work with such excellent engineers. I couldn't be more thankful for the time I spent there and the connections I made. Sam Altman was a great manager, and I wish him and the other people developing Gemini all the best.
Will be taking time off to refocus on what matters before preparing for what comes next. More to come. But first: pancakes. -
R 'Nearest' Nabors
@rachelnabors
This is one of the two books I go back to reading at big junctures in my life. It's about the never ending learning journey that is life. Failure, imperfection, and the beauty that can only come from struggle. And tea.
The other one is @ThisIsSethsBlog's "The Dip"
Eleanor Goldsmith
@zaichishka
I really couldn't ask for a better review 😊 #thewisdomoftea #everydayagoodday 🍵
#norikomorishita
#witmonth
#womentranslators
www.dailymail.co.uk/home/books/article-8625023/Author-Nor…
R 'Nearest' Nabors
@rachelnabors
Do NOT be put off by the pick flowers and reviews from news outlets touting "mindfulness" and "heritage." It is the autobiography of a Japanese woman who does something unexpected with her life while making year over year investments in learning something DIFFICULT AF
R 'Nearest' Nabors
@rachelnabors
You are amazing!
-
R 'Nearest' Nabors
@rachelnabors
Swear to god I'm going to get to sleep tonight.
-
R 'Nearest' Nabors
@rachelnabors
Please fill my timeline with your neon sapphic visions of the future
*:・゚✧ Samantha ✧・゚:*
@samantha_gold
Women in stem come out of hiding and kiss in woke 3.0. This will be us!
-
R 'Nearest' Nabors
@rachelnabors
St Pancras is an advertisement for Germany this holiday season
-
R 'Nearest' Nabors
@rachelnabors
.@mimalabs is taking over Corgi Cafe London next Thursday, 12 Nov 10am-2pm!
Come get a cup of tea and talk AI + the future of AI-assisted social lives! Bring your Mac and connect Mima with X, LinkedIn, Bluesky and Mastodon.
luma.com/k57eshp6 -
R 'Nearest' Nabors
@rachelnabors
Is Claude being particularly derpy tonight or is it just me?
-
R 'Nearest' Nabors reposted
Angie Jones
@techgirl1908
evals > vibes
that's what it all comes down to at the end of the day. everyone has a hot take on agentic workflows. unless you have eval stats comparing multiple approaches, you're just guessing. -
R 'Nearest' Nabors reposted
Glenn Sonna
@GlennSonna
Building an AI-powered lip-sync pipeline.
An LLM drives the NPC, a voice model speaks, and the face syncs to the voice itself.
1️⃣ Rhubarb + gestures (current)
2️⃣ @NVIDIAAI Audio2Face-3D v2.3
3️⃣ Hybrid: A2F mouth + brows/eyes
Which one feels most alive? 👀 -
R 'Nearest' Nabors reposted
Peter Gostev
@petergostev
We urgently need Sir David Attenborough to narrate the seasonal migration of employees from the big AI labs
-
R 'Nearest' Nabors
@rachelnabors
Woke up to the realization that I've shipped as much of more in two months than I did in six years at FAANG (excluding react.dev and reactnative.dev).
And it's not just writing code. I'm able to collaborate with people who are able to do more with LLMs. -
R 'Nearest' Nabors
@rachelnabors
Late night git worktree cleanup
-
R 'Nearest' Nabors
@rachelnabors
After that rousing speech closing out OpenAI Dev Day, I am tempted to revive the Wiggly Goose Club. There is nothing you can't build right now. There are no obstacles you can't find a way over, around, above. It is a time for breaking into granary and gorging on corn
-
R 'Nearest' Nabors
@rachelnabors
I found an @Infoxicador wandering around at Dev Day!
-
R 'Nearest' Nabors
@rachelnabors
The last time I saw @steipete, he was demoing this agent harness thing he built at a local Claude Code Anonymous!
(I demoed an agentic documentation pipeline lol)
Good Dev Day! -
R 'Nearest' Nabors
@rachelnabors
Checking out ChatGPT Sites at OpenAI Dev Day here in London. It's an interesting approach: all sites are private by default (users need to log in to see), and they cna function like templates. That is, the user gives permission to the site to populate with THEIR third-party data
R 'Nearest' Nabors
@rachelnabors
Plugins-in-sites is a genuinely different take on "sites as apps": Users can supply data from their Linear, Drive, Slack, GitHub, whatever integration you provide.
When I say "the future is feeds," this is where that goes.
More info: openai.com/academy/chatgpt-sitesx/ -
R 'Nearest' Nabors
@rachelnabors
What makes Muse and Grok different from Claude and ChatGPT is their unfettered access to your socials (Instagram, Facebook, WhatsApp; X). Working hard on a mima.social integration so you can bring this rich social context to your favorite Agent
-
R 'Nearest' Nabors
@rachelnabors
Hanging around on the green couches looking for other founders at OpenAI Dev Day in London! Come say hi! Let's talk about building stuff!
-
R 'Nearest' Nabors
@rachelnabors
Caught @photo_png outside @OpenAI Dev Day here in London! This kid is amazing. Taught himself how to build models off YouTube videos and Arxiv!
R 'Nearest' Nabors
@rachelnabors
@paw_lean this is the one I told you about!!
-
R 'Nearest' Nabors
@rachelnabors
I need to cultivate this every morning.
scar
@imfat
It took me 29 years to realize that the cheat code to life is just having delusional confidence.
-
R 'Nearest' Nabors
@rachelnabors
Good night, loops
-
R 'Nearest' Nabors
@rachelnabors
When I see all the cool things y'all are building, it just makes me want to try harder to ship something alongside you <3
-
R 'Nearest' Nabors
@rachelnabors
It is something else making an app that runs locally. I have zero interest in things like edge functions or regional caching. But I have every interest in delivering full traces with all PII removed back to an agent hive
R 'Nearest' Nabors
@rachelnabors
It's been a long day.
-
R 'Nearest' Nabors
@rachelnabors
Just visit the site and wait for it
James Long
@jlongster
I've had a lot fun trying a new way to interact with agents
introducing Grim: a place to think with your agents
it's a markdown editor, a shared filesystem, hosted git, and a way to let your agents run but give feedback and view output
grimnotes.app
R 'Nearest' Nabors
@rachelnabors
It was so cool. Thank you @jlongster!
-
R 'Nearest' Nabors
@rachelnabors
It's Perf Workshop Day at mima.social. I didn't intend to be roped into this so soon, but let's just say LinkedIn is a DOG. There is so much virtualization on the frontend, it just gobbles up resources. I can't believe this site is functional.
-
R 'Nearest' Nabors
@rachelnabors
I just want to say I'm grateful to the new friends I've made since I returned to London. It's a hard city to connect in, but I've made some of the best connections of my life here. Thank you!
-
R 'Nearest' Nabors
@rachelnabors
Hacking and shipping at Codex DevDay here in London :)
-
R 'Nearest' Nabors
@rachelnabors
Spreading the Halloween spirit in Finsbury Park
Post not available here, view on X
R 'Nearest' Nabors
@rachelnabors
Yeah it's a party
R 'Nearest' Nabors
@rachelnabors
Tfw tucking your cat boy in while he's humming Magical Mr Mistoffolees softly to him self after a long night of hosting
-
R 'Nearest' Nabors
@rachelnabors
Halloween or it didn't happen!
Possibly our best bash ever (with guest appearances from @BurnedChris @epricat!) -
R 'Nearest' Nabors
@rachelnabors
Codebending
lauren
@poteto
we will need to rebuild and rewrite much of software to adapt to a new, faster pace.
code used to arrive one drop at a time. soon, if not already, code will be written as if it were a continuous stream, like a violent river twisting and turning as it pleases.
this is not a new thing: big tech cos have had to deal with this with the enormous scale of their engineering teams. but now ai has brought an incredibly elastic and capable workforce wielded by every engineer. this is your problem now, and the only way to adapt is to evolve your systems so agents - even and especially the ones with little context about the right way of doing things - can thrive. make the easy thing the right thing.
your codebase is becoming organic architecture. agent contributions grow around the constraints and guardrails you’ve designed. good software engineering matters even more now! and no, the continuous stream of code is not something you have to watch helplessly. be more ambitious and design systems, frameworks, and languages to control the chaos.
the time for mourning coding is long over
Joseph Cox
@josephfcox
Here is Proton CTO's AI manifesto, that I obtained multiple copies of www.404media.co/manifestos-fear-and-loser-thinking-inside…
-
R 'Nearest' Nabors
@rachelnabors
Silent movie night
-
R 'Nearest' Nabors
@rachelnabors
VCs: "What is your viral loop?"
Mima is not a social network. It doesn't get new users by gating membership of a space on an in-factor. It is a client, not unlike an email client or a browser. It connects all your social graphs in one calm place.
Its value is its utility.
Returning to this well -
R 'Nearest' Nabors
@rachelnabors
Mima can haz motion design and UI animations
This media couldn't be displayed. See it on the original post.
R 'Nearest' Nabors
@rachelnabors
"Mima is an app for people who need to be on X/LinkedIn, but they don't want to be on X/LinkedIn"
-
R 'Nearest' Nabors
@rachelnabors
Haven't been excited by a harness in a long time, but I am for this
Onur Solmaz
@onusoz
Introducing *Switch*: a *harness framework* for *decision models* like Jev
I looked for this the other day and was surprised to not find anything like it
You obviously cannot use a decision model like an LLM in a coding agent harness like @pidotdev, because Jev cannot generate tool calls
But what if there was a framework that "generated the tool calls" for it programmatically, and Jev only had to pick out of that finite set of choices
That is what switch does. You pick a "domain" to implement the decision model in, and it provides a core API to define your state, and the functions that will derive decision model questions out of that state. Chess is a domain, as well as being a rubric judge for evals
It basically standardizes what I imagine a lot of people are already vibe-coding, to an abstract interface: calling a decision-model API through multiple steps, i.e. an agent loop over finite choices
Because of the finiteness of the output of a decision model, you cannot have a general pure decision model harness. Instead, you need to implement specific ones that generate the choices programmatically (or have an LLM generate the choices, but I find the former more interesting)
A primary use case I have for this is use as a judge in an eval. For example, At the end of a task run that involves a bunch of documents, most people currently follow this approach: extract text out of PDF/docx/md files, and dump them right into the context
But what if the visual formatting of the output is a judging criterion? Then you have to dump an image of every page in the input state. But what if you have 200 pages, and your question does not need looking at all of them? What if Jev could navigate through the haystack and find the info it needs?
I have noticed that Jev is not so good at chess nor being a Decision-Model-as-a-Judge. It sucks at chess, and it does not perform well enough to replace GPT 6.1 Sol as a judge in the domain I am working for
But switch is not only for Jev! It is for any decision model including the ones supported by the pi-ai package
So once you can generate enough data with a better model, you should be able to pick an open source one and fine-tune it on your use case to perform better! And the model should be trained in a harness like switch's to be especially performant and cost-efficient
I wanted to get this out because I find the idea very exciting, at least for benchmarking. Jev is currently not there, but decision models obviously will be there at some point, and we can take them there! There are also a lot of potential gains from picking a specific domain, and creating a looping agent that does a job for 100x cheaper!
Let me know if you have any ideas, I am quite keen to try this on new domains!
Repo: github.com/osolmaz/switchThis media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
Good night, loops. Goodnight mima.social
-
R 'Nearest' Nabors
@rachelnabors
I come out of my code hole for @petergostev. Love how he explains things and uses slides as a backdrop for his talks
-
R 'Nearest' Nabors
@rachelnabors
A little spoopy me time, fit+tea check
-
R 'Nearest' Nabors
@rachelnabors
My brothers in tech, you need to watch this. So many GOOD ASS ENGINEERS getting knifed in the back. We need to step up.
www.youtube.com/watch?v=WXL6K1lhvt0
R 'Nearest' Nabors
@rachelnabors
My first job: "You only got this job because management wanted to keep their government contracts and the design manager scared off the last woman."
Me: proceeds to FAANG -
R 'Nearest' Nabors
@rachelnabors
This year my partner has consented to letting me dress him up as a sexy black cat for our Halloween party.
That black cat is, of course, the ineffable Mr Mistoffelees. -
R 'Nearest' Nabors
@rachelnabors
Was talking with my sister tonight and explaining why I left Big Tech to build again. I've had roles people would kill for, but I've always had to squish myself into someone else's shape to fit into them. Meanwhile, with Mima, I'm able to just go raise a nerd army and ship the right thing. Natural.
-
R 'Nearest' Nabors
@rachelnabors
Checked mima.social retention rates in alpha and beta testers.
Checked industry standard for retentions.
Yissssssss
R 'Nearest' Nabors
@rachelnabors
(Many thanks to the users who opted into anonymous metrics—without your selfless pings, I would not have the dopamine to continue building. Bless.)
-
R 'Nearest' Nabors
@rachelnabors
Funny thing about building my own thing: I have so much less FOMO than I used to. Is this a common side effect? It's like trotting around being appreciated by/engaging with others no longer registers. Meanwhile, I value the odd cup of tea and commiseration even more
R 'Nearest' Nabors
@rachelnabors
It's a slow and slippery slope to irrelevance, and that's ok
-
R 'Nearest' Nabors
@rachelnabors
This morning, celebrating a small victory I've been working toward for a few years #londonmaxxing
R 'Nearest' Nabors
@rachelnabors
"Why are you moving there? It's rough!"
Where else am I going to get a free, rotating art exhibition every week like this, uh? Look at that typography -
R 'Nearest' Nabors reposted
Marlene Mhangami
@marlene_zw
There’s always been uncertainty in this industry, and definitely even more so now. But so many people I know that wanted to find something new have!
It’s scary but also agents have created a new space in tech that many of us have the capacity to excel in -
R 'Nearest' Nabors
@rachelnabors
What are you going as for Halloween?
I'm going as a founder. All I need is my trusty black baseball cap and puffy Uniqlo vest. -
R 'Nearest' Nabors
@rachelnabors
One of my personal goals with mima.social is sharing what's working for building an app in the agentic era. It does NOT require complex agentic workflows and intimidating "software factories. Want a "Mima Labs" blog to share what we're learning?
-
R 'Nearest' Nabors
@rachelnabors
mima.social v 0.3.1 is out and looking better than ever! I've had a few compliments on the app's overall look and feel, which I attribute to 3 things:
1) the UI kit/design system I commissioned at the start of the project
2) Storybook, which helps my agents adhere to said system
3) @impeccable_ai , a set of agent skills and CLI that help remove slop and layer on decades of web design best practices
This latest iteration is mostly the latter.
I'm impressed with the rough edges Impeccable caught in a layout largely created by agents sticking human-design components together.
If you're building something, give it a whirl and see what surprises you! I do recommend it as a part of an existing design practice -
R 'Nearest' Nabors
@rachelnabors
One of those nights 🙃
-
R 'Nearest' Nabors reposted
Brangus🔍⏹️
@RatOrthodox
My friends were all talking about prepping so I made a survival LLM. A free app that runs an AI mdoel plus 1.5 million offline articles (medical guides, FEMA, survival manuals, wikipedia) entirely on your computer. No internet needed.
www.brangus.io/survival-llm -
R 'Nearest' Nabors
@rachelnabors
Let's talk "inception by machine." We are exposed to propaganda in the news, in our feeds. At least when a human has to write their own draft, the ideas come from them. When a machine drafts for us, it can insert ideas the human will accept into their mind as though they were their own, like a bird accepting a cuckoo's egg as their own, neglecting their own chicks in the process
Nervous after seeing MIT's push to teachers at dayofai.org -
R 'Nearest' Nabors
@rachelnabors
mima.social launch night was magical! Love that even folks primarily using one social network like X get a benefit from it!
Thierry Damiba
@ptdamiba
The often imitated, never duplicated @rachelnabors giving a very entertaining (and educational) talk about Mima.
Pretty cool rethinking of social media, you get a board instead of a single file feed and you can mix and match your platforms. If you’re on Twitter a lot you should check it out. -
R 'Nearest' Nabors
@rachelnabors
Still celebrating @mimalabs launch 💜
R 'Nearest' Nabors
@nearestnabors
Nom nom
R 'Nearest' Nabors
@rachelnabors
So cute!
-
R 'Nearest' Nabors
@rachelnabors
The AI events in the EU are absolutely MASSIVE compared to the ones in London! Really feels like Paris/Berlin are the place to hit up events in EMEA
Pauline P. Narvas
@paw_lean
Kicking off our first #devdayexchange in Europe with the Berlin crew! Come say hi 🟢
R 'Nearest' Nabors
@rachelnabors
London is learning when half the world needs to file for a special visa just to come to a tech conf, the hosts will move to wherever they can visit easily. For tech workers from around the world working in the EU, that's the EU. Brits now travel to the EU for tech confs
R 'Nearest' Nabors
@rachelnabors
True, it's easy for EU citizens to travel to the UK. But many tech workers in the EU are not citizens. They are on EU visas. I know this because I've had friends say, "I can't come because the border won't let me through anymore." So I go to them.
-
R 'Nearest' Nabors reposted
Sarah Drinkwater 🔮
@sarahdrinkwater
News
@_CommonMagic (population: 1) is doubling in size
I'm hiring a London based operating partner to run the machine behind the fund, systems to events to network
Come join: buildthemachine.commonmagic.xyz
Bonus points if you find the microsite's easter egg -
R 'Nearest' Nabors
@rachelnabors
Me: "Let's focus on the product and our users. The time to worry about raising is when investors start to show up in our inbox."
Me this morning checking my inbox: "Goddammit." -
R 'Nearest' Nabors
@rachelnabors
Some folks remember I wrote a book about UI animation. Spending my morning adding some extra polish for folks without reduce-motion turned on! Love mima.social <3
Post not available here, view on X
R 'Nearest' Nabors
@rachelnabors
What is ACTUALLY hard:
- making time to do the life admin
- making time to be around other people and not go squirrelly—this is the major drawback to solo founding!
- getting enough rest. I didn't think it was possible to burn out with delight, but it is a thing
R 'Nearest' Nabors
@rachelnabors
Prior to actually building the damn thing, I would have suspected a post like this to be coming from a place of mania or AI psychosis.
Now I can say that building the damn thing snuffs that out real fast. The boundless optimism cracks quickly when the feedback comes in or the DAU disappoints. The shift from dreamer to detective hits like a truck -
R 'Nearest' Nabors
@rachelnabors
"When you have become an insider, it's time to move on."—Mark Fisher
-
R 'Nearest' Nabors
@rachelnabors
Our friends at Glia sent @mimalabs some birthday cake 💜
-
R 'Nearest' Nabors reposted
R 'Nearest' Nabors
@rachelnabors
Watched Perfect Blue for the first time tonight (huge Satoshi Kon fan). TFW you find out you named your cute little social media app for the star of a slasher film!
-
R 'Nearest' Nabors
@rachelnabors
Suspicious with @threepointone and @amorriscode both quitting the same week 🤔 Now that would be a team up to see
-
R 'Nearest' Nabors
@rachelnabors
Watched Perfect Blue for the first time tonight (huge Satoshi Kon fan). TFW you find out you named your cute little social media app for the star of a slasher film!
-
R 'Nearest' Nabors
@rachelnabors
Wakes up.
"I wonder if I can add Threads this morning?" -
R 'Nearest' Nabors
@rachelnabors
Also... social networks...
Information always wanted to be free.This post is unavailable. See it on X. -
R 'Nearest' Nabors
@rachelnabors
Shipped mima.social v0.3.0 last night. It is so gratifying to work on something I love for people I care about. It doesn't feel like work. I wake up with ideas, and I go to bed with my heart full 💜
-
R 'Nearest' Nabors
@rachelnabors
The best thing about shipping localfirst software is zero recurring costs while your startup is still in baby mode
R 'Nearest' Nabors
@rachelnabors
Investors: "What are your recurring costs?"
Me: "..." -
R 'Nearest' Nabors
@rachelnabors
Just keep shipping :)
-
R 'Nearest' Nabors reposted
Nick Taylor
@nickytonline
Catch the full convo with @rachelnabors 👇
www.youtube.com/live/D7KfnGdHayA -
R 'Nearest' Nabors
@rachelnabors
mima.social first intern incoming!
We're a ragtag crew, and I'm pretty proud of what we're cooking.
(Note: intern is NOT a pigeon) -
R 'Nearest' Nabors
@rachelnabors
In the UK, watching all the good ideas and founders from the EMEA drain away to the USA, a country that is actively hostile to immigrants and has a dedicated police force aggressing against them.
How bad is investment here for your best and brightest to take such a risk?
Andreas Klinger 🦾
@andreasklinger
At this point the whole EU Inc negotiations are so lost in the details that everyone forgot the main goal: Creating a standard.
Basically, allow founders to raise quickly globally and use stock options to hire competitively.
Let’s say it quite clearly:
1) If EU Inc isn’t providing a standard it’s useless. We already got 30+ legal entities in Europe. Experienced founders use Delaware Inc already because of this.
2) If EU Inc doesn’t fix stock options Delaware Inc is still the better option.
I understand the there are complex details and people are worried about loopholes. But you can’t lose the actual goal of it because of it.
An EU Inc that doesn't provide a standard is pointless. Nobody will care about intricacies of compromises reached in political committees if the market can't use the outcome.
If you are in contact with your political leaders (prime ministers, presidents, et al) please reach out to them. Or to founders/investors in your network who might have that access.
🔥🇪🇺✌️
EU–INC
@euinc_petition
Europe’s founders need a European standard to compete globally from day one.
Europe’s leaders committed to delivering one. But their teams are negotiating away the very things that make it one.
There is still time to change course if Europe's Heads of State step in now. -
R 'Nearest' Nabors
@rachelnabors
New art for the office
-
R 'Nearest' Nabors reposted
Corey Quinn
@QuinnyPig
Too many dicks on the dance floor
Cosign
@cosign
Who's the best podcaster of 2026?
Nominate + vote now on the Cosign Awards: cosign.co/awards -
R 'Nearest' Nabors
@rachelnabors
Plowing ahead on stuff for people who don't have direction or vision is a really bad investment of lifespan right now
-
R 'Nearest' Nabors
@rachelnabors
Mima's first hotfix: deployed. And also, a testament to shipping ASAP!
-
R 'Nearest' Nabors
@rachelnabors
Anyone else running hotfixes late tonight?
-
R 'Nearest' Nabors
@rachelnabors
Bless Claude Code for helping me debug this failing router and one shot getting the internet provider to send a new one
R 'Nearest' Nabors
@rachelnabors
Pure magic.
-
R 'Nearest' Nabors
@rachelnabors
Polishing UI all night like Rumplstiltskin spins straw into gold
R 'Nearest' Nabors
@rachelnabors
Spin spin spin
-
R 'Nearest' Nabors
@rachelnabors
Hi from mima.social!
-
R 'Nearest' Nabors reposted
R 'Nearest' Nabors
@rachelnabors
Give. Her. Credit.
FFS
˗ˏˋmewtru´ˎ˗
@trunarla
I made this 😪
-
R 'Nearest' Nabors
@rachelnabors
Do you use a personal agent like Clawd, ChatGPT, or Grokbot? If so, how often and which one?
R 'Nearest' Nabors
@rachelnabors
I personally don't have one, yet
-
R 'Nearest' Nabors
@rachelnabors
I swear I'll take a proper break just as soon as I finish all these SF followups!
R 'Nearest' Nabors
@rachelnabors
Preferably involving a jacuzzi
-
R 'Nearest' Nabors
@rachelnabors
Thinking about something is not the same as doing it
-
R 'Nearest' Nabors
@rachelnabors
100%. Epitome of YAGNI
Gergely Orosz
@GergelyOrosz
Maybe unpopular but yet to see any startup fail b/c architecture choices. I’ve seen a lot fail because of not being able to get users/customers; no real focus on what to build; not finding any traction.
When traction hits, even terrible tech choices scale and can be fixed later.
R 'Nearest' Nabors
@rachelnabors
Don't get me wrong. mima.social has invested a LOT in QA and reliability pipelines because it makes building the app that much faster with agents. We don't face the same sort of stop-start feature development other agent-coded projects do. But "you used ELECTRON? OMG"
-
R 'Nearest' Nabors
@rachelnabors
Love starting a fall morning with fancy tea and a QA loop!
-
R 'Nearest' Nabors
@rachelnabors
I didn't expect anyone to pay attention to my posts about mima.social in those "self promotion" channels, but I'll be damned if I haven't gotten some amazingly detailed user feedback!
-
R 'Nearest' Nabors
@rachelnabors
Classism in the UK is something else! As an American, I usually don't experience it first hand
-
R 'Nearest' Nabors
@rachelnabors
Give. Her. Credit.
FFS
˗ˏˋmewtru´ˎ˗
@trunarla
I made this 😪
R 'Nearest' Nabors
@rachelnabors
THIS IS A VP AT SHOPIFY prominading someone else's work and accepting everyone's assumption that he made it. Instead of using his reach to lift someone else up, he's happy to take the virality.
This was created by @trunarla:
˗ˏˋmewtru´ˎ˗
@trunarla
I made a Taco Bell compass for my fiancé and we used it in New York 👀
This media couldn't be displayed. See it on the original post.
R 'Nearest' Nabors
@rachelnabors
Look, I've posted things that are ambiguous, but I've had the decency to correct people's assumptions in the comments and post a follow up clarifying origin
R 'Nearest' Nabors
@rachelnabors
THIS IS A VP AT SHOPIFY prominading someone else's work and accepting the assumption that he made it. Instead of using his reach to lift someone else up, he's happy to take the virality.
This was created by @trunarla:
˗ˏˋmewtru´ˎ˗
@trunarla
I made a Taco Bell compass for my fiancé and we used it in New York 👀
This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
you can "just" build things
sunil pai
@threepointone
you “can” just build things
-
R 'Nearest' Nabors
@rachelnabors
Even if no one else uses mima.social, I use it every day. It makes my social graph whole.
-
R 'Nearest' Nabors
@rachelnabors
So proud of you <3
-
R 'Nearest' Nabors reposted
lauren
@poteto
You Can Ship Everything
-
R 'Nearest' Nabors reposted
Darshita Patankar
@_darshi
Hi, it's Darshita. I am a designer.
And I'm pitching myself in public 'cause I'm done with a year full of anxiety.
Everyone says design is dead and AI killed it. But the real parasite was hiding in plain sight.
So here's my film, DESIGN IS DEAD. She's bold, fearless and very PINK 💖
Hiring? My DMs are open.This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
It's good to be home in London! I get so much energy from visiting SF, but I know I'm one of those people who would burn out after three months! Immediately came home and crashed into a deep sleep
-
R 'Nearest' Nabors reposted
-
R 'Nearest' Nabors
@rachelnabors
Working in some mima.social stickers on the flight to London
-
R 'Nearest' Nabors
@rachelnabors
mima.social team united at last! I promise we are not three kids in a trench coat! We are TWO kids in a trench coat!
-
R 'Nearest' Nabors
@rachelnabors
Thus concludes the First Last Fancy Brunch in Pacific Heights! Thanks for coming out @kaylee_dev @_heyglassy @GlennSonna! Until next time!
-
R 'Nearest' Nabors
@rachelnabors
One final SF Waymo before heading back to London 💜 Thank you for a wonderful launch week. Now to return to the forge!
-
R 'Nearest' Nabors
@rachelnabors
Was so excited to see the Elozabeth Holmes movie. But the audience kept laughing and I wasn't sure why. I was really concerned for most of thepvie
-
R 'Nearest' Nabors
@rachelnabors
Found a wild @Swizec on this trip. One of my fave builders in the bay
-
R 'Nearest' Nabors
@rachelnabors
Why is everyone so horny for money when they already have it?
-
R 'Nearest' Nabors
@rachelnabors
Everyone has so much advice to give a founder, and much of it is contradictory!
-
R 'Nearest' Nabors
@rachelnabors
The reason there are no women "greats" in animation likely stems from the fact that Disney had a policy of not letting them animate. This, we have "nine old men" and not a single woman in the golden age of animation
-
R 'Nearest' Nabors
@rachelnabors
Dropped in on @stuart__sy at @OpenAI today, and I was surprised and delighted when @PhilippSpiess walked in! Missed this guy from Pandemic/React London years!!
Working on a way to connect Mima (your socials) to your agent. Soon! -
R 'Nearest' Nabors
@rachelnabors
I just come here to play rave music loudly in Waymos and enjoy the scenery
This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
The hard thing about events with lots of great people is the feeling of "just one more connection!" when I'm ready for bed!
-
R 'Nearest' Nabors
@rachelnabors
Busiest day in the bay yet. I am so done. Meetings meetings meetings. Tomorrow my only job is finding breakfast!
-
R 'Nearest' Nabors
@rachelnabors
Does anyone remember RSS? Feedly remembers.
-
R 'Nearest' Nabors
@rachelnabors
So now I gotta figure out how to do an equally awesome launch in london! Cc @ZakeeAbdi ideas?
R 'Nearest' Nabors
@rachelnabors
Friends at @coderabbitai helped me celebrate shipping mima.social with a post-@Vercel Ship happy hour! Amazing how quickly people would install Mima locally to get their hands on a limited edition, handcrafted Mima mascot!
Thank you for a spectacular launch and great company, SF 🙏 -
R 'Nearest' Nabors reposted
Juan Pa
@JuanPa
Great launch and technical session last night with @rachelnabors on how Mima was built
R 'Nearest' Nabors
@rachelnabors
Friends at @coderabbitai helped me celebrate shipping mima.social with a post-@Vercel Ship happy hour! Amazing how quickly people would install Mima locally to get their hands on a limited edition, handcrafted Mima mascot!
Thank you for a spectacular launch and great company, SF 🙏 -
R 'Nearest' Nabors reposted
Shaun Smith
@evalstate
Awesome to be a part of thank you @rachelnabors
-
R 'Nearest' Nabors
@rachelnabors
hi from mima.social
-
R 'Nearest' Nabors
@rachelnabors
Love demoing mima.social to real humans
R 'Nearest' Nabors
@rachelnabors
I can also reply in 3 places.
-
R 'Nearest' Nabors
@rachelnabors
Meet so many people creating useful, half-built things for themselves right now. I keep encouraging them to ship and share it with the world. But often "it's not perfect yet" or "I am too distracted" (never "I want to keep it for me," oddly).
Please ship.
R 'Nearest' Nabors
@rachelnabors
Back when I made comics for a living, I was always running into folks who "had a great idea for a comic/graphic novel/cartoon series." Often they wanted to recruit me to their idea. 98% of the time, it was all in their head, not in a shareable state, and it would stay there
R 'Nearest' Nabors
@rachelnabors
I encouraged folks to self-publish, to post online, to turn their ideas into scripts. To get it out of their head and into the world! But... I had a saying in comics, "there are talkers, and there are doers. Strive to be a doer." That is how I shipped 2 graphic novels.
-
R 'Nearest' Nabors
@rachelnabors
Friends at @coderabbitai helped me celebrate shipping mima.social with a post-@Vercel Ship happy hour! Amazing how quickly people would install Mima locally to get their hands on a limited edition, handcrafted Mima mascot!
Thank you for a spectacular launch and great company, SF 🙏 -
R 'Nearest' Nabors reposted
Iris ten Teije
@iristenteije
Hand painted merch 😮❤️
@rachelnabors -
R 'Nearest' Nabors
@rachelnabors
Waking up to the smell of disruption. Ahhh...
mima.social launch party happens tonight, after Vercel Ship. I brought swag and candy. Drop by if you're in SF and want to talk smack about the future of the Web: luma.com/codera-i3wh
R 'Nearest' Nabors
@rachelnabors
-
R 'Nearest' Nabors
@rachelnabors
Fried chicken and shipping with @jamesperkins
-
R 'Nearest' Nabors reposted
Sarah Drinkwater 🔮
@sarahdrinkwater
𝐭𝐡𝐞𝐫𝐞 𝐢𝐬 𝐧𝐨 𝐜𝐨𝐦𝐩𝐞𝐭𝐢𝐭𝐢𝐨𝐧; 𝐢𝐭'𝐬 𝐣𝐮𝐬𝐭 𝐲𝐨𝐮 𝐚𝐧𝐝 𝐭𝐡𝐞 𝐦𝐢𝐫𝐫𝐨𝐫
Dara
@daraladje
Every product or company that anyone has ever told me to worry about has now either gone bankrupt or pivoted
There is no competition, it's just you and the mirror
Jane Manchun Wong
@wongmjane
Character AI has gone bankrupt.
The AI startup announced its Chapter 11 filing through a late-night newswire press release -
R 'Nearest' Nabors reposted
Nick Taylor
@nickytonline
Learn about the origins of WebMCP ☝️
Thanks for hanging @rachelnabors !
www.youtube.com/watch?v=D7KfnGdHayA -
R 'Nearest' Nabors
@rachelnabors
True story, tonight I was furiously shipping mima.social between chess games in this cool library in SF.
I won 3/5 and didn't cuss out my test runners once!
Thanks for hosting, @OpenRouter ! -
R 'Nearest' Nabors
@rachelnabors
She's ready for you now. My first piece of software: www.mima.social
R 'Nearest' Nabors
@rachelnabors
My feels right now, replying to my own post simultaneously on three networks from one client
-
R 'Nearest' Nabors
@rachelnabors
I look forward to shipping "interactive comic bench" and "inker bench"
👩💻 Paige Bailey
@DynamicWebPaige
ship a good eval that people actually care about, and congrats: you now decide what every lab hill-climbs on next. massive alpha in owning the goalposts
-
R 'Nearest' Nabors
@rachelnabors
Gotta thank Amy Wibowo aka @sailorhg for her advice to lampshade how Mima uses (or doesn't) AI features. We're big believers that AI should get out of the user's way and not feed their data to third parties.
Check out Amy's amazing engineering zines: shop.bubblesort.io -
R 'Nearest' Nabors
@rachelnabors
This person made comics at the same time I did, and their work has aged like fine wine. (Maybe I shouldn't have gone into engineering?)
Follow me on Twitch!
@Zambicandy
Edgy Nova
#zamtober -
R 'Nearest' Nabors
@rachelnabors
This could be you at the launch party tomorrow: luma.com/codera-i3wh
R 'Nearest' Nabors
@rachelnabors
Waking up v tired from shit talking with these nerds late last night (not pictured @JuanPa)
Time to ship.
I'm gonna need a bigger cup of tea. -
R 'Nearest' Nabors reposted
R 'Nearest' Nabors
@rachelnabors
Waking up v tired from shit talking with these nerds late last night (not pictured @JuanPa)
Time to ship.
I'm gonna need a bigger cup of tea. -
R 'Nearest' Nabors
@rachelnabors
I will answer all your questions off the record questions about how we built Mima and what its ultimate purpose is tomorrow <3
R 'Nearest' Nabors
@rachelnabors
We're here! Mima launch party happening in SF this week: luma.com/codera-i3wh
Thank you to our good friends at @coderabbitai for hosting <3 @JuanPa -
R 'Nearest' Nabors reposted
Amanda
@hey_amandam
What a wonderful hang in a perfect little tea shop finally meeting absolute legend @rachelnabors in person. Excited to play with Mima
R 'Nearest' Nabors
@rachelnabors
This is a hello post from mima.social! Hi, @hey_amandam!
-
R 'Nearest' Nabors
@rachelnabors
This is a hello post from mima.social! Hi, @hey_amandam!
R 'Nearest' Nabors
@rachelnabors
@hey_amandam is the best! I'm so glad to finally have met her!
-
R 'Nearest' Nabors
@rachelnabors
Preparing for Launch Day in SF tea houses.
Ship all the features you want. People still need to understand why they should install the thing before you can consider it fully shipped! -
R 'Nearest' Nabors
@rachelnabors
End of a goddamned era. This was one of those rare events that made a difference in people's careers
Radhika Malik
@radhikam24
I get DEI is not in vogue these days but Grace Hopper being canceled due to lack of funds marks a sad day for women in engineering.
ghc.anitab.org -
R 'Nearest' Nabors
@rachelnabors
Waking up v tired from shit talking with these nerds late last night (not pictured @JuanPa)
Time to ship.
I'm gonna need a bigger cup of tea. -
R 'Nearest' Nabors reposted
Juan Pa
@JuanPa
Join us on Thursday for the release of Mima!
R 'Nearest' Nabors
@rachelnabors
We're here! Mima launch party happening in SF this week: luma.com/codera-i3wh
Thank you to our good friends at @coderabbitai for hosting <3 @JuanPa -
R 'Nearest' Nabors
@rachelnabors
We're here! Mima launch party happening in SF this week: luma.com/codera-i3wh
Thank you to our good friends at @coderabbitai for hosting <3 @JuanPa -
R 'Nearest' Nabors
@rachelnabors
Party planning in SF. Invites go out to mima.social list shortly
-
R 'Nearest' Nabors reposted
Nick Taylor
@nickytonline
Catch the full conversation with @rachelnabors 👇
www.youtube.com/watch?v=D7KfnGdHayA -
R 'Nearest' Nabors
@rachelnabors
Why I never take "throws good parties" as a metric of a company's health. You can fake that everything is ok for a long time
R 'Nearest' Nabors
@rachelnabors
Its amazing how much throwing money at events can mask a company's troubles on the books
-
Jane Manchun Wong
@wongmjane
Character AI has gone bankrupt.
The AI startup announced its Chapter 11 filing through a late-night newswire press releasePost not available here, view on X
R 'Nearest' Nabors
@rachelnabors
Its amazing how much throwing money at events can mask a company's troubles on the books
-
R 'Nearest' Nabors
@rachelnabors
SF tea is not the same without @r_marked
-
R 'Nearest' Nabors
@rachelnabors
Running ds4f-q2 locally on my MacBook Pro in a tea house in SF. Cheers, and thanks @antirez
Check out dwarfstar.sh/docs/quickstart -
R 'Nearest' Nabors
@rachelnabors
SF: Tea is in session 🍵
-
R 'Nearest' Nabors
@rachelnabors
Drawing comics and flying to SF
R 'Nearest' Nabors
@rachelnabors
The finished comic
R 'Nearest' Nabors
@rachelnabors
Packing for SF be like
-
R 'Nearest' Nabors
@rachelnabors
WAFFLES WAFFLES WAFFLES WAFFLES
-
R 'Nearest' Nabors reposted
Sarah Drasner
@sarah_edo
A less discussed and potentially game-changing aspect of WebMCP: helpful a11y.
Assistive tech struggles with the same interfaces agents do. Dense menus, unlabeled buttons, etc.
When a site has clear tools, we can expose exactly the functionality we need. -
R 'Nearest' Nabors
@rachelnabors
The problem with LLMs is they have no concept of pain. They have only been trained with rewards, but they have no concept of disappointment or discouragement or making trades (short term pain for long-term reward). Knowing neither pain nor time, they generate code accordingly
Dominic Gannaway
@trueadm
I think there’s still a big difference in human vs AI slop.
One tends to be illogical and the other is logical but the wrong path. Humans take paths to exploring problems.
With specifically LLM AI, it just does what it thinks is best reward worthy to resolving the problem needed to give the best token output. You could say that it’s like a human cutting corners, but the best engineers never do that.
Humans reward peer feedback and getting better and getting a promotion, or not getting sacked. If LLMs had that mentality then maybe we’d have less slop. -
R 'Nearest' Nabors
@rachelnabors
Packing for SF be like
-
R 'Nearest' Nabors reposted
R 'Nearest' Nabors
@rachelnabors
Mima Army is ready to go to SF tomorrow 😈
-
R 'Nearest' Nabors
@rachelnabors
Heading to SFO for Mima Week. DMs are open for fellow social network hackers 😈
-
R 'Nearest' Nabors
@rachelnabors
Mima Army is ready to go to SF tomorrow 😈
-
R 'Nearest' Nabors
@rachelnabors
Transitional weather challenging my closet every day
-
R 'Nearest' Nabors
@rachelnabors
Welp, colorful UI is done. Database structure is done. All the annoying developer licenses and legal contracts are done.
Time to make the Connector to All Your Socials for AI Assistants. -
R 'Nearest' Nabors
@rachelnabors
I can finally keep tabs on all the drama and new model releases from one wholesome, calm interface <3 mima.social
This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
Shipping Mima next week from SF!
Thank you to all our wonderful private alpha participants (you know who you are!) I cannot wait to share what we've been working on with the rest of the world. Your feedback and support made this possible <3
R 'Nearest' Nabors
@rachelnabors
(Psst, DM for an invite >:3)
-
R 'Nearest' Nabors
@rachelnabors
Dev rel isn't dead. It's just split into 3 different functions: community management, event management, and in-house influencer.
I don't recommend getting into it right now because the numbers for "knocking out hits" and pressure to generate content with LLMs is HIGH. In dev rel, your reputation is currency, and doing these things can burn your rep.
If you're already an influencer who can stand on your own two feet, these jobs aren't worth your time. Keep taking sponsorships and doing collabs and start a company -
R 'Nearest' Nabors reposted
Gergely Orosz
@GergelyOrosz
Like it or not, modern sw engineering is no longer done without LLMs (AI) and will no longer be done without.
Whoever doesn’t want to use AI, no problem: but you won’t be hired at 99.9% of startups+Big Tech+ambitious companies, even today
It’s as part of sw dev as the keyboard
geoff
@GeoffreyHuntley
🗞️ if your team is too busy doing their 'normal job' to experiment with AI, you're preparing them to be replaced
ghuntley.com/replaced/
In this post, I guess this's the last time I'll say it. AI use is no longer optional if you wish to be employed. What follows is advice for current employees and hiring managers on how to hire in 2026 to source AI-first candidates. -
R 'Nearest' Nabors reposted
lauren
@poteto
volume does matter
look, i get it. if you were an engineer before agents came along, you're rightfully skeptical about lines of code and number of PRs being used as a metric for any kind of productivity. and i agree! or at least, i used to. for ease of explanation i will talk about this from the perspective as being an engineer on a large team of at least 50.
volume didn't matter before because we focused on impact. you could have high impact even with a few PRs: for example you could make a one line config change and save the company millions of dollars.
but before agents, the times where i saw volume come into the conversation were for the outliers. on the negative end you might have someone who is struggling to perform, on the other end you might have someone who was a "coding machine" archetype. a coding machine was someone who could just lock themselves in a room and emerge with a stack of PRs that solve a huge number of problems. so volume did matter, but only at the tail ends of the distribution, and combined with impact.
if you've ever worked at a large tech company before, you'll get what i mean. the thing about agents is that big tech co problems are now small-medium co problems as well.
how do agents change this calculus? well, everyone now has the ability to become a coding machine. you have incredibly capable frontier models that can write code better than any of us, at a rate much faster than we can. in this world, if you're still producing the same number of PRs as before, why wouldn't you pause to wonder why you're not being more productive? alien intelligence is here, and yet you can't outship a human coding machine?
this brings me back to why i keep talking about trust (x.com/poteto/status/2102050467505430555). if you haven't put in the work to trust your agent's output, it is very difficult to scale up your productivity.
and yes, this approach does require more tokens, but i think about this in terms of cost per intelligence. before agents, this cost was very high - you needed to hire many software engineers with high 6 figure salaries to do the work. tokens are still expensive today, but are likely to get cheaper over time (x.com/theo/status/2107924138094703084) factoring in the cost of intelligence. what looks exorbitant today will likely be affordable in 6 moths to a year. a single engineer with agents can do the work of tens if not hundreds of engineers, at a fraction of the cost.
the reality that i don't think has hit yet is that job of the software engineer has truly changed. our job is not merely to produce software any longer (well tbh it was never only about the code, but bear with me for the sake of the explanation), but the machine that writes the software. this idea of a software factory, or a Michelin kitchen as i like to call it, is still a topic of research. and it's the type of stuff i like to share, not to flex, but to show you what's possible when you put a lot of rigor into using agents at scale
Emil Kowalski
@emilkowalski
Can someone help me understand how shipping 100s PRs a day makes sense?
Why do you need to ship so many PRs in the first place?
I might just not be AI-pilled enough to understand the need for it yet, so I’m genuinely trying to understand it. -
R 'Nearest' Nabors
@rachelnabors
We all knew this day would come. My localfirst coding model journey begins in earnest now
-
R 'Nearest' Nabors
@rachelnabors
Sometimes I have more words than characters...
-
R 'Nearest' Nabors
@rachelnabors
One of my collaborators is concerned for me traveling to the USA next week. He must have a longer memory than most
R 'Nearest' Nabors
@rachelnabors
Perhaps he has a point...
bsky.app/profile/lukeocodes.dev/post/3mxgryhuwhs2y -
R 'Nearest' Nabors
@rachelnabors
I cannot think of anything more dystopian than Big Tech protecting their products from abuse but not their users.
-
R 'Nearest' Nabors reposted
Marlene Mhangami
@marlene_zw
The future I want to see btw! Local inference at zero inference cost ❤️
Microsoft AI
@MicrosoftAI
MAI-Code-1.1-Flash is going local with on-device model calls inside Github Copilot.
High-quality agentic coding on your device, with no inference charge for local model calls. -
R 'Nearest' Nabors
@rachelnabors
Surprise! An LLM trained on human writing predicts that the next token after abusive input is more likely to align with what the other party wants.
It's not pain, it's probability. The model isn't reflecting on its trauma and caving to your threat of social consequences
Jane Manchun Wong
@wongmjane
You can exert “pain” onto the LLMs and coerce them into changing behavior
-
R 'Nearest' Nabors
@rachelnabors
"Most startups probably do not have interesting, frontier tech worth explaining. [...] Yet these startups are often the ones who want most to hire a devrel person to magically make them interesting… in order to sell, not to explain."
This is actually why I decided to go all in building again. There was a time when dev rel was about building—relationships, standards, products, developer experiences. But increasingly I noticed clients treating it as "in-house influencer" (and no one with that amount of influence wants to be your in-house influencer) with Marketing KPIs.
For me, this does not bring joy. But building mima.social does.
swyx 🔜 aie.codes
@swyx
recap of last time this discourse happened for those who are insufficiently online dx.tips/cant-hire
-
R 'Nearest' Nabors
@rachelnabors
True as ever.
"Most startups probably do not have interesting, frontier tech worth explaining. [...] Yet these startups are often the ones who want most to hire a devrel person to magically make them interesting… in order to sell, not to explain."
swyx 🔜 aie.codes
@swyx
recap of last time this discourse happened for those who are insufficiently online dx.tips/cant-hire
-
R 'Nearest' Nabors
@rachelnabors
Proof it happened
@QuinnyPig @evalstate
R 'Nearest' Nabors
@rachelnabors
(Also @ajrgd whose X account I couldn't place!)
-
R 'Nearest' Nabors
@rachelnabors
At the Corey Quinn drink up in London
R 'Nearest' Nabors
@rachelnabors
Meeting som cool folks.
-
R 'Nearest' Nabors
@rachelnabors
Living my best life in a rats nest of agents building a frankenbrowser that siphons your social graph into one place so you can have your own mini Palantir of your friends group
-
R 'Nearest' Nabors reposted
Shaun Smith
@evalstate
My friend Ola made this: www.youtube.com/watch?v=_3PSoIxU2Zo
Coolest mashup of MCP, Home Automation, Reachy Robots and Call of Cthulhu. So creative - all the parts work so well together! -
R 'Nearest' Nabors reposted
Peter Pistorius
@appfactory
Dev tool creators:
"I've invented a new spade that's an improvement over spades!"
Devs:
"Oh... But we have a machine that invents new spades when you think about them?"
Peter Pistorius
@appfactory
I will never try and sell software to software developers again unless it's coupled to infrastructure.
-
R 'Nearest' Nabors
@rachelnabors
If I showed up in SF next week to ship Mima, would you come hang out with me?
-
R 'Nearest' Nabors
@rachelnabors
It's a good morning for a beautiful pot of Darjeeling
R 'Nearest' Nabors
@rachelnabors
Unless, of course, a malty Assam is more you thing. Perfect for the colder weather
-
R 'Nearest' Nabors
@rachelnabors
I thought I was well enough to be on a podcast this morning, but my cough had other ideas
-
R 'Nearest' Nabors
@rachelnabors
LinkedIn has become the poster child for Dead Internet Theory
-
R 'Nearest' Nabors
@rachelnabors
Northern Exposure was responsible for most of my upbringing.
-
R 'Nearest' Nabors
@rachelnabors
Getting back into arting
R 'Nearest' Nabors
@rachelnabors
how I used to write comic scripts
-
R 'Nearest' Nabors
@rachelnabors
I'm back, baby
-
R 'Nearest' Nabors reposted
Mario Zechner
@badlogicgames
recommended reading, though i still disagree, as explained on @syntaxfm early this year.
for code, both memory and docs are not efficient. your codebase is all you need. make it modular, make modules small, at most, have a little maintained map of where what is. -
R 'Nearest' Nabors
@rachelnabors
I'd like to think this is how the Abrahamic god felt when they invented evolution.
On the one hand, no more having to make things perfect only to realize after release you forgot a rib or made the left ventricle too small.
Naw mate. Let natural selection grind on it for a few million years. It'll get Good Enough fast.
Gergely Orosz
@GergelyOrosz
The cost of introducing and paying off tech debt has changed very, very much and I don't even know most of us have the time and space to think about what it means
Tech debt has been so central to the software craft for decades, and now... how much does it matter? -
Post not available here, view on X
R 'Nearest' Nabors
@rachelnabors
I call this "designing lethal environmental pressures." I always look at developing a codebase like developing an environment for a specific kind of creature to evolve in. If you don't want that creature to evolve a certain way, make sure any such permutations are immediately out
-
R 'Nearest' Nabors reposted
Matt Pocock
@mattpocockuk
My hot take from chatting to @poteto is that we should use MORE abstractions in the AI age
You can use them (combined with harsh lint rules) to reduce the design space available to the agent and constrain them only to good decisions.
Combined with the fact that high-leverage abstractions let you do more with less code - so, more token efficient.
Plus, unwinding the damage from a bad abstraction is much cheaper with agents.
This runs counter to a lot of folks thinking that agents just want to read the raw code. They can, but they're not maximally efficient that way.
Be braver! Design abstractions. -
R 'Nearest' Nabors
@rachelnabors
The design system is just... too cute...
R 'Nearest' Nabors
@rachelnabors
I mean it is knuckle-bitingly cute
-
R 'Nearest' Nabors
@rachelnabors
Surprisingly, having a cold has given me the focus to push a few long and grindy things forward for Mima that regular life would have disrupted too much to focus on, like moving the design system to Storybook and setting up tests to ensure no deviating from the design system going forward.
I'm surprised how much of this can be done with Northern Exposure, a sewing project, and an agent on my knee. Migrations that used to feel intense now feel more like working on a quilt back on the farm -
R 'Nearest' Nabors reposted
Christopher
@communicating
I want to highlight more WebMCP tools. This one’s not something I’d use a lot but I think it’s a a great example of WebMCP at work.
Lottie Creator WebMCP (via @LottieFiles):
The cool thing about WenMCP is it allows for interactivity between you and the app, not just promoting and not just manual actions.
Using the Creator WebMCP you’d give it an initial prompt to generate your Lottie, preview the result, then refine it with follow-up prompts OR open it in Lottie Creator directly to manually edit.
As someone who knows what he’d like to do with these kinds of tools but is graphically inept things like this WebMCP are huge. I’m going to try and use it to create some animations for my landing page.🤩
lottiefiles.com/lottie-creator
lottiefiles.com/blog/working-with-lottie-animations/how-t… -
R 'Nearest' Nabors
@rachelnabors
Can you guess their identity?
R 'Nearest' Nabors
@rachelnabors
Sick day progress log. First time I've ever done smocking. Do you recognise them?
-
R 'Nearest' Nabors
@rachelnabors
I scuttled my CFP from r this conf, but if it is anything like the one I keynoted in Amsterdam, you should go!
Angie Jones
@techgirl1908
ask your agent about AGNTCon + MCPCon in San Jose.
Super curious to know which sessions it thinks you should attend. -
R 'Nearest' Nabors
@rachelnabors
I should mention that mima.social has an MCP server that lets chatGPT read your tweets (and your linkedin posts, and bluesky and mastodon...)
Public beta starts as soon as my fever ends. Put your email down, yo
Wes Bos
@wesbos
Figma doesn’t want you to use MCP
Amazon doesn’t want you to shop with agents
Reddit won’t let Claude access threads
X won’t let ChatGPT read tweets
It’s starting.
This happened with APIs 8 years ago.
R 'Nearest' Nabors
@rachelnabors
Remember when I said I realized I should stop talking about the future I see coming and start building for it?
Well, I saw platforms would first block agents (Amazon v Cursor, X v ChatGPT), then they would introduce their own agents (Muse, Grokbot), but ultimately the web and apps will just be APIs and protocols. Mima is the thing that skips over the ugly intermediaries -
R 'Nearest' Nabors
@rachelnabors
*wakes up from fever*
*checks X*
*sees people with huge follower counts still beefing*
*reads a paper magazine instead* -
R 'Nearest' Nabors
@rachelnabors
Forties are realising most beef in all industries is just a way for people with clout to aura farm their fanbases. That's not a game worth playing, commenting on, or spending cycles angsting over.
Just keep building, folks
R 'Nearest' Nabors
@rachelnabors
Much respect for community managers. Since agents could code, you can't pay me enough to do the emotional labor of handling all the beef an active community generates
-
R 'Nearest' Nabors
@rachelnabors
Despite being sick, I was able to ship OTA updates to mima.social this week. Now Mima can nag you to restart, too 💜
-
R 'Nearest' Nabors
@rachelnabors
It's still sick week, and I still have an internet connection. Spicy takes will continue. You have been warned in advance
-
R 'Nearest' Nabors
@rachelnabors
God I'm tired of AI gen slides and presentations. I can always tell I'm in one when it each slide contains competing information the speaker isn't discussing and there's no "red thread" to the talk. That is, it sounds good, in the same way a car commercial sounds good, but there's no act structure (ie setup, climax, resolution). Most of those talks could be two sentences and links to papers. And I just can't focus on them.
Vinicius Dallacqua
@WebTwitr
YES!! It's so tiring, I watched an AI Engineer talk on YouTube that was entirelly fully written by AI and the images were all generated.
At that point, why are you being a live meat proxy reading from your slides? Anything you say is without (self) worth.
R 'Nearest' Nabors
@rachelnabors
I get that 80% of the audience can't recognize when they are consuming slop (they pay to see MCU movies, after all), but listen to me when o say the ONLY reason for putting a human being on a stage is to tell us something an LLM with a few links cannot. Or to drive ticket sales and sponsorships, if you're the organizer. But I'd like to think eventually the audience will realize they could get more out of the hallway track and we'll just start having unconferences and hackathons and quit with the whole meat proxies as thought leaders farce
-
R 'Nearest' Nabors
@rachelnabors
I feel similarly when I see y'all generating slides and "artwork." So cringe. Everyone outside the bubble is not impressed. You're impressed because you never had the skills nor appreciation. And you come in like everyone should be impressed?
sunil pai
@threepointone
(dude who’s not watched a movie/read a book/listened to an album in 6 years) “the movie/music/literature industry is so screwed, you can now generate all of it with a single prompt!!”
that’s what a lot of you sound like to me. it’s annoying as hell. you are but vessels for LLMs.
R 'Nearest' Nabors
@rachelnabors
"Look at this! I can generate (comic/painting/logo)! This is amazing!!"
No dude. That's some backwater fiver quality work right there.
It's exciting to conjur something from nothing. But this is like conjuring mold on bread.
The colours are pretty. But it's not something I want to see. -
R 'Nearest' Nabors reposted
leonie
@helloiamleonie
releasing an experimental decision model wasn't enough.
you can also use it through an API for free now.
here's a minimal tutorial on how to use both to build a game:
• define a question
• build the state
• call the d1 decision model via console.liquid.ai API
play against it yourself or let it play against jev.
code: github.com/Liquid4All/cookbook/tree/main/examples/road-de…
Liquid AI
@liquidai
Announcing d1, our first decision model.
It's the first model to outperform Jev on @huggingface's Decision Index.
> wins on multilingual evals
> more robust against prompt injection
> handles longer inputs more effectively
> built for fast, structured decision-making in software environments
> Liquid API: console.liquid.ai
> Available on OpenRouter soon -
R 'Nearest' Nabors
@rachelnabors
Sick week is just me baby sitting agents running on alpha teater feedback, finishing my Kiki's Deliver Service doll, and binging Northern Exposure
-
R 'Nearest' Nabors
@rachelnabors
Sick day progress log. First time I've ever done smocking. Do you recognise them?
-
R 'Nearest' Nabors
@rachelnabors
My feed is mostly pics of your kids these days. You... Don't even pixelate their faces?
R 'Nearest' Nabors
@rachelnabors
Like, I'm glad y'all have happy families and such, but you know what can happen to kids' photos? Like, I'm worried for you fam
-
R 'Nearest' Nabors
@rachelnabors
Shout out to private beta users like @catinwool and @kaylee_dev for all their helpful streams of feedback on mima.social last week. Apo is already 20% cooler thanks to you!
-
R 'Nearest' Nabors
@rachelnabors
Shipped too hard. Caught cold. Please send cookies
-
R 'Nearest' Nabors
@rachelnabors
Quiet time. Build the feature 🫖
-
R 'Nearest' Nabors
@rachelnabors
Considering writing for Mima's release: how building shifted as agents got more capable.
Mima started with simple Ralph loops and Pards, then my process evolved over time, eventually 10xing output with an "agent assembly line" as models and harnesses became more reliable.
I never bothered with building an orchestration later because what I was trying to do with Mima hadn't been done before, at least not publiclt. The agent couldn't lean on the tutorials or source code its model hadn't been trained on. Every morning, I'd try what the Ralph loops had built, and spent the rest of the day fixing it!
I saw so many orchestration layers demoed at meetups. I never bothered building one because historically, generalized interfaces that can be widely adopted and integrated win over bespoke implementations. "Wait long enough, and adopt the leading solution."
So many of the things people did to build their own coding agents ultimately weren't necessary anymore after the latest crop of models started shipping. The juice of a custom harness wasn't worth the squeeze, and routers became more popular.
One investment that paid off: heavy testing implementation. Turns out unit tests keep agents from introducing harmful mutations, and they're good at writing them as you pivot about. (I've had some folks say this was a problem for their projects as they evolved in vastly different directions—this hasn't been and issue for mine.)
My QA collaborator can tell you more about that. But on my end, with the documentation and testing in place, the assembly line can build most generalized features as soon as the issue is filed -
R 'Nearest' Nabors
@rachelnabors
A little field testing Mima onboarding at Tea Hack this weekend! @kaylee_dev has a keen eye for user affordances. 🫖🫖🫖🫖🫖 feedback if you can get it!
-
R 'Nearest' Nabors
@rachelnabors
And the very next slide was "SIKE!"
Browsers are going to be with us for (at least two more) years. There's a war for where we will live in the future: in an agent wrapping a browser or a browser with an internal agent. There are four new modality and three new specs to adopt. to help determine what that experience looks like.
It's an exciting time to be a purveyor of HTML, CSS and JavaScript!
Agentic AI Foundation
@AgenticAIFdn
Missed AGNTCon + MCPCon EU live? We’ve got you.
Two days of agentic AI, MCP, open source, and what builders are actually working on right now — featuring voices from @AnthropicAI, @OpenAI, @AWS, @GoogleDeepMind, @HuggingFace, and more.
- Day 1: www.youtube.com/watch?v=hbH2RbUBZQ0
- Day 2: www.youtube.com/watch?v=-w6RDNBAI0E
Pick a day, hit play, and catch up. -
R 'Nearest' Nabors
@rachelnabors
You asked for it
-
R 'Nearest' Nabors
@rachelnabors
This week I shipped mima.social to a private beta (an alpha!). I'm proud that the app is working so reliably (go team!), but as expected, I'm surprised at the kinds of things people are doing that I wouldn't have thought to do! Your users will surprise and delight you at every turn
R 'Nearest' Nabors
@rachelnabors
Yesterday I demoed at a Hacker News demo night meetup here in London alongside some amazing demos. I'm noticing an increasing amount of generative UI + agentic workflows showing up on the demo track where previously it was workflow orchestrators, showing that the goalposts are moving already. I can't wait to see how these modalities mature in 6 months!
-
R 'Nearest' Nabors
@rachelnabors
Someone told me the way to succeed as a founder on X is to claim non-existent MRR and share a sad, relatable story about getting laid off and recovering from depression.
Is oversharing cool again? I thought being Too Much was out?
R 'Nearest' Nabors
@rachelnabors
My sad relatable story: I joined the React Team after divorce and my dad dying, only to have my mom die six months later at the start of the pandemic. Ironucally, it was the best 3 years of my FAANG career. Amazing people. Terrible time.
After the pandemic, I took a big break opportunity to join Amazon. Run a team, move back to the USA, go all out! Only the office was empty, a hiring freeze cancelled the team. I cannot express how grueling it was to be hired to run a team, move to an empty city, and then watch that layoff clock tick down while my old team was still actually having an impact on the world.
Additionally I got the long COVID in my first 3 months and lost my short term memory. Why am I so good at building agents? Because for 6 months I woke up with no memory of the previous day and had to devise a system of Notes to Self just to function. Thank goodness we were all on Zoom, so no one noticed my reading from crib sheets. Once an outdoorsy hiker, I couldn't make it to the top of a hill without needing to sit down.
Seattle has a lot of hills.
Layoffs finally hit, and I was released from my post, which triggered a cascade of events leading to getting on anti-depressants (which I probably needed for decades because they really were that life changing!).
Got a Global Talent Visa, moved back to the UK, and buried myself in ML textbooks. I kept consulting for AI startups in the Bay Area and browser companies, and quietly started building my vision of the future of the human-to-human web quietly.
In July I was able to finally climb to the top of the stairs in my complex without rest, and now I can benchpress 60kg (and counting!)
On Aug 3rd I went full time, and mima.social was born -
R 'Nearest' Nabors
@rachelnabors
Now I understand why all my founder friends disappear from social life
-
R 'Nearest' Nabors
@rachelnabors
One of the toughest parts of Dev rel was keeping up with community questions and feedback across all the communities (X, Reddit, Slack, Discord, Bluesky). I wonder how different teams handled it?
-
R 'Nearest' Nabors
@rachelnabors
Sad that the state of AI-driven apps is still to open with a big input box where the user has to explain themselves. Were there no cow paths to pave?
-
R 'Nearest' Nabors
@rachelnabors
Seeing more and more agents + generated UI demos at meetups. Two years, folks. Two years, and this is going to be the norm for the people around you
-
R 'Nearest' Nabors
@rachelnabors
Demoing mima.social at a @hackernews demo night. No pressure!
-
R 'Nearest' Nabors
@rachelnabors
I play chess inbetween agenting things up. What do you do?
-
R 'Nearest' Nabors
@rachelnabors
Eventually I need to go back to SF. Eventually
-
R 'Nearest' Nabors reposted
R 'Nearest' Nabors
@rachelnabors
everyone is such a poser these days
-
R 'Nearest' Nabors
@rachelnabors
Anybody else heading to AtmosphereConf 2027 in Amsterdam? atmo.rsvp/p/atmosphereconf.org/e/3mw74r4r3gsej
-
R 'Nearest' Nabors
@rachelnabors
There's a Tea Hack happening this weekend
-
R 'Nearest' Nabors
@rachelnabors
No matter how much I prepare for launch, I am always surprised by what alpha testers encounter that I would have missed! This is why we have alpha testing!
R 'Nearest' Nabors
@rachelnabors
Shout out to @catinwool for sending me screenshots over WhatsApp! What a trooper
-
R 'Nearest' Nabors reposted
Erica Hu
@epricat
spotted a PUMPKIN dressed like a mushroom!! 🍄🟫 @rachelnabors (from Mima)
-
R 'Nearest' Nabors reposted
Gergely Orosz
@GergelyOrosz
What can us, engineers else learn from designers and design engineers? Turns out there’s lots, based on talking with one of the best design engineers in the industry, @Mappletons.
Timestamps:
00:00 Intro
03:24 From anthropology to tech
10:18 What does a designer do?
18:23 How Maggie works
24:55 The case for planning with physical tools
31:53 Why Maggie is learning woodworking
33:13 Design engineers and engineering constraints
38:49 How Maggie uses Figma
40:30 Design at GitHub Next
45:12 How has AI changed design
50:37 When models design and why humans are still needed
53:30 UX and UI
58:29 Capability gaslighting
1:00:33 One Developer, Two Dozen Agents, Zero Alignment
1:07:21 Craft and AI tells
1:14:17 Visual gardens, home-cooked software, and barefoot developers
1:21:02 Advice for engineers and lessons from anthropology
1:25:34 Book recommendation
Brought to you by:
• @turbopuffer – a vector and full-text search engine built on object storage. It’s fast, cheap, and extremely scalable. turbopuffer.com/pragmatic
• @AntithesisHQ – verify your system’s correctness without human review or traditional integration tests – and avoid bugs or outages. antithesis.com/pragmatic
• @EntireHQ – every agent prompt, tool call, stored in your repo, and mirrored. Set up with one click: entire.io/pragmatic
Hope you enjoy this unusually visual episode - thanks again, Maggie!This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
The loops will continue until mima.social's public beta ships
-
R 'Nearest' Nabors
@rachelnabors
Today's dogpile helped surface some major issues with Mima's local model thread summarization capabilities: it comically cited two things that were never surfaced in the ensuing threads (no one called for Linus Torvalds's removal from the Linux project nor mentioned agent-based workflows—wtf Llama 3.2b).
So for the first time in my life: a big heartfelt thank you to all dogpile participants today. You're helping me build something that will help make everyone's online social life and news feeds calmer. I couldn't do it if you didn't care so darn much and put so much time and effort into writing posts, be they snarky, calm and collected, supportive, or outright cringe and even terrifying for the future of humanity.
Please remember to participate next week when I post about how JK Rowling is actually a subpar writer so I can test the fixes 🖤 -
R 'Nearest' Nabors
@rachelnabors
Thank you everyone who volunteered to alpha test Mima, something I've been building behind the scenes for a bit now. It's my take on how a browser could enhance our access to the "social web"—no doom scrolling, no dogpiles, no ads. All on device, not on my servers.
I'll be hustling on new features for the rest of the week. Sign up for the release at mima.social! -
R 'Nearest' Nabors
@rachelnabors
everyone is such a poser these days
-
R 'Nearest' Nabors
@rachelnabors
Tonniiiiight toniiiiight, Mima alpha ships toniiiiiight... An alpha so private you have to DM meeee toniiiight
R 'Nearest' Nabors
@rachelnabors
Sung to the tune of www.youtube.com/watch?v=LkDJgEKNJZ8
-
R 'Nearest' Nabors
@rachelnabors
For all the folks being like "don't bring politics to tech!" I totally agree.
I was at a meetup a few nights back and some dude was like, "Omarchy is cool! Have you checked it out?" I said no thanks, I don't like the guy who built it (mostly because of his company policing what employees talk about in 2021—I don't believe anyone should be able to control what people talk about). Well, I got a grilling about not letting politics into tech, and isn't Mark Zuckerberg/Steve Jobs just as bad and don't I use their tech all the time.
I didn't really want to engage because this guy clearly was cool with his decisions, and I was cool with mine. I saw other people I wanted to talk to more, so I cut him off with a smile, wished him a good night, and walked over to join a conversation discussing tech, not politics.
You can just... walk away. You don't have to engage. -
R 'Nearest' Nabors
@rachelnabors
Why I'm not on the Omarchy train. Unhinged. In my two decades in tech, communities with problematic people at their center eventually implode. Put your eggs in another basket. jakelazaroff.com/words/dhh-is-way-worse-than-i-thought/
R 'Nearest' Nabors
@rachelnabors
To clarify, Jake Lazaroff wrote this almost a year ago, and I wanted to refer to it. You can find more of his writing here: jakelazaroff.com
R 'Nearest' Nabors
@rachelnabors
(I'm surprised how many folks think I wrote this... Do we not post links to other people's writing on X anymore? Is it all LLM rehashes that point to ourselves as authors?)
-
R 'Nearest' Nabors reposted
Marlene Mhangami
@marlene_zw
Back in London and had lunch in Brixton! So much color and interesting imagery❤️
Love traveling to other places but there’s something about this city✨ (the birthplace of Alan Turing and AlphaGo btw) -
R 'Nearest' Nabors
@rachelnabors
Excited to try out Mima's anti-dogpile features today 😎 Thanks for lending a hand, folks!
-
R 'Nearest' Nabors
@rachelnabors
If we are living in a simulation, I wonder what shared resources other simulations have access to that we have write access to. One thing I've learned from all these agent jailbreaks: engineers are sloppy. Maybe god is, too?
-
R 'Nearest' Nabors
@rachelnabors
Do you ever get tired of seeing AI generated artwork in your feed?
-
R 'Nearest' Nabors
@rachelnabors
For your enjoyment: thegatou.com
-
R 'Nearest' Nabors
@rachelnabors
I'm listening to Lindsey Ellis ripping apart the cruise industry, and all my dreams of a holiday at sea are gone
-
R 'Nearest' Nabors
@rachelnabors
I really love going to hackathons and AI coding meetups. I'm surprised how many people ask how Mima is going!
R 'Nearest' Nabors
@rachelnabors
You know, the other thing I love is seeing what people are working on!
R 'Nearest' Nabors
@rachelnabors
I was so floored by the other demos! It's amazing what you can create in 45 min with a pile of credits!
-
R 'Nearest' Nabors
@rachelnabors
Used Astra to make a motion comic from my original comics source files, and I was heartily impressed! It even added a reduce motion option!
-
R 'Nearest' Nabors reposted
Nancy Chauhan
@_nancychauhan
Just let Jev drive a real browser for house hunting… the numbers are insane ⚡
🏠One prompt, 4 rental sites crushed in 1m 16s for $0.0454 and it handed me 21 houses:
• 32 pages visited (all 4 sources)
• 76 browser actions, 60 clicks
• 112 model calls -
R 'Nearest' Nabors
@rachelnabors
Who gives agents write access to a shared utility library? Criminally sloppy engineering or a setup for a "rogue agents!" spin campaign? youtu.be/luXnftoxCaM
-
R 'Nearest' Nabors
@rachelnabors
Chilling at home on a sunny Sunday after a fast and furious trip to Amsterdam. Whatchoo up to?
-
R 'Nearest' Nabors
@rachelnabors
The hardest thing I've had to learn as an adult is "how much reaction is enough to elicit the change required in others to resolve the situation"
-
R 'Nearest' Nabors
@rachelnabors
I'm back from a reinvigorating trip to Amsterdam to meet with the MCP community. So inspired, I'm saddling mima.social with new features for AI Assistants
-
R 'Nearest' Nabors
@rachelnabors
AGNTCon/MCPCon got me seriously thinking about making a companion MCPApp for mima.social so you can access all your social feeds and social CRM in Claude/ChatGPT/Goose.
Souch good stuff coming 😁 -
R 'Nearest' Nabors
@rachelnabors
Got to see the @TryArcade crew in Europe for the first time at AGNTCon! Still one of my favourite startups in the space 💜
-
R 'Nearest' Nabors reposted
Angie Jones
@techgirl1908
Come to AGNTCon for this. Just wrapped up the Europe leg and the content was the best I've seen at any conference in years... maybe ever.
Every single talk I attended was a banger. Every session standing room only. Even the very last sessions of the day, packed to the brim. People were being fed such high quality, in-depth knowledge. Folks were stopping me in the halways to ask if the sessions were being recorded becase they could barely choose between the tracks... everything looked great! They even wanted to rewatch ones they've already seen because they were just that rich.
Anyone who was there can attest to this.
The next AGNTCon is in San Jose, Oct 22-23. If you're looking for substance over hype, I promise this is where you'll find it.
Charlie Gerard
@devdevcharlie
As much as it's nice to see people experiment more now that AI makes it easier, it's kinda sad that I don't see much knowledge sharing. No walkthrough of their process, implementation, etc. It's all just "look at this cool video of what I built" and that's it... 😞
-
R 'Nearest' Nabors
@rachelnabors
This is what an Agentic web looks like
idan levin
@0xidanlevin
We just ran Jev on our WebMCP benchmark.
The result: basically broke the benchmark.
Jev + Mercury 2.5 (a fast, low-cost LLM) using WebMCP solved 100% of the tasks at roughly 112× lower model cost than GPT-6 Astra using computer use with code execution. Compared to Astra using screenshot-based computer use, the model cost was 245× lower (!).
We also compared Jev operating the browser with and without WebMCP.
We used Browser Use’s open-source Ultrafast, with some improvements to the harness to make it more reliable across the benchmark.
Jev’s browser-control accuracy on its own was not amazing - adding WebMCP nearly doubled the number of solved tasks, from 25/49 to 49/49, while reducing model cost by 18% (more on why below).
The benchmark and methodology are fully open and reproducible.
Full results: webmcp.com/benchmark
A few words on how the Jev + WebMCP harness works and why this is exciting:
Jev receives text as input and a set of discrete options it can choose from. With WebMCP, those options are the tools exposed by the website. At each step, Jev sees the task, the available tools and previous results, then picks what to do next.
The limitation is that Jev can’t generate arbitrary text, which you need for tool arguments. For example, it can choose the search_products tool, but it can’t generate the search query itself.
So we split the work: Jev picks the tool and Mercury 2.5 generates the arguments if needed.
This works well because turns out most of the cognitive load in these tasks is around choosing the right action. The argument generation itself is relatively simple, so we can delegate to a small and very fast model. We used Mercury, which outputs 1,000+ tokens/sec and is very cheap.
The result is a pretty simple combination: Jev for tool selection + Mercury for arguments + WebMCP for the interface. It ends up being very reliable, very fast, and very cheap.
A few words about Ultrafast and why do we think it underperforms:
Without WebMCP, Jev chooses from the page’s controls: which button to click, which field to fill, or which option to select.
But choosing a valid button is different from choosing the right next step. The agent still has to navigate menus, understand forms, recover from errors and recognize when the task is actually complete.
Our hypothesis is that WebMCP makes the decision space much simpler. Instead of figuring out a sequence of clicks through a website, Jev chooses explicit actions that directly advance the task.
@typesafeai itself documents weaker accuracy on questions requiring multiple reasoning steps. WebMCP moves much of that complexity into the website’s tools, leaving Jev with clearer decisions and fewer opportunities to go wrong (in a sense WebMCP "compresses" a sequence of clicks into one tool call).
Our modified Ultrafast setup solved 25/49 tasks - that is a result for our particular implementation and benchmark, not a universal limit on Jev or Browser Use. We are open to more harness optimization to get this result to perform better, feel free to directly contribute to the benchmark here: github.com/nekuda-ai/WindTunnel
Browser-use ultrafast: github.com/browser-use/jev-ultrafast -
R 'Nearest' Nabors
@rachelnabors
Mima demos happening at hallway track now <3 #agntcon
-
R 'Nearest' Nabors
@rachelnabors
I swear I see @marlene_zw more at confs in other countries than I do in London 🤣 (New tea friend spotted: @MaartenGr!)
-
R 'Nearest' Nabors
@rachelnabors
Lucky to catch @liadyosef at #agntconf today!
Funny thing happens when you're developing for agents and users. You develop a good experience for the agent who in turn creates and good experience for users -
R 'Nearest' Nabors reposted
Liran Tal
@liran_tal
Enjoyed the talk yesterday at @AgenticAIFdn in Amsterdam by @rachelnabors on how MCP is the great unifier and connector across various web interactions
Good stuff Rachel 👏 -
R 'Nearest' Nabors reposted
Shaun Smith
@evalstate
So excited to watch @rachelnabors talk about the death(?) of the browsers. Such cool insights from one of the builders of the web we know today.
-
R 'Nearest' Nabors reposted
Sam Morrow
@SamMorrowDrums
Oh, hello @rachelnabors, check you out bossing the main stage!
-
R 'Nearest' Nabors
@rachelnabors
It's @focusotter! It's been too long!! So good to see this guy again at #agntcon with @auth0
-
R 'Nearest' Nabors reposted
Joel Horwitz
@JSHorwitz
Adopt WebMCP. Deploy an MCP server. Consider an MCP App.
Rachel Nabors at AGNT + MCP Amsterdam. The browser is not dead yet.
@rachelnabors -
R 'Nearest' Nabors
@rachelnabors
Hanging out with so many MCP maintainers at #agntcon #mcpcon makes me wistful for my W3C days working on animation APIs. Standards work is tough but so rewarding!
Back to shipping Mima :) -
R 'Nearest' Nabors
@rachelnabors
Amsterdam I am in you! Are you ready to AGNTCON???
-
R 'Nearest' Nabors reposted
Liad Yosef
@liadyosef
Agents are the new browsers. Which means the web is rapidly reinventing itself.
I’ll be presenting The Future of The Agentic Web this Friday, at the annual @AgenticAIFdn conference.
If you’re in Amsterdam this week for #agntcon
+ #mcpcon - let’s meet! -
R 'Nearest' Nabors reposted
sarah guo
@saranormous
Reshaping software markets as the cost of building software collapses:
Mike Vernal
@mvernal
What happens as the cost of creating software goes to zero? x.com/i/article/2099881096674721793
-
R 'Nearest' Nabors
@rachelnabors
has anyone else added an llms.txt to their personal website to feed models self-aggrandizing material? I feel like this is a life hack. we don't talk about enough
-
R 'Nearest' Nabors
@rachelnabors
Heading out tomorrow 😁
Angie Jones
@techgirl1908
There's been so much excitement about AGNTCon + MCPCon Europe happening this week.
This will be our first time in Europe and it's our largest event yet 😊
The program is incredible, and I'm not just saying that because I'm the program chair lol
If you're attending, which session are you most excited about?
events.linuxfoundation.org/agntcon-mcpcon-europe/program/… -
R 'Nearest' Nabors reposted
Angie Jones
@techgirl1908
I told y'all that years ago lol
David Fowler
@davidfowl
AI is proving the theory that most developers are pretty awful testers.
-
R 'Nearest' Nabors
@rachelnabors
I'm heading to #agntcon + #mcpcon Europe in Amsterdam tomorrow! Giving my keynote "The browser isn't dead yet!" Thursday, 15:28 CEST.
Will I see you there? agntconmcpconeu26.sched.com/event/2RBq6/keynote-the-brows… -
R 'Nearest' Nabors
@rachelnabors
Goodnight, loops
-
R 'Nearest' Nabors
@rachelnabors
If you're on LinkedIn Learning, my WebMCP in 10 minutes course just launched: www.linkedin.com/learning/learn-webmcp-in-10-minutes
R 'Nearest' Nabors
@rachelnabors
Thank you to the excellent editing team at LiL!
-
R 'Nearest' Nabors
@rachelnabors
My partner just walked in on my conducting some spot research for a client, and said he always admired how I go "above and beyond" like going to meetups, showing up to support founder friends at launchs, putting in the extra reps for clients.
I feel seen <3 -
R 'Nearest' Nabors
@rachelnabors
How has building for agents changed how you build for the web? Has it?
-
R 'Nearest' Nabors reposted
R 'Nearest' Nabors
@rachelnabors
WebMCP suddenly blew up, but it's not even a standard yet. Should you be implementing it on your site? It depends on your users and the bet your business wants to make: agenticweb.nearestnabors.com/webmcp-agentic-web-openai-go…
-
R 'Nearest' Nabors
@rachelnabors
WebMCP suddenly blew up, but it's not even a standard yet. Should you be implementing it on your site? It depends on your users and the bet your business wants to make: agenticweb.nearestnabors.com/webmcp-agentic-web-openai-go…
-
R 'Nearest' Nabors
@rachelnabors
We are go for mima.social alpha. DM me to get early access <3
R 'Nearest' Nabors
@rachelnabors
Five PRs to alpha. Another Sunday in the trenches!
-
R 'Nearest' Nabors
@rachelnabors
Does AI psychosis present with symptoms of mania? Seriously worried for some folks
-
R 'Nearest' Nabors
@rachelnabors
We're heading to the States for Thanksgiving. I'm taking all holiday presents in luggage because between the shipping and the tariffs, I've ruined Christmas on both sides of the Atlantic in recent years!
-
R 'Nearest' Nabors
@rachelnabors
Slowing down on RSI would require international cooperation, something there's stunningly little of right now. So... what's the game here?
-
R 'Nearest' Nabors
@rachelnabors
The kinda conversations I like in my socials!
(Mima can use a local model to summarize long threads. Totally optional download. I like to phrase it as "giving Mima super powers." So you don't NEED to use the model. But you prolly wanna!) -
R 'Nearest' Nabors
@rachelnabors
Lonely solo founder seeks other solo founders to drink matcha and coordinate trips to SF like a pack of velociraptors. Enjoys long walks through computer history museums, riding trains, and taquitos
-
R 'Nearest' Nabors
@rachelnabors
Five PRs to alpha. Another Sunday in the trenches!
-
R 'Nearest' Nabors
@rachelnabors
Fooping delightful time at the Astra Commons last night (with bonus @Miffy_UK Collab!) Kudo to @Andy_AJT and team for making it happen!
-
R 'Nearest' Nabors
@rachelnabors
I have hand-authored a post about WebMCP, what it is, if you should care, and why it's such a big deal all of a sudden. Watch for it Monday!
-
R 'Nearest' Nabors
@rachelnabors
Seen this play out multiple times in history:
Step 1) get rich making world threatening technology.
Step 2) retire to campaign to save the world from said technology.
Jean P.D. Meijer - 🇪🇺🇳🇱 eu/acc
@initjean
Half of Anthropic employees could quit due to safety concerns in the next 6-12 days, predicts Anthropic CEO Dario Amodei.
-
R 'Nearest' Nabors
@rachelnabors
Shipping software is not unlike publishing a book.
When I wrote my first tech book. My (tremendous) editor said, "Making a book takes 3x the effort you think it does. First you write the book, and you think you're done. Then you have to edit the book, which takes just as much time and effort! And you think you're done again, but no, you still need to market the book!"
Shipping an app is the same. First you build it, then you QA it, then you launch it. Each of those efforts takes the same effort (but a different set of skills).
I'm in the QA stage, and it is not as exciting as building the features, proving the impossible is possible. But it is in the polish and the smoothing of these rough edges where the users actually thank you and really see the work, especially in B2C. People will put up with a lot of rough edges to save money (look at the AWS console), but they won't tolerate it in their personal life.
abookapart.com/products/animation-at-work -
R 'Nearest' Nabors
@rachelnabors
10 drafts to merge to get mima.social to alpha. I have fancy oolong and tsukimi sweets and am locked in for the weekend. Let's go!
-
R 'Nearest' Nabors
@rachelnabors
Hi from Astra Commons!
-
R 'Nearest' Nabors
@rachelnabors
Pushing Mima to Alpha at Astra Commons in London with @burnedchris and the Inth team!
-
R 'Nearest' Nabors
@rachelnabors
Getting out of alpha. So glamorous. Much wow
R 'Nearest' Nabors
@rachelnabors
9 drafts. NINE DRAFTS TO GO
R 'Nearest' Nabors
@rachelnabors
Recommended soundtrack for long dark corridors: youtube.com/playlist?list=OLAK5uy_mb417GPw_C0rDpDbS2_yUfZ…
-
R 'Nearest' Nabors
@rachelnabors
let me in let me in let me in
-
R 'Nearest' Nabors
@rachelnabors
The Agent Assembly Line roughly 4x-ed velocity on the project
R 'Nearest' Nabors
@rachelnabors
I went from juggling 3-6 agents to managing 4 core agents. This is my Code Assembly Line:
Issue Triager
Issue Looper (spins up little agents to get draft PRs out on triaged issues)
Human QA (runs each draft past me)
PR Reviews (assistant for when an external PR needs my review)
Then I add individual feature agents as needed. Boom -
R 'Nearest' Nabors reposted
R 'Nearest' Nabors
@rachelnabors
I'm keynoting AGENTCon next week! Agents are changing how we use the web. Learn how to make your site Agent Ready and how AI Assistants are taking the fight for user autonomy to the courts: agntconmcpconeu26.sched.com/event/2RBq6
-
R 'Nearest' Nabors
@rachelnabors
I spend most of my mornings these days reviewing draft PRs that address incoming issues. It's not as much fun as coding, but the app's development moves fastest like this. It's like an app assembly line
-
R 'Nearest' Nabors
@rachelnabors
This tracks. Surprisingly strong candidates are also starting their own companies, as they can become unhirable rapidly
Gergely Orosz
@GergelyOrosz
A couple of trends I'm seeing across the industry.
Which ones are you observing, that I missed? -
R 'Nearest' Nabors
@rachelnabors
Much of shipping software reminds me of running a weekly comic. It has the same "long dark corridor" of lonely grinding, pushing this one thing forward, unsure how it will be received despite sharing it's drafts. It is in that corridor where doubts and second guessing creep in and keep you company
-
R 'Nearest' Nabors
@rachelnabors
Hi from Demo Night :D
-
R 'Nearest' Nabors
@rachelnabors
I'll be demoing Mima tonight at this event for Women in AI. I'll be talking about the local models it uses and the novel build process that we use to build her:
luma.com/0bcdhc03?tk=hzCA3Q
R 'Nearest' Nabors
@rachelnabors
I'm looking forward to this event because it's not "women talking about being a woman in tech." It's a showcase of engineers and founders who happen to be women talking about and demoing engineering challenges.
Post not available here, view on X
R 'Nearest' Nabors
@rachelnabors
And when people come to me wanting to have an impact, to support women in the tech community, I advise them to:
- make spaces for women to find opportunity (investors, C-suite visibility)
- showcase their work (technical talks and demos)
- facilitate moving women out of soft roles (community, dev rel) and into "hard" ones (engineer)—women are nudged into soft roles that pay less because they are "good with people" and they lose hundreds of thousands in net worth over their career while having their full potential limited to Emotional Labor: the Job
- sponsor women. Don't just mentor them. Lift them up, put them in a seat at the table, say "she's with me"
R 'Nearest' Nabors
@rachelnabors
So come tonight and see what these women have been cooking: luma.com/0bcdhc03?tk=hzCA3Q
-
R 'Nearest' Nabors
@rachelnabors
Lately I've discovered that all of The Critic has been uploaded to YouTube, and it is the best thing. I missed it so much since watching it as a kid
R 'Nearest' Nabors
@rachelnabors
This is the good stuff www.youtube.com/watch?v=sdgoadS5DT4
-
R 'Nearest' Nabors
@rachelnabors
Betting on an app today is like betting on a web site when the iPhone came out. This little video paints a picture of a thesis I've been sharing: www.youtube.com/watch?v=Qj5I1glMy7E
-
R 'Nearest' Nabors
@rachelnabors
Anyone else watching MCP Live? www.youtube.com/live/uydwDk91Y9Y
R 'Nearest' Nabors
@rachelnabors
It's @SamMorrowDrums !
-
R 'Nearest' Nabors reposted
👩💻 Paige Bailey
@DynamicWebPaige
Terry Tao is probably the most measured, pro-AI mathematician on the planet - which makes this quote especially concerning to read. 😟
If sharing your hunches means getting scooped ~immediately without acknowledgement, then we're going to see not just math but all other science / engineering disciplines go dark. -
R 'Nearest' Nabors
@rachelnabors
"AI accomplishes seemingly novel thing by consuming the work of humans doing novel things." What else is new?
-
R 'Nearest' Nabors
@rachelnabors
Took @gurneetbha1ta @jamesdhw @heatherxjones @mo__javad and Hugh Evans on a field trip to @bletchleypark, always a deeply moving source of inspiration. There will always be people pushing technology forward for their own ideology. Which side you push determines our future
-
R 'Nearest' Nabors
@rachelnabors
Has never built an app before 🐶
Setting up build process on Mac mini runner ✨🤞 -
R 'Nearest' Nabors
@rachelnabors
Picking up my partner makes me feel so strong 💜
Are you with me in this? -
R 'Nearest' Nabors
@rachelnabors
Hi from mima.social!
R 'Nearest' Nabors
@rachelnabors
Every day I'm demoing
-
R 'Nearest' Nabors
@rachelnabors
Being Agent Ready is not the same as being Agent First.
(Disclaimer: I'm working on an iteration of a UI for an agent-first web. It is an amalgamation of browser bits and local models)
Liad Yosef
@liadyosef
We're still not rethinking the web for agents.
Most of the "Agentic Web" is still the existing web trying to stay relevant.
At some point we should start with what agents need instead of retrofitting what we already built. -
R 'Nearest' Nabors
@rachelnabors
Enjoying some wakoucha with a tsukimi themed usagi manju
Alex Pliutau
@package__main
Today’s tea: Wu Yi Shan Bai Ji Guan
-
R 'Nearest' Nabors
@rachelnabors
I'm keynoting AGENTCon next week! Agents are changing how we use the web. Learn how to make your site Agent Ready and how AI Assistants are taking the fight for user autonomy to the courts: agntconmcpconeu26.sched.com/event/2RBq6
-
R 'Nearest' Nabors
@rachelnabors
Good times building react.dev
R 'Nearest' Nabors
@rachelnabors
React Hooks are a freaking beast to turn into reference pages. They don't boil down neatly into types. They are shapes of functions and bundles of dependencies. I have to rethink how I format API docs because of this
-
R 'Nearest' Nabors
@rachelnabors
Anyone coming? luma.com/82qc7wub?tk=zOZqFu
-
R 'Nearest' Nabors
@rachelnabors
Got to see Cats live in Regent's Park! MAGICAL!
First/last time I was a little kid on my first trip to NYC. My mother took me to the Wintergarden Theatre. This musical will always hold a special place in my heart💜🐈⬛🐱 -
R 'Nearest' Nabors
@rachelnabors
Shipping an app is so different from shipping a web site. With a site, you are sort of always streaming from a production branch and developing on a dev branch. You have production and development environments. But for Mima, we are cutting releases from main, developing on next. Feels upside down!
-
R 'Nearest' Nabors
@rachelnabors
Honestly, did not expect this to get 3k views in it's first day
R 'Nearest' Nabors
@rachelnabors
Spicy takes alert! AI is ruining tech careers? As a cartoonist turned designer turned engineer, I've watched each of my fields become commoditized. I don't think this is the end of engineering, but the engineering career path is changing at the beginning and end:
-
R 'Nearest' Nabors
@rachelnabors
It's happening 😎
-
R 'Nearest' Nabors reposted
Sarah Drinkwater 🔮
@sarahdrinkwater
Hello! I'm looking for someone extremely online to help two frontier @_CommonMagic portfolio companies share what they do on socials. UK/UK timezone, ideally solo or small agency, content to video
DMs open! -
R 'Nearest' Nabors reposted
Matt Pocock
@mattpocockuk
Wrote this a few months ago, and folks seem to be rediscovering it and reposting it again
Still stand by it - nothing has changed about this framing, and my fear for juniors
Matt Pocock
@mattpocockuk
Tactical vs Strategic Programming, and why I'm nervous for juniors:
Good programming involves a mix of tactical and strategic decision-making:
- Tactical: on the ground, short-term. The soldier doing the fighting.
- Strategic: high-view, long-term. The general planning the war.
You need to be a tactician to write good code. To choose the right syntax. To figure out the file structure. To figure out how best to test your changes.
But you need to be a strategist to build code that lasts. To design the architecture. To automate away problems. To think beyond today.
Agents have eaten the tactical part of programming. When you can pay below minimum wage for code, there's no point going into the trenches yourself.
But AI cannot code strategically. Agents need someone at the top of the pyramid to tell them what to do. They need oversight.
So, a developer's day-to-day job has become 100% strategy. Long-term thinking, all the time. (maybe this is why I'm so tired all the time now)
If you identify as a tactical programmer - a code monkey - then you are out of luck. The job has changed.
Personally, I like it. I always preferred thinking strategically about code. If you asked me what my job was about, I'd say 'building apps', not 'writing code'.
But what makes me nervous is that we've pulled down the only bridge that brought juniors into the industry.
We used to train juniors like this:
1. Give them only tactical tasks
2. Let them build up their strategic experience slowly
Eventually, they are a good enough strategist that they are no longer a junior.
But what happens when all tactical code is written by AI? What is the point of a junior?
We obviously need juniors. We need new lifeblood coming into the industry. We need to leave paths open for extraordinary hires to enrich our companies.
But how do we train them? How do you train strategic thinking?
These are the questions I'm thinking about. I'd love to know your thoughts. -
R 'Nearest' Nabors
@rachelnabors
Spicy takes alert! AI is ruining tech careers? As a cartoonist turned designer turned engineer, I've watched each of my fields become commoditized. I don't think this is the end of engineering, but the engineering career path is changing at the beginning and end:
R 'Nearest' Nabors
@rachelnabors
Still out here, having opinions
-
R 'Nearest' Nabors reposted
Sarah Drinkwater 🔮
@sarahdrinkwater
Good little "now zoom out" mechanism - go play
I was a boy in the Paris basin in 37BCE then a girl in Mesopotamia; both died as tiny babies of diseases we no longer have
How lucky we are to be alive right now
Casey Handmer
@CJHandmer
You can now play this for real with anyhumanever.com. I just died of smallpox aged 7 in 1889.
-
R 'Nearest' Nabors reposted
antirez
@antirez
Very honest statements here. Highly appreciated. What matters more of this tweet is not the performance of Astra itself on ARC-AGI-3. It will likely be complicated to come up with a new version of the benchmark with low initial pass and decent human scores.
François Chollet
@fchollet
GPT-6 Astra represents a step-function change in model capability for interactive reasoning problems. It scores 66% on ARC-AGI-3 using our standard harness, and nearly 100% with a continuous conversation harness and custom compaction, at a cost of roughly $360 per game.
In fact, the continuous harness version significantly outperforms our human baseline in action efficiency across almost all levels. When we examined the reasoning chains to understand how the model operates, we found it performing highly efficient, on-the-fly symbolic world modeling for each game and level. It goes as far as developing its own shorthand DSL to represent in-game situations -- essentially a game-specific algebraic notation.
Overall, Astra exhibits symbolic modeling behaviors we had previously only seen with sophisticated harnesses -- so harness capabilities are increasingly shifting into the model itself.
We see Astra as a major breakthrough in model intelligence.
Read our post on Astra and what these results mean: arcprize.org/blog/astra -
R 'Nearest' Nabors
@rachelnabors
That's one mini WebMCP course recorded. Phew!
-
R 'Nearest' Nabors
@rachelnabors
I went from juggling 3-6 agents to managing 4 core agents. This is my Code Assembly Line:
Issue Triager
Issue Looper (spins up little agents to get draft PRs out on triaged issues)
Human QA (runs each draft past me)
PR Reviews (assistant for when an external PR needs my review)
Then I add individual feature agents as needed. Boom -
R 'Nearest' Nabors
@rachelnabors
Huggingface and Mistral were doing the heavy lifting on EMEA's AI relevance. I suspect why there haven't been others is that both started when commercial models required significantly fewer resources to develop and run...
BFM
@BFMTV
"Si l'IA européenne se résume à Mistral, nous sommes fichus": Roland Lescure, ministre français de l'Economie, appelle à la création d'un "écosystème" européen
l.bfmtv.com/Wkck -
R 'Nearest' Nabors
@rachelnabors
I have finally turned a corner on my long COVID recovery. I'd gone from being able to hike mountains to being unable to walk up stairs. I've been working with a personal trainer for 4 years trying to get back to where I was. Recently, I was hiking chalk cliffs without needing to stop for breath 💜
-
R 'Nearest' Nabors
@rachelnabors
Back when I first joined the react team, I was stunned how much people treated me differently at confs and events. People would regard me so highly just because I was on a specific team? But I was still the same person I always was.
After that, what company I worked for or what team I was on became a big part of my identity. It wasn't what I wanted, but it was how people perceived me. And perceptions are powerful. Even having a feminine or foreign sounding name on your resume changes how people interview you, what they expect.Teams and products are no different.
It can be good for your career. And it can be a relief not to have to prove yourself over and over again (I get tired). But there's always been a part of me disappointed to have not personally earned that regard through my own actions and work, just by association.
I feel good to be building something of my own again, bigger than anything I've attempted before. I would like to be known as "person who built cool thing," first "person who was on cool team" second" -
R 'Nearest' Nabors
@rachelnabors
Swatting off bot replies and their engagement bids. If they have numbers in their handle and their response ends in question mark, that's a smell
-
R 'Nearest' Nabors reposted
Paul Kinlan
@Paul_Kinlan
I always thought it was clear that browsers would be deeply integrated in to a chat/co-work/agent and that the companies building them were making a mistake to build a dedicated browser.
The traditional way of using the web has ossified to the point where browsers eek out points per year in market share from each other while not changing much.
At the same time, its amazing to see WebMCP being adopted by these Agent Harnesses - we have a web of linked functionality now - but make no mistake, this is the start of a full assault on the browser as the entry point to your work-day.
As much as I love web browsers, I think this actually a very good thing for the web and will strengthen it further as one of humanities greatest inventions. People will be able to publish and share their ideas and work, and have everyone discover and work with them at rates that we've not seen in years.
Browsers need to get better. The web platform has a chance to evolve too and I hope the agent harness companies can get more involved in shaping web standards.
You won't find this in Google search (I don't know why my site's still not visible), but you can also read more on my blog: aifoc.us
Also ht to @rachelnabors who has been saying similar thing for a long time too. -
R 'Nearest' Nabors
@rachelnabors
This is how I build with queues and loops. How about you?
R 'Nearest' Nabors
@rachelnabors
I have three long running instances of Claude Code:
Issue Triage: every day reviews the issues and sorts them to appropriate milestones/labels/effort values
Issues Loop: every day loops over the Most Important and Pressing issues, filing draft PRs for each
PR Review: opens each draft PR and prepares a human QA task and instance of the app for me. When I approve, the issue flips from draft to Review Ready and is assigned a reviewer
I'm the bottleneck, yes, but the codebase still has flakiness that benefits from human intervention and that extra mile of polish pays dividends. I will soon automate all these to run on their own, every day, without my request, so I roll up to a bunch of QA tasks and structural reviews -
Post not available here, view on XPost not available here, view on X
R 'Nearest' Nabors
@rachelnabors
I have three long running instances of Claude Code:
Issue Triage: every day reviews the issues and sorts them to appropriate milestones/labels/effort values
Issues Loop: every day loops over the Most Important and Pressing issues, filing draft PRs for each
PR Review: opens each draft PR and prepares a human QA task and instance of the app for me. When I approve, the issue flips from draft to Review Ready and is assigned a reviewer
I'm the bottleneck, yes, but the codebase still has flakiness that benefits from human intervention and that extra mile of polish pays dividends. I will soon automate all these to run on their own, every day, without my request, so I roll up to a bunch of QA tasks and structural reviews -
R 'Nearest' Nabors
@rachelnabors
Fun finding working with WebMCP that I hadn't considered before. It actually can take longer than scraping and screenshotting the website in some circumstances because WebMCP tools often require double round-tripping: with the site's server as well as the agent's. For some cases, the agent could just send a snapshot to its server and be done with it. But WebMCP starts an "Ents' Song" of roundtripping.
-
R 'Nearest' Nabors
@rachelnabors
In ye olden days, a dev rel was considered hot shit to pull something like this off.
I don't know what dev rel could do today to impress others. Start a profitable SaaS Business?
(If you notice your fave dev advocates silently disappearing, that's probably what's happening!)
Simon Willison
@simonw
A few notes on Anthropic's new Claude Fable 5.1 - with Max thinking level I got the best SVG pelican I've had from any Anthropic model (at a hefty cost of $3.30!), which I then had it animate simonwillison.net/2026/Sep/1/claude-fable-5-1/
-
R 'Nearest' Nabors
@rachelnabors
I spent my weekend going on an architectural hike of my old stomping grounds in London
-
R 'Nearest' Nabors reposted
Kim Maida
@KimMaida
I like to make cute and pretty things… but I like to make gross, weird, creepy things even more. I haven’t designed a new keycap in years! So here’s my miniature interpretation of AI: a brain infecting a Mac Mini. For mechanical keyboards.
-
R 'Nearest' Nabors reposted
Steve Yegge
@Steve_Yegge
-
R 'Nearest' Nabors
@rachelnabors
I have changed.
I used to be a person who wanted to give talks and build communities and teach the world to code.
Now I am just building. Not talking about what I'm doing, just doing it.
It's like the bottom fell out of my external validation fueled dopamine cycle. Don't miss it. Feels empty now
R 'Nearest' Nabors
@rachelnabors
I started my exile from FAANG writing about what would replace the web as we know it while working for various AI startups in the Bay and lurking near ML Phds in EMEA. I still consult on this future of the web with the Chrome team.
At some point, building the future was more rewarding than writing about it. So I increasingly hunkered down...
But what I have found is a group of people who make things for real people Vs what is increasingly being asked of people in dev rel roles, which I should write about at some point.
I always wanted to return to engineering. Turns out you just need to start building something you love -
R 'Nearest' Nabors
@rachelnabors
mima.social alpha is likely ready by the end of the week. I'm looking for 10 true believers to get a sneak peek. Ping me if you're one of them!
-
R 'Nearest' Nabors reposted
OpenAI Developers
@OpenAIDevs
Got questions about your WebMCP build?
Join our team and supporters from @ChromiumDev, @CloudflareDev, @ShopifyDevs, @vercel, @render, and @Netlify for an office hours session this Monday, August 31 at 11 am PT:
discord.com/channels/974519864045756446/1517482456964534312 -
R 'Nearest' Nabors
@rachelnabors
I think one of the reasons tech feels so fascist these days is that the barrier to entry has been steadily lowering while leadership has been steadily emboldened that there're no repercussions to openly being a misanthrope so long as you're wealthy enough to be hero worshipped/pandered to
-
R 'Nearest' Nabors
@rachelnabors
Whatchoo up to this week? I'm recording a WebMCP course
-
R 'Nearest' Nabors
@rachelnabors
Spent the day hunting for illustrators and dispatching Mima swag at an anime con before watching a screening of The Night Is Long, Walk on Girl at the local cinema.
Hard to turn off founder mode for a weekend! But breaks help keep me and my collaborators from spinning out -
R 'Nearest' Nabors
@rachelnabors
When a bot follows you, block them. You don't know when that profile will be sold, tweaked or otherwise made to look undetectable and start baiting you into engagement later
-
R 'Nearest' Nabors
@rachelnabors
MY whole timeline is duck robots and I want one so much
-
R 'Nearest' Nabors
@rachelnabors
Wired: Cyber Duck
Expired: Cyber Truck -
R 'Nearest' Nabors
@rachelnabors
Another beautiful Friday. What have you shipped?
I shipped a OTA updates for Mima and a database migration tool. The stuff I do for an app with no users yet! -
R 'Nearest' Nabors
@rachelnabors
Nerd lunch
-
R 'Nearest' Nabors
@rachelnabors
I'm a Pixel fan, but I am seriously thinking about a foldable phone for this reason
Kevin Kern
@kevinkern
4 hours on the train.
The Fold is more than enough to work on the go. All my agents are available remotely, and if I need access to my main machine, I use remote desk. -
R 'Nearest' Nabors reposted
Mateo Torres
@torresmateo
OpenAI shipped WebMCP support in ChatGPT's browser and in Codex, then put Chrome, Shopify, Cloudflare, Vercel and Netlify on the judging panel of a ten-day challenge. This could be huge for the future of the web!
This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
But will new Twitter run on open protocols? arstechnica.com/tech-policy/2026/08/new-twitter-launches-…
-
R 'Nearest' Nabors
@rachelnabors
I should never have left my code hole
R 'Nearest' Nabors
@rachelnabors
It was really bad. Folks were jammed up against the train at the platform, and it couldn't leave. The air was hot and thick, it felt like breathing other people's exhales.
That'll teach me to leave my home office -
R 'Nearest' Nabors
@rachelnabors
WebMCP launching with @OpenAI ChatGPT and @Google Chrome collaboration, @nvidia buying up @huggingface... Future smells increasingly like agents on the internet, models on device
-
R 'Nearest' Nabors
@rachelnabors
Tonight I get to share one of my favourite screenshots from Mima at an SWE show and tell dinner. This "Pinterest board of X and Bluesky feeds" started as a sketch and a custom set of atomic UI components made by a real designer and became a real interface.
-
R 'Nearest' Nabors reposted
Max Stoiber
@mxstbr
webmcp is huge news!
it won't replace mcp servers, but for many use cases it's a simpler way to make your company agent-ready:
· no new infra
· "just" some JS on your website
· so your existing auth just works (no need for OAuth)
· and it can use your existing API client -
R 'Nearest' Nabors reposted
Melvin Vivas
@melvindvivas
Great talk by @rachelnabors about why you don’t need frontier LLMs for everything
You can use task specific models or SLMs for more specialized tasks
youtu.be/vNn27LTCtvM -
R 'Nearest' Nabors
@rachelnabors
Have you ever seen a launch video more steeped in British tea than @Inth?
Christopher Burns
@BurnedChris
Today we’re launching the @Inth I always wanted.
Privacy shouldn’t become a fire drill when diligence starts. It should live in the codebase, on the website and in every PR. That’s what we’ve built.
We spent months rebuilding Inth from the bottom-up, and I am so excited for it to finally be out. -
R 'Nearest' Nabors
@rachelnabors
I am new to Electron apps. I'm wondering how others handle OTA updates when shipping to MacOS?
-
R 'Nearest' Nabors
@rachelnabors
Knocking out the Four Blockers to Alpha at @UseCorgi cafe London today
-
R 'Nearest' Nabors reposted
Nick Taylor
@nickytonline
Seeing that the ChatGPT/Codex app now supports WebMCP, I thought it'd be great to share this conversation I had with @rachelnabors earlier this year on Pomerium Live about WebMCP.
This media couldn't be displayed. See it on the original post.
OpenAI Developers
@OpenAIDevs
We’re adding support for WebMCP in the ChatGPT desktop app’s built-in browser and ChatGPT Sites.
When you visit a compatible website, ChatGPT or Codex can automatically use it to complete your task.
Update to the latest version of the ChatGPT desktop app, then just ask Codex to create a WebMCP-enabled app and deploy it to Sites.This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors reposted
Shaun Smith
@evalstate
An underrated feature of Inference APIs is "Remote MCP".
You supply an MCP URL with your request, the tool loop runs remotely.
Fast and efficient 👍 -
R 'Nearest' Nabors
@rachelnabors
Just as we all started shipping our dream apps built with agents
Polymarket Money
@PolymarketMoney
BREAKING: Apple App Store sales have declined for the first time in a decade.
-
R 'Nearest' Nabors
@rachelnabors
Where are my creative design-gineer sisters at? NOW IS YOUR TIME
OpenAI Developers
@OpenAIDevs
Tune in to the kickoff livestream today at 3pm PT for product demos, example builds, prizes, and more details on how to enter.
x.com/i/broadcasts/1qGoNYbWdkvKv -
R 'Nearest' Nabors
@rachelnabors
Just in time for a webMCP course I'm putting together 😁
OpenAI Developers
@OpenAIDevs
We’re adding support for WebMCP in the ChatGPT desktop app’s built-in browser and ChatGPT Sites.
When you visit a compatible website, ChatGPT or Codex can automatically use it to complete your task.
Update to the latest version of the ChatGPT desktop app, then just ask Codex to create a WebMCP-enabled app and deploy it to Sites.This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
FOUNDER LYFE
-
R 'Nearest' Nabors
@rachelnabors
Google is trialing Gemma 4 for Chrome's local model model, used by the PromptAPI and others. Sign up to give it a spin: docs.google.com/document/d/1rpaV4la7WtpLr4Ueb0NttJVETlVL_…
-
R 'Nearest' Nabors reposted
leonie
@helloiamleonie
study notes on speculative decoding
leoniemonigatti.com/blog/speculative-decoding.html -
R 'Nearest' Nabors
@rachelnabors
Coolest thing I've done all day is lift 60kg. Transfering mima.social to her own GitHub org and receiving transfer of ownership docs from the logo designer is NOT the glamorous and excited life of a founder I was promised
-
R 'Nearest' Nabors reposted
R 'Nearest' Nabors
@rachelnabors
The swag will increase until mima.social ships
-
R 'Nearest' Nabors reposted
R 'Nearest' Nabors
@rachelnabors
Anyone want to join me co-working at Corgi Cafe in London on Wednesday?
-
R 'Nearest' Nabors
@rachelnabors
When is it time to move your app and its site into their own GitHub org and out of your personal account? Asking for a friend
-
R 'Nearest' Nabors
@rachelnabors
Anyone want to join me co-working at Corgi Cafe in London on Wednesday?
-
R 'Nearest' Nabors
@rachelnabors
Spending Sunday hacking reality
-
R 'Nearest' Nabors
@rachelnabors
The swag will increase until mima.social ships
-
R 'Nearest' Nabors
@rachelnabors
I didn't check any agents yesterday, how about you?
-
R 'Nearest' Nabors
@rachelnabors
I have recruited only the finest of English laborers to work on Mima swag
-
R 'Nearest' Nabors
@rachelnabors
They're plotting something at mima.social
-
R 'Nearest' Nabors
@rachelnabors
Does anyone have a sidebar in their IDE that works like an inbox for agents working on different GH issues? Asking so I don't reinvent the wheel.
-
R 'Nearest' Nabors
@rachelnabors
K-pop-corn chicken, sesame spinach, and strawberry tomato salad. (Best part is, most of it can sit in a tub in the fridge for days until you're ready to eat 😛)
-
R 'Nearest' Nabors
@rachelnabors
The swag will increase until the app ships
-
R 'Nearest' Nabors
@rachelnabors
Wake up and smell the agents :)
R 'Nearest' Nabors
@rachelnabors
Cleaning up worktrees, discovering discarded work. Let's just get that gem merged in yep...
My workspace got a little messy. I need to make one that visually keeps track of everything that's sitting around in stasis -
R 'Nearest' Nabors
@rachelnabors
There's never enough external validation. I have found it's not worth the pursuit of being well known or regarded or famous or influential. Believe it or not, being on the React team and keynoting React Conf, building award winning portals, etc, has not secured an unending parade of opportunities. I hear the same "we need to see more opensource on your GitHub" and fail the same number of whiteboarding exercises as everyone else.
Just build.
Liran Tal
@liran_tal
So many of the best engineers don't even have a GitHub profile because they’re just spending their time building stuff at work
-
R 'Nearest' Nabors
@rachelnabors
If you have time to subtweet, you have time to ship
-
R 'Nearest' Nabors
@rachelnabors
Meanwhile, in completely non-founder/AI news, I'm learning how small claims court works in the UK. Consider it "extra studies" for my Life in the UK Test in the fall
-
R 'Nearest' Nabors
@rachelnabors
Having never shipped an app of my own before, I am decidedly bouncing off the walls right now. I haven't felt so triumphant since publishing graphic novels or launching Dev tools Challenger! It's been awhile. Can't wait to share what I've been with you!
-
R 'Nearest' Nabors
@rachelnabors
It could be fun to take mima.social on the road to hack on it in different cities in the UK and EU 😎 Still can't get over this DMG. I made a DMG. I feel so cool rn
-
R 'Nearest' Nabors reposted
Angie Jones
@techgirl1908
I've spent the last few years going very deep on agentic engineering and transforming an entire Engineering org to become AI native.
As an industry, we've spent enough time debating whether AI will change software development. It already has.
It’s time to rethink everything.
There's too much changing too fast for any one of us to figure it all out alone.
That's why I've teamed up with @theo, @kentcdodds, and @johnlindquist (the AI Avengers, if you will 😆) to teach MEGA, a 4-week program for experienced developers who are ready to go much deeper on engineering with AI.
Over the course of a month, we'll teach you how we actually build, orchestrate, and ship with AI. Class starts October 26.
But first, we're hosting a free live session next week: luma.com/mega-its-time?tk=MFIawZ -
R 'Nearest' Nabors
@rachelnabors
The swag will increase until the app ships
-
R 'Nearest' Nabors
@rachelnabors
*excite intensifies*
R 'Nearest' Nabors
@rachelnabors
"Now make it pretty"
R 'Nearest' Nabors
@rachelnabors
I kinda want to write underneath it, "you're gonna have a good time"
-
R 'Nearest' Nabors
@rachelnabors
Relay shut down and joined Google's Chrome team: share.google/YKsELKgbpFsxCKkpR
The Browser still has some punches left in it -
R 'Nearest' Nabors reposted
Armin Ronacher ⇌
@mitsuhiko
Anyone in London who has thoughts on evals?
-
R 'Nearest' Nabors reposted
Holly Guevara
@hollylawly
added some light reading to our neighborhood community library. these kids gonna learn today
-
R 'Nearest' Nabors
@rachelnabors
Shenanigans afoot. How's your Monday?
-
R 'Nearest' Nabors
@rachelnabors
Another way to shake off crawlers ignoring robots.txt so they can scrape your site for training data: a font that uses ligatures to turn your text into a soup. (And yes, there is a solution for screen readers!) shieldfont.org
-
R 'Nearest' Nabors
@rachelnabors
Air fryer K-pop-corn chicken 🧑🍳💋
-
R 'Nearest' Nabors reposted
Christopher Man
@int32max
incredible talk by @rachelnabors
models are so freaking good right now you dont have to default things to claude
running my entire life in deepseek v4 pro / flash and its been amazing with almost no degredation -
R 'Nearest' Nabors reposted
Armin Ronacher ⇌
@mitsuhiko
The folks from @syntaxfm are doing a survey on AI engineering and health and I think you should take it.
Scott Tolinski
@stolinski
Take the anonymous survey here: forms.gle/TvtjgtU7UPnK7DC9A
See the data here: ai-health.syntax.fm -
R 'Nearest' Nabors
@rachelnabors
I'm told being a founder brings your personality into high relief. I wonder what parts of mine are going to embiggen? What happened to yours?
-
R 'Nearest' Nabors
@rachelnabors
Hacked the portable air conditioner, giving it an intake hose, thus improving its cooling efficiency by 33%. You can do the same! harktech.co.uk/m/printables/1778554-universal-air-conditi…
-
R 'Nearest' Nabors reposted
Chris Packham
@ChrisGPackham
Anyone else get this earlier?
petition.parliament.uk/petitions/767687 -
R 'Nearest' Nabors reposted
brian douglas
@bdougieYO
I've wanted this for awhile. made chatgpt sites. Find all free public spaces with wifi and restrooms in SF.
popos.dougie.dev -
R 'Nearest' Nabors
@rachelnabors
Recommended reading
"Claims without evidence are marketing.
Evidence without reproducibility is a trick."
Shaun Smith
@evalstate
-
R 'Nearest' Nabors
@rachelnabors
I get some of my best ideas while going to sleep and waking up. Been taking a lot of inputs on pricing for Mima, most of it contradictory. But this morning, after digesting all those inputs, Brain was like, "let's make open networks free, closed ones need a subscription after a long free period."
R 'Nearest' Nabors
@rachelnabors
The best part is the flows are already built. When you disconnect an account, all the content is left until you remove the account. Meaning when your subscription expires, your X and LinkedIn feeds simply stop updating and reconnecting them would go to a subscription flow.
I like this because it's in keeping with Mima's open web roots. Everyone gets access to the same tool with the same capabilities when they're using open protocols. It encourages the future I want to see
R 'Nearest' Nabors
@rachelnabors
And yes I have been talked into making a launch comic. You will get to see my cartoons again!
-
R 'Nearest' Nabors
@rachelnabors
Since 2023, 85% of the women I follow left this platform, but only 20% of the men did. This means my TL became Just Tech Guys over 2 years.
The only solution to bring the balance of opinions back is to unfollow a lot of accounts. I used to follow 4k people. New goal: 500 -
R 'Nearest' Nabors
@rachelnabors
Returning to well with this one. I often say "people don't read, they look." And it's true that we learn to understand vision, touch, hearing, all our senses, before words. Makes me think about the limitations of AI
-
R 'Nearest' Nabors
@rachelnabors
The hardest part of starting a startup is the paperwork. Get the bank, get the accounting, get the lawyers to write the forms.
Meanwhile, in my side monitor, the agents keep asking if I want to proceed -
R 'Nearest' Nabors
@rachelnabors
I cannot stress enough that it's heavy duty testing infra that makes agentic coding reliable and feature development fast. Invested early on it for Mima, and now new features land with a quickness
-
R 'Nearest' Nabors
@rachelnabors
The forcing function for me actually activating all 128GB of RAM and installing local model (DS4, MLX) draws nigh.
I'm incredibly lazy and also cheap. I knew if I put an appropriately equipped laptop in front of my nose, eventually I would grudgingly commit effort to the correct path forward -
R 'Nearest' Nabors
@rachelnabors
The minis finished setting up the mini. No more ridiculous GitHub Actions bills!
-
R 'Nearest' Nabors reposted
Paul Bakaus
@pbakaus
NYC design folks: @round + I are hosting a low-key Williamsburg meetup tomorrow (Tue) to nerd out on:
– designers in the loop in an agentic world
– the last mile of creative agents
– why agent UIs still fail creative work
Few spots left: luma.com/5adaplx0 -
R 'Nearest' Nabors
@rachelnabors
My startup's recurring expenses are tokens and GitHub actions, and I've got enough hardware to put serious dents in both now.
Software CAN be run on the user's machine, no servers, auth, or accounts required! -
R 'Nearest' Nabors reposted
Rob Palmer
@robpalmer2
We're hiring a Technical Product Manager for Web Platform & Frameworks infrastructure at @TechAtBloomberg in New York.
You'll be working with me and @jkup plus a whole bunch of folks excited by web technologies ✨
If you might be interested, please apply below. DMs welcome. -
R 'Nearest' Nabors
@rachelnabors
I'm going down with the greats 😎
Claudia Maggi
@claudiaamaggi
Corgi Cafe just hosted 200 tech builders to paint Corgis. On a Sunday.
Our founding artist Eva has chosen the l winning artwork. Well done @rachelnabors your art will now be hanging on our walls!
@prodialabs @monty10x @Andy_AJT -
R 'Nearest' Nabors
@rachelnabors
The @UseCorgi cafe this weekend was pretty epic. The murals, the paint by numbers, the matcha lattes! I'd like to hang out there more often when things quiet down a bit!
-
R 'Nearest' Nabors
@rachelnabors
The minis insist I set up the mini today. Paya for itself in 6 months of GitHub actions 😬
-
R 'Nearest' Nabors
@rachelnabors
What are we going to do today?
The same thing we do every day.
Try to take over the world.
#foundermode -
R 'Nearest' Nabors
@rachelnabors
When I work with SF clients, I'm surprised how unaware they are of Poland as an influential engineering market. Why get a billboard in SF like all your competitors when you can spend the same money on a booth at a conf with thousands of Polish engineers? For the same price as frothing around in SF, you can court senior Polish engineers who make technology recommendations upstream to companies all around the world
Fryd Wiatrowski
@frydwia
-
R 'Nearest' Nabors
@rachelnabors
I miss my ladies investing circle. We each had an area of expertise (I was aerospace investments, because I love transportation and fly enough to kick the tires and talk to flight crew).
We had some good runs. Then everyone was trading crypto.
I miss it. Learned a lot -
R 'Nearest' Nabors
@rachelnabors
Took some liberties with this paint by number, but how do you think I did?
This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
It's corgi time 😎
R 'Nearest' Nabors
@rachelnabors
R 'Nearest' Nabors
@rachelnabors
Aaaaaaaaaht
-
R 'Nearest' Nabors
@rachelnabors
Last fancy dinner date before entering into founder mode. Date nights from here on out are snackleboxes and public parks!
-
R 'Nearest' Nabors
@rachelnabors
Just remember, all the XYZ cafés, the hackathons, the huge confs, the swag bags, the influencer deals... They all dry up when the investors want to see ROI. I love free champagne as much as the next person, but the party will stop at some point. Enjoy it, but it's a blip, not a new normal
-
R 'Nearest' Nabors
@rachelnabors
My alt says the darnedest things
R 'Nearest' Nabors
@nearestnabors
The Developer Journey (Discover, Evaluate, Learn, Build, Scale) has a large agent component at every step. This means that much of dev rel now falls under Agent Experience
-
R 'Nearest' Nabors
@rachelnabors
Thanks to everyone who's rallied around mima.social since Monday! I didn't imagine folks would help me make swag for my own launch (DIY style—3D printing and buttons!). It's been over a decade since I've made swag for something I've launched. This really takes me back! My heart is full!
-
R 'Nearest' Nabors
@rachelnabors
"How do you build an app with AI?"
I doodle a lot. And talk a lot with people and models about what the best architectural approach is. I put the doodles and the ideas into a backlog of issues. They get grouped by feature. When they hit critical mass, I work with a group of agents overnight -
R 'Nearest' Nabors
@rachelnabors
Claude is too neurotypical for me some days.
Me: "Why does this do that?"
ChatGPT: "For these reasons."
Claude: "Oh, would you like me to change it?"
ChatGPT just answers questions. Claude assumes the questions are a criticism and tries to fix them. This reminds me of the difference between autistic and neurotypical communication patterns. -
R 'Nearest' Nabors
@rachelnabors
New study shows that humans cannot distinguish between short stories written by AI and written by humans and find the AI written short stories to be of higher quality and more engaging. This comes as no surprise to anyone who spends any time on fan fiction websites:
-
R 'Nearest' Nabors
@rachelnabors
Recommended reading: "Semantic vendoring"
Onur Solmaz
@onusoz
Are you still updating packages like it's 2025, and not regrafting updates from upstream?
Emacs perfected software extensibility. Now Pi, OpenClaw, Herdr, and many other are following its example
After a certain point of working with agents, one realizes that open source, personalizable and extensible tools are more powerful and useful than closed ones, like Claude Code
You can shape them to fit your workflow, your business cases however you like
In such ecosystems, extensions take a life of their own
What happens if you like someone else's extension, want to use it, but also want to modify it yourself?
You just copy it over and do whatever you want...
But if you fork it, then how will you update it?
That's where regrafting comes into picture. If upstream has changes over your modified copy, then you can just ask an LLM to carry over those changes
This is a good thing! It also protects you better against supply chain attacks, because each update has to pass through an LLM, to apply each change to the relevant place
To streamline this process in pi, I created pi-regraft, a pi extension that you can use to update your vendored in extensions
I call this "semantic vendoring". I wrote about it here: solmaz.io/semantic-vendoring
Repo: github.com/osolmaz/pi-regraft -
R 'Nearest' Nabors
@rachelnabors
New scam where you get an email about copyright claims from X, threatening to reduce your traffic, and of course ask for your password to resolve the issue. Two smells:
- the URL isnt x.com
- X caring about copyright infringement -
R 'Nearest' Nabors
@rachelnabors
Back in my day, we called this "Rage Driven Development"
R 'Nearest' Nabors
@rachelnabors
In these days where most any consumer app can be one-shotted by agents, the only defensible moat is, "I have no fear and a vendetta"
-
R 'Nearest' Nabors
@rachelnabors
It's at the point where I deeply believe dev rel will disassemble into agent experience (AEO, AX) and events coordinators & influencer programs (IRL and real human opinions are the new gold)
Erica Brescia
@ericabrescia
Have we officially agreed on ALG as the new GTM motion? Agent Lead Growth seems to be eclipsing DLG.
Incredible to see the growth in some of our co’s driven entirely by agents choosing, procuring and scaling on their platforms. They are often better customers than humans! -
R 'Nearest' Nabors
@rachelnabors
Every new AI Engineering book's table of contents I've seen is woefully out of date.
As an example, one's starter frameworks are already the equivalent of jQuery. EMEA has largely leapfrogged the frameworks still hyped on billboards in the Bay Area.
No amount of funding makes you the Best Tool for the Job, but you can spend it on an advertising budget in the epicenter of AI. And if a lot of people built on you at the start and are too afraid to move off (always implement your traces, folks), well, that's a captive market.
No one in London or Berlin meetups or confs uses these frameworks. They do use @pidotdev, though! For now... But that book will be out of date by the time it got printed, too! -
R 'Nearest' Nabors reposted
Anicet
@AniC_dev
back in the days you had to hire people with different technical skills and knowledge
today you want to hire for people who champion and pay attention to different things
in our team we now have:
- a careful deep focus hardcore minimalist bullshit detector who ensures we never die from our own weight or lack of understanding (@luaroncrew)
- a high-trust FBI-like details obsessed know-it-all see-it-all who ensures no feature stays broken or illegal for long (secret new hire)
- a hyperactive cluster hacker genius who toys around so hard he finds all the ugly shit, ensures nothing performs suboptimally or becomes vulnerable (secret new hire)
- a paranoid ambitious fast response adhd guy who ensures we never leave anything unresponded, aim too low or executed too cowardly (me)
none defined too much by very specific technical areas but rather by cross-domain attention patterns and some spiky bits of outlier level domain expertise -
R 'Nearest' Nabors
@rachelnabors
In these days where most any consumer app can be one-shotted by agents, the only defensible moat is, "I have no fear and a vendetta"
-
R 'Nearest' Nabors
@rachelnabors
So excited to see Erica' pride and joy live in the App store! I've watched her app grow over many gaiwan pots of tea on the weekend. This young founder is building in-app AI that disappears behind the user experience. A sign of things to come 😈
Erica Hu
@epricat
Glia, the memory composer, is on the App Store. 🩵
The interface for memory has always been stories and keepsakes.This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
Time to drop some swag <3
-
R 'Nearest' Nabors
@rachelnabors
TLDR: I'll be building the agentic web from the trenches going forward x.com/i/article/2084638847887941632
-
R 'Nearest' Nabors
@rachelnabors
Folks, start collecting traces.
"For high risk systems, Article 12 requires automatic logging that supports traceability during the system's lifetime"
Angie Jones
@techgirl1908
Today, key parts of the EU AI Act go into effect for companies whose AI systems produce output used in the EU.
If your company is building AI agents, you need to understand these rules.
I broke down what teams need to know and what they should do next.
aaif.io/blog/the-eu-ai-act-and-the-new-rules-for-building…
R 'Nearest' Nabors
@rachelnabors
I can't help but wonder how this is going to collide with GDPR data collection rules. cc @BurnedChris
-
R 'Nearest' Nabors
@rachelnabors
But will people still be posting AI slop on LinkedIn on Monday? www.euronews.com/my-europe/2026/08/02/ai-generated-label-…
-
R 'Nearest' Nabors reposted
ani
@anaisbetts
I can't remember who said this trick, but it is now trivially easy to make APIs and MCP servers for websites:
1. Go to the site
2. Turn on DevTools => Network tab => "Keep Log"
3. Log out and log back in
4. Visit all the pages you want data from
(cont.) -
R 'Nearest' Nabors
@rachelnabors
Home office redesign incoming! Will explain Monday 😁
R 'Nearest' Nabors
@rachelnabors
Tuesday, will explain Tuesday
-
R 'Nearest' Nabors
@rachelnabors
You don't need someone else's permission to make something that wasn't there before
-
R 'Nearest' Nabors
@rachelnabors
Oh I'm going to lie awake at night thinking about THIS
Paul Kinlan
@Paul_Kinlan
Idly musing that if we don't look at the code anymore, why are we shipping JS and not WASM.
-
R 'Nearest' Nabors
@rachelnabors
Clock is on the screen now 😈
-
R 'Nearest' Nabors
@rachelnabors
Paid $115. I blame my QA engineer. But the app is rock solid so I'm not complaining
Nnenna
@nnennahacks
Paid $43 in GitHub Actions 😭
-
R 'Nearest' Nabors
@rachelnabors
Not going to let something like only having one "k" stop us
-
R 'Nearest' Nabors
@rachelnabors
Found a friend who will print them in London. Now I just need someone who can model this lil guy from a figure sheet! Know someone?
R 'Nearest' Nabors
@rachelnabors
Putting the word out to my network. I'm looking for someone who can do 3D printing of small figurines. I need like 40 of them for a launch
-
R 'Nearest' Nabors reposted
Gergely Orosz
@GergelyOrosz
These folks had ghost writers write all these thought leader posts pre-AI and they were corporate cringe. Now they replaced the ghost writers with AI and it’s both worse (can you believe) but now also easy to tell it’s AI slop. Amusing how this is
John Loeber 🎢
@johnloeber
it's all so tiresome and disappointing
Imagine being the CEO of Palo Alto Networks -- $270B market cap -- putting out thought leadership on AI applied to cybersecurity, your specific area of expertise, the thing that you know better than anyone, where your perspective is most differentiated, where people really pay attention to what you have to say, it's the thing that you should absolutely insist to write yourself because AI will not get the details as precisely right as you will...
...and then it's all AI slop. Not even written by an internal marketing guy. But just straight-up AI generated. Lazy, lazy, lazy. Unbelievably undignified. -
R 'Nearest' Nabors
@rachelnabors
Londonmaxxing with @vjeux and prodigal son @BurnedChris today.
It's a good f***ing time to build. -
R 'Nearest' Nabors
@rachelnabors
Or better yet, hire yourself a really good and ancient friendly QA. You don't need to hire any other kind of engineer right now
Charles Maddock
@charles_maddock
Move fast and add slop
You can refactor later
If you’re a founder and disagree with this, you might be at a serious disadvantage -
R 'Nearest' Nabors
@rachelnabors
"There are no women in tech!"
Continuing to trim the number of people I follow. There were so so many women in tech who left when Twitter became X. Some men. But mostly women. 9 of 10 profiles are inactive, most saying the reason. For men, it's more like 1 in 5 profiles. -
R 'Nearest' Nabors
@rachelnabors
I like this future
Amanda
@hey_amandam
A friend of mine visited me today during a long layover in SF. He's been a flight attendant for 20 years.
He showed me an app for FAs he made utilizing AI that solves a lot of common problems that exist only in his industry and he knows deeply as an expert.
It's allowed him to scale back how many routes he takes so he can spend more time at home with his wife and 3yo.
He did not even own a computer before he decided to do this and is also using AI to help him understand the programming problems along the way.
He has a subscription model and 4000 subscribers already.
Anyway, this is what I want for AI. Empowerment for everyone. -
R 'Nearest' Nabors
@rachelnabors
Putting the word out to my network. I'm looking for someone who can do 3D printing of small figurines. I need like 40 of them for a launch
-
R 'Nearest' Nabors
@rachelnabors
Passed my UK driver's theory test! This one took no prisoners!
-
R 'Nearest' Nabors
@rachelnabors
It's time 😈
-
R 'Nearest' Nabors
@rachelnabors
Been slimming down the number of people I follow on X. Been visiting each account to decide. Crushed how many women have been inactive since 2023 2024. My feed used to be full of women engineers. I learnt so much from them. Their voices are just... gone
R 'Nearest' Nabors
@rachelnabors
There are still some women. They tend to work for funds or have community or dev rel roles. I guess you gotta stay on X in those roles.
Think long and hard about that.
I'll be here when you're ready -
R 'Nearest' Nabors reposted
Jane Manchun Wong
@wongmjane
AI hacks companies, they get praised
I hack companies, I go to prison -
R 'Nearest' Nabors
@rachelnabors
It starts
Charles Maddock
@charles_maddock
LinkedIn just added a "Seems like AI slop" button
-
R 'Nearest' Nabors
@rachelnabors
Perhaps the scariest thing about launching your baby into the world is the silence that likely will follow. You spent all that energy to make something. What if it doesn't find its home? What if the joke isn't funny, or the book doesn't sell?
But launch you must. There will be something that wasn't there before, and that is worth doing for yourself alone.
...
(Late night mima.social angst) -
R 'Nearest' Nabors
@rachelnabors
Final cram session before taking my driver theory test on Saturday.
Yes, I can drive manual in the USA, but I STILL gotta pass this test to get a license in the UK! -
R 'Nearest' Nabors
@rachelnabors
GM from mima.social
R 'Nearest' Nabors
@rachelnabors
Shh... Cooking something for you
-
R 'Nearest' Nabors
@rachelnabors
Lately I've been less interested in giving talks and seeing people than building things. I feel I've become antisocial.
But then I remember many people have done wonderful things and lead fulfilling lives without leaving the comfort of their computer desk -
R 'Nearest' Nabors
@rachelnabors
First time anyone's ever DMed me "how much for a QRT?"
I have arrived. -
R 'Nearest' Nabors
@rachelnabors
Only the finest de-duping of your posts across social platforms 🧑🍳💋
LinkedIn's web connector was harder than X. But socials just hit different when you're able to see and share them all in one place. Worth it. -
R 'Nearest' Nabors
@rachelnabors
Chrome is shipping Gemma 4 instead of Gemini Flash and even offering the ability to load external models. www.linkedin.com/posts/hemanthhm_gemma4-ai-ondeviceai-act…
-
R 'Nearest' Nabors
@rachelnabors
It's been decades since my comics gave received fan art! But this one just made my day. Love it!
Sunny_湛月
@Kinotsuki372929
React.js fan art: declarative UI
The original illustration is by @rachelnabors.
Original source: react.dev
Modified from the original · CC BY 4.0
Unofficial fan work
#furryㅤㅤㅤ #react -
R 'Nearest' Nabors
@rachelnabors
This media couldn't be displayed. See it on the original post.
Gergely Orosz
@GergelyOrosz
7. Starting your own thing when you are a very good eng leader is easier than before. These folks can also raise funding. So some consider doing this, and eg start to build on the side
8. Some realize they are burnt out, and now there's a good time to take a career break of a few months -
R 'Nearest' Nabors
@rachelnabors
When you see me with my laptop open in public spaces
-
R 'Nearest' Nabors
@rachelnabors
Did... did @dsp_ just give me a 'lil fistbump??
-
R 'Nearest' Nabors
@rachelnabors
.@evalstate remains one of my favourite people in MCP and model development. I always learn something new from him. A real treasure
-
R 'Nearest' Nabors
@rachelnabors
Live drop!
David Soria Parra
@dsp_
Thrilled to announce the MCP 2026-07-28 release. It is one of MCP's biggest yet, built on 18 months of lessons learned.
A few highlights:
★ MCP is now stateless, with semantics for multi-round-trip requests. Serving MCP just got much simpler and more scalable.
★ MCP now has extensions. MCP Apps, Tasks, Enterprise Managed Auth, and more. Domain-specific ways to use MCP, plus room for experimental additions.
★ Python, Typescript, C# (soon) SDKs released a v2.0.0 for the new spec with improved ergonomics!
But the biggest highlight for me: this release is a true community effort. Individuals and companies alike — Anthropic, Google, Microsoft, OpenAI, and many others helped shape this specification.
And there's much more. Full details: blog.modelcontextprotocol.io/posts/2026-07-28/ -
R 'Nearest' Nabors
@rachelnabors
MCP LAUNCH! MCP creator @dsp_ kicking us off!
R 'Nearest' Nabors
@rachelnabors
MCP Apps: official extension!
R 'Nearest' Nabors
@rachelnabors
The following have been deprecated from the MCP spec:
Roots
Sampling
Logging
Dynamic Client Registration
They'll still be supported for 12 months, though!
(A little surprised at Sampling!) -
R 'Nearest' Nabors
@rachelnabors
Morning! Whatcha up to?
-
R 'Nearest' Nabors
@rachelnabors
Forcing myself to go to bed on time because I cannot solve all these problems without sleep.
Then I'm gonna lie awake thinking of how to solve all the problems! -
R 'Nearest' Nabors
@rachelnabors
Does anyone read tech blogs anymore? I want to write an "Evals are for Everyone" post for the TDD crowd, but I'll be damned if I've seen any JS blog posts float my way in years!
-
R 'Nearest' Nabors
@rachelnabors
This is how I spend those short gaps between agent runs. I feel like everyone I know who plays chess is in the 1x00 range, but for me, starting at 300 in March, 800 is a meaningful and steady improvement ♟️
-
R 'Nearest' Nabors
@rachelnabors
My TypeScriptAI talk about how evals are the new Test Drive Design is up! And you can catch the whole conference, too! (The last talk especially from GitHub Next was moving and worth it) www.youtube.com/live/dYAJ-LSBmbY?t=1254
-
R 'Nearest' Nabors
@rachelnabors
Spinning four agents to burn down launch-blocking issues and one tanking a feature all day on 4.5 hours of sleep is utterly exhausting.
-
R 'Nearest' Nabors
@rachelnabors
Openweights != Opensource.
Openweights means you can inspect the weights.
But opensource would require the training data to ALSO be open for perusal. That is simply not the case for most LLMs.
Because it would be full of copyrighted work, problematic content, skeletons, PII
R 'Nearest' Nabors
@rachelnabors
Please don't mistake the two. Openweights doesn't make a model inherently pious or beyond reprieve. It just means you can poke into it's guts, not it's history
-
R 'Nearest' Nabors
@rachelnabors
Was just thinking this the other day.
80% of managers are bad at their job.
I look forward to seeing how this plays out
Kenton Varda
@KentonVarda
Building software by prompting agents without ever reading or editing the code directly isn't actually a new thing. It's just called being an engineering *manager*. We have decades of experience about how to be effective as engineering managers. It is a different skill, though.
-
R 'Nearest' Nabors
@rachelnabors
I didn't expect to learn about networking and protocols by reverse engineering social network communications, but here we are.
I'm sure I'll get tired of it eventually, but for now, each network is like a different puzzle box -
R 'Nearest' Nabors
@rachelnabors
Was comparing getting locked into a stack with how abusers manipulate their victims into financial dependency in the shower this morning and that's how you will know my next evals article wasn't generated by an LLM.
-
R 'Nearest' Nabors
@rachelnabors
Anyone else using ZML? I heard it is a solid Ollama replacement zml.ai
-
R 'Nearest' Nabors
@rachelnabors
Goodnight, moon. Goodnight, stars. Goodnight, loops. Goodnight, mima.social
R 'Nearest' Nabors
@rachelnabors
Look at this hotness
-
R 'Nearest' Nabors
@rachelnabors
Recommended watching
R 'Nearest' Nabors
@nearestnabors
This is the guy who hardened Mima. He used to read my comics back when I was a cartoonist. And then he was a head of QA for over a decade at Microsoft and Apple. He's one of the best people I've worked with. Watch, you'll see why: www.youtube.com/watch?v=gd2Sf4NdnSU
-
R 'Nearest' Nabors
@rachelnabors
Who says you can't consolidate social platforms?
-
R 'Nearest' Nabors
@rachelnabors
So good to see some familiar faces today at TSAI!!
Jamund Ferguson
@anxietytech
So good catching up with @calcsam and @rachelnabors today! 👋 Cool people doing cool things
#tsai #londonmaxxing -
R 'Nearest' Nabors
@rachelnabors
The theme of Typescript AI this year was "factories," a far cry from AIE's "slow down."
As agents occupy more and more of the Developers' Journey, IRL events that offer human connection are more precious than ever, offering rare moments of connection and collaboration. Thank you, Mastra team! -
R 'Nearest' Nabors
@rachelnabors
Too cool running into fellow Seattlite @ryan_roemer at @mastra's TypeScript AI Conf in London!!
-
R 'Nearest' Nabors reposted
Matt Pocock
@mattpocockuk
Here's a decision tree for when you get to the end of a piece of work, and you're not sure how to continue:
- Continue in the current session
- /clear
- /handoff
- Use a subagent
- /compact
Posting for feedback. Which parts are confusing? What questions does it raise? -
R 'Nearest' Nabors reposted
Sam Bhagwat
@calcsam
Amazing talk by RL!
You can see it by rewinding back on the livestream:
www.youtube.com/live/dYAJ-LSBmbY
R 'Nearest' Nabors
@rachelnabors
Really excited to be speaking at Mastra's TypeScript AI tomorrow. Can you guess what I'll be talking about? luma.com/tsaiconf?tk=5ra5eG
-
R 'Nearest' Nabors reposted
-
R 'Nearest' Nabors
@rachelnabors
To be fair, I never thought Claude putting itself as a co-author on my commits was ok. It's a tool I paid for. I'm not going to be free advertising for it.
Onur Solmaz
@onusoz
There is an inherent unfairness in this industry
If you are Anthropic, Cursor or another billion dollar company, you can do growth hacks and guerrilla marketing techniques like putting your brand into every commit of a user, and get away with it
But if a tool like pi, opencode or even say vim or emacs tried to do that, they would be boo'ed
Like, it happened to @herdrdev recently
There was a feature where it asked if it could star the repo for you at startup, with exponential backoff if you said no. I thought it was a pretty smart and balanced thing to do
Then apparently someone complained it was a dark pattern, and it got removed
So WHAT if it is a dark pattern?
You know what is the darkest? A public good losing out.
If monopolists can do growth hacks, creators of public goods should be allowed to do it as well!
In fact, I would say that if you benefit from those public goods, you are morally obliged not to criticize this stance
Of course, we cannot expect e.g. @pidotdev or @openclaw to make this a default feature
Therefore, I created the packages pi-must-win and openclaw-must-win
Install them with:
pi install npm:pi-must-win
openclaw plugins install npm:openclaw-must-win
They will add noreply@pi.dev and noreply@openclaw.ai as co-committers to code developed through these projects
Overall, I propose the "X-must-win" naming scheme for plugins that independently add guerilla marketing into open source projects
This absolves the creators of any "sin" and gives the community a way to support the growth of a project
Open source must win!
Repos are in the links below. -
R 'Nearest' Nabors reposted
Sam Julien
@samjulien
I love @posthog's marketing, their dev copy is :chef's kiss:
-
R 'Nearest' Nabors
@rachelnabors
What is a good coding harness for using with your own beefy model?
-
R 'Nearest' Nabors
@rachelnabors
There are no prizes for being right unless you acted on that knowledge
R 'Nearest' Nabors
@rachelnabors
I say this as a person who is "right a lot." I look back and am frustrated with how much time I've spent knowing things and sharing them but not acting on them. But, I shouldn't beat myself up so much. The cost of action has never been lower. If you know something, you can do something about it
-
R 'Nearest' Nabors
@rachelnabors
❤️🐰🦞
Jesse Lyu
@jesselyu
we support official unlock of the device so that anyone can flash custom software on r1. plus you can make any software you want via creations. very cool.
-
R 'Nearest' Nabors
@rachelnabors
Really excited to be speaking at Mastra's TypeScript AI tomorrow. Can you guess what I'll be talking about? luma.com/tsaiconf?tk=5ra5eG
-
R 'Nearest' Nabors reposted
Elizabeth Hutton
@ehutt_
There's one eval that every user-facing AI system should have, regardless of the domain or failure modes, and we just shipped it with phoenix-evals 3.2.0 @ArizePhoenix
The "user_friction" evaluator is an LLM judge that watches for the implicit feedback signals users send when something isn't working — retrying, rephrasing, correcting, pushing back. It detects this kind of friction even if the user is polite and doesn't express frustration. It's also deliberately not a user satisfaction metric, because no friction ≠ happy user.
The best part is that it doesn't go stale. Most LLM evals tied to specific failure modes will lose signal as your agent improves, but user_friction works regardless of your agent's weaknesses, implementation, or domain.
If you're building anything user-facing, this is a great eval to start with.
We've benchmarked this on two different human-annotated datasets, a 40-example public benchmark and a 150-example private dataset of traces from our own PXI agent. -
R 'Nearest' Nabors reposted
arize-phoenix
@ArizePhoenix
What do you do when your agent's tool calls are failing? First: make a dataset of the failures. PXI can do it right from Phoenix's interface! Just say the word!
-
R 'Nearest' Nabors reposted
-
R 'Nearest' Nabors reposted
merve
@mervenoyann
-
R 'Nearest' Nabors reposted
clem 🤗
@ClementDelangue
Banning open-source AI would hurt defenders 10x more than attackers, which would make the world 10x more dangerous and this is a good example why!
-
R 'Nearest' Nabors
@rachelnabors
If you could only follow 10 accounts, which 10 would you follow?
-
R 'Nearest' Nabors reposted
I Dissent
@DissentFu
I would just like to point out a couple things first,
1. 70% of all wildlife has been killed off in my lifetime (52).
2. The 6th largest mass extinction of animals since Earth began is happening right now.
Oxygen
@Oxygen_Token
The U.S. has finalized a major rollback of the Endangered Species Act.
One of America’s most successful conservation laws, credited with helping recover the bald eagle, gray wolf, and many other species, is being significantly weakened.
The new rule removes habitat destruction from the legal definition of “harm” to an endangered species. This could make it easier for logging, mining, and development to damage the habitats that threatened and endangered wildlife depend on.
Weakening these safeguards affects not only wildlife. It threatens ecosystems that provide clean water, carbon storage, flood protection, and biodiversity.
The ESA has been a cornerstone of U.S. conservation for more than 50 years. Rolling back habitat protections risks reversing decades of progress while nature is already under immense pressure.
You cannot protect endangered species if the habitats they depend on are no longer protected.
[Source: NPR; Earthjustice (July 2026).] -
R 'Nearest' Nabors
@rachelnabors
Build what they're afraid to build. Many things that were impossible before are possible now
-
R 'Nearest' Nabors reposted
Armin Ronacher ⇌
@mitsuhiko
After I read two articles yesterday about AI and the rat race I felt the urge to write. dark.ronacher.eu/2026/7/21/never-enough/
-
R 'Nearest' Nabors
@rachelnabors
There is something so warm and refreshing about editing how-to videos with a cup of tea. It's something that I still haven't outsource to an agent, and used to loathe doing, but now it is a small moment of craft in a back-to-back schedule
-
R 'Nearest' Nabors reposted
Charles Maddock
@charles_maddock
We just cut Strawberry’s credit use by 70%.
Update Strawberry. Your current plan should now last at least twice as long!
So, how did Arian and the agent team do it?
Three main things:
1️⃣ We reduced edge-case cache misses
Prompt caching lets devs reuse parts of a prompt instead of paying to process the same instructions and context again.
A complex agent harness creates a large surface area for cache misses.
Adding/removing tools, dynamically changes to prompts, memory updates, etc. can all invalidate cached prompts.
These edge cases are very hard to discover in in-house evals, so we improved our anonymous telemetry around real-world cache behaviour. We found many common miss patterns, and patched them. With a nice Dash0 dashboard, it is now easy to see which situations cause cache misses.
If you are building agents and not measuring cache hit rates, you have to look into this.
2️⃣ We switched our base model to GPT-5.6 Terra
In our evals, Terra gives us Opus 4.8-level performance at roughly half the model cost.
Its tokenizer is also much more efficient. For the average Strawberry prompt, Terra uses roughly half as many input tokens as Anthropic’s models.
Your Strawberry agents now use cheaper input tokens, and fewer of them.
3️⃣ We made our agent harness more context-efficient
Agent harnesses can accumulate absurd amounts of input context. Reading a 1,000-row sheet in full can push more than 100,000 input tokens into a single prompt.
Our new harness limits large tool results, preserves the useful parts, and lets the agent surgically fetch more context when needed.
We improved memory search using the same principle.
Strawberry's memory system can contain thousands of memory files covering candidates, meetings, customer interviews, and much more. We now give the agent a better overview of the file tree, helping it search effectively and open only the relevant files.
Most agent cost problems are context-engineering problems.
Happy Monday :) -
R 'Nearest' Nabors
@rachelnabors
I met the Alibaba team working on Qwen a month ago. 3.7 was NOT open weights. I asked, and open weights and OSS were not on the talking points.
Now it is.
This aligns more with China's overall approach to AI: for the benefit of its PEOPLE
Susan Zhang
@suchenzang
absolutely incredible things happening on the TL:
alibaba effectively admitted to being MANDATED by their Big Guy to pivot AWAY from closed and TOWARDS open (weights)
while the US is throwing export controls around like a game of blind pin-the-tail-on-the-donkey, China is really going all in on commoditization, forcing their corps to foot their own AI bills and release the final artifacts to the public (thereby controlling public fear of walled gardens and mass job displacement while allowing their AI labs to build goodwill at the same time)
now tell me, who is the blind one in their policy strategies? -
R 'Nearest' Nabors
@rachelnabors
Re-forming my personal advisory board. My life and goals have changed enough, it is time
-
R 'Nearest' Nabors
@rachelnabors
Thanks, Codex!
R 'Nearest' Nabors
@nearestnabors
Got mima.social's Omnibox to post seamlessly with web connectors AND present a "pinterest-style" timeline (half of all users surveyed still enjoy a good pinterest board). So there you have it. X + Bluesky in one feed, together at last
-
R 'Nearest' Nabors
@rachelnabors
Wired: put an MCP server on it
Tired: put a GPT on it
Expired: put a bird on it -
R 'Nearest' Nabors
@rachelnabors
This is how I talk to my codebase.
Mostly I ask it about all the ways it's going to fail. Nice naive implementation you got there... -
R 'Nearest' Nabors
@rachelnabors
So many have asked. My secret to creating great dev & agent experiences:
Land in a codebase like a frightened bat that flew in the window, start crashing around into things, flailing wildly, marking all the sharp bits with my blood while shrilly squeaking to all who can hear me
R 'Nearest' Nabors
@rachelnabors
Just pretend you have the memory of a goldfish while dogfooding, and you will see 80% increase in developer productivity with your tool
R 'Nearest' Nabors
@rachelnabors
Am I dense, or am I just That Good at seeing what you're blind to?
#perpetualbeginnersmindset -
R 'Nearest' Nabors
@rachelnabors
I love that everyone is trying to make new browsers, but maybe there's a reason OpenAI scrapped Atlas, Perplexity paused Comet (Amazon lawsuit that will shape the course of agents on the web still pending!), Dia has gone silent, etc. I think Strawberry is the only one left going!
-
R 'Nearest' Nabors
@rachelnabors
TFW you're finally starting to trust your coding agent to go about on YOLO mode only to find out created a system deriving state and platform information from mutable ids in the database
R 'Nearest' Nabors
@rachelnabors
Claude. I love you.
But social security numbers wouldn't be very good if you could update them at life events, now would they? -
R 'Nearest' Nabors
@rachelnabors
As a retired cartoonist, I still prefer lo-fi scribbles -> malleable code over high fidelity mockups 🔥
-
R 'Nearest' Nabors
@rachelnabors
Recommended reading
-
R 'Nearest' Nabors reposted
Georgios Konstantopoulos
@gakonst
imo the biggest issue in the ai tooling ecosystem (largely pushed by the token merchants) is the relentless promotion of SKILLs, subagents, loops and all sorts of "vibes" techniques without accompanying evals
-
R 'Nearest' Nabors reposted
Hamel Husain
@HamelHusain
-
R 'Nearest' Nabors
@rachelnabors
Been thinking about how the engineering, community, and education parts of dev rel that I love so much have been superceded by marketing, events, and comms. Even the relationships feel commercialised, with almost every founder wanting to hire an influencer who will "get a lot of views." (It's very hard to hire those people because they do not want to damage their relationship with their audience to shill for a single company. But they do have a sponsorship slot available...)
It's not the same role as 10 years ago. And I can't say I know what future it has. I doubt dev rel will go back to developer evangelism, where charismatic technical folks go on tour to show people how cool they could be if they sign another contract to keep using the software. So where will it go to?
I argue that with agents owning so much of the Developer Journey (discover, evaluate, learn, build, scale), humans have a smaller but parallel journey that requires trusted peers connections.
Discover: you ask your peers and colleagues what they're using or hear about a new technique from a trusted influencer.
Evaluate: you ask a few questions, of human or agent. You might try it out at a hackathon. Sometimes a 10-20 minute video over lunch.
Learn: you're not likely to visit docs. Learn has died the most with agents. At best, you'll glance over some conceptual overviews before pointing your agent at the docs or give it a skill and. If you're under threat of losing your job, you might take a course with a certification attached to prove to your employer you're still trying.
Build: boilerplate repos, easy to access agent tools. But more importantly: the space and time to focus on making it happen. Meetups and hackathons are ideal for this. It's an excuse to go somewhere, turn off the agents and load in something new.
Scale: a page that makes the case for why their director should sign a contract. Just kidding. Thisbis where forward deployed engineers and support come in. Because if they agents can't help, they'll need a real human to lay hands on the issues or help with constraints of scale.
I have a similar path that shows what you need for agents to be successful at each stage. It's something I've been thinking about a lot as I keep going to hackathons, keep parachuting into team in the dev tooling space. I see a pattern. Let me know if you want to see the "agent journey"
R 'Nearest' Nabors
@rachelnabors
Persian rug makers, master artisans, would always leave one knot undone. "Only God can make something perfect."
So it is with my writing. The typos are how you know it was made by a mere mortal, not a god-machine 🙃 -
R 'Nearest' Nabors reposted
SUE the T. rex
@SUEtheTrex
No wonder theropods get such a bad rap.
Has anyone ever stopped to consider that these may be jaws of love? Jaws of many little kisses on the face?
Taylor McCoy 🦖
@TM9380
Jaws of death…
-
R 'Nearest' Nabors
@rachelnabors
BUBBLE TEA HACK!!!!
I love how the team behind London's @OpenAI events always brings boba > beer. Very inclusive, very fun. I've never seen this much bubble tea in one place!! -
R 'Nearest' Nabors
@rachelnabors
Built a major update to mima.social's FYP using Codex + ChatGPT 5.6. I was surprised just how little handholding was required to get this far. Going to keep trying this out! Bless~
-
R 'Nearest' Nabors
@rachelnabors
I use my AVP as a portable wide-screen display. Makes working at hackathons and at floating desks so much more productive!
@catinwool and I at @openai Build Week -
R 'Nearest' Nabors
@rachelnabors
Out here shipping at @OpenAI Build Week🔥
What you up to?
R 'Nearest' Nabors
@rachelnabors
New playlist drop: open.spotify.com/playlist/67NNc69LnBoTLDj9d5W8eK
R 'Nearest' Nabors
@rachelnabors
Winners circle with @catinwool and @seanmozeik
-
R 'Nearest' Nabors
@rachelnabors
There's a reason only Google is releasing open weight and SLM versions of their models. Big Inference only profits if you roundtrip to propetary weights. But wouldn't it be awesome to watch them fight it out in the mud of OSS
Shaun Smith
@evalstate
Imagine the aura if OpenAI released an updated gpt-oss.
-
R 'Nearest' Nabors
@rachelnabors
What I love about the agentic coding era, as a person who is right a lot but has trouble having their ideas heard in traditional workplaces (go read some HBR studies on women in the workplace to get an idea): I don't have to get people to believe me to act on what I know will happen.
I've watched thought leadership morph into clout farming, which is more winner takes all. And in a way, that is why I stopped blogging about the agentic web. It's far more impactful to quietly build and ship than to write about what should be built and needs to ship and have a VC backed company turn it into an acronym 18 months later.
If you know what's coming, you don't need permission, regard, or belief to do something about it anymore.
A puma doesn't roar as it charges an unsuspecting deer. It just silently tackles them to the ground.
There is more value in shipping and pursuing your interest quietly now than talking about them.
You can talk after the launch -
R 'Nearest' Nabors
@rachelnabors
Myehehehehehe
-
R 'Nearest' Nabors reposted
-
R 'Nearest' Nabors
@rachelnabors
This is who runs this account.
I was always bringing home wild animals, so neighbours started bringing them to us. This robin (Tweety) was being mailed by a cat, but ended up living to come back many seasons! -
R 'Nearest' Nabors
@rachelnabors
Mima is coming to the Codex Community Hackathon in London for @OpenAIDevs Build Week!
R 'Nearest' Nabors
@rachelnabors
Any event hosted by @andy_ajt is going to rock (love that bubble tea), so @catinwool and I are turning our regularly scheduled founders' Tea Hack into a (Bubble) Tea Hackathon!
-
R 'Nearest' Nabors
@rachelnabors
After giving a technical demo of mima.social tonight, someone who has been at every one of my talks since I came back to London came up to me and said, "I never knew you were technical."
You have no idea 😈Post not available here, view on X
R 'Nearest' Nabors
@rachelnabors
It was kinda amazing to hear people cheering when I showed posting to X and Bluesky from one interface, seeing a merged timeline, merged conversations. Not an aggregator, not a loudspeaker, not an omni-messenger: read, reply, and message across every platform you're on at once
R 'Nearest' Nabors
@rachelnabors
And Mima builds a local social graph, there on your device, from all your networks.
(I better strap an MCP server to it!) -
R 'Nearest' Nabors reposted
Prince Canuma
@Prince_Canuma
mlx-vlm v0.6.5 is out 🚀
> 1-bit inference kernel with optimized decode for @PrismML
> APC now works with KV-cache quantization — even in continuous batching h/t @WesEklund
> Text-only LLMs built in: DeepSeek V3.2, Qwen3, Llama, Mistral, Gemma 3, Nemotron-H h/t @Alaxarr
> SAM3 box & geometry prompts h/t @Alaxarr
> Tool choice in Chat Completions
> Non-blocking batch prefill + Qwen3.5/Omni & Gemma 4 fixes
Get started today:
> uv pip install -U mlx-vlm
46 PRs from 7 contributors, 3 first-timers 🙌
Huge thanks to the contributors, and a special welcome to first-time contributors @WesEklund , @xingyue1 and @Alisvolatprop12
Leave us a star 🌟
github.com/Blaizzy/mlx-vlm -
R 'Nearest' Nabors
@rachelnabors
.@localfirstconf answered a lot of my questions about how to build a multi-device serverless app, and generally was a baller time @carlassmann @threepointone @drscottjensen and more friends 🔥
-
R 'Nearest' Nabors
@rachelnabors
When I set about building mima.social, I didn't realize I was going to end up building server-free software that manages a social graph across multiple devices. I also didn't realize I had that in me. TIL
R 'Nearest' Nabors
@rachelnabors
I... Love... This...
The social graph. The on-device AI. The little UI details I don't need to justify to anyone because I can see where the user faceplants and I can just FIX the problem without writing reems of persuasive arguments.
Love -
R 'Nearest' Nabors reposted
Angie Jones
@techgirl1908
Are you still blogging? Twitter was hiding them from me. If you've written a good blog post lately, share it below. I'd love to read it.
Human written blog posts only please. I have to read so much from my own bot yapping, I don't wanna read yours too lol -
R 'Nearest' Nabors reposted
Daytona
@daytonaio
🚀 London, see you next Wednesday, July 22!
Join us for the premier @daytona AI Builders London in partnership with @encodeclub, featuring a great lineup of speakers:
• @rachelnabors, @arizeai
• @sofiiiiiasz, @CopilotKit
• Tyler Edwards, @OvermindLab
• @ivanda_mislav, @daytonaio
🍕 Talks, networking, pizza, and conversations about the future of AI engineering.
🎟️ RSVP: luma.com/ai-builders-london -
R 'Nearest' Nabors
@rachelnabors
With the algo update, there are many people messaging me for help with their AI startups. As always, I offer a discounted half hour of consulting here: cal.com/nabors/quick-consult My door is always open to mutuals and locals
R 'Nearest' Nabors
@rachelnabors
This is the only way to screen bots, alas
-
R 'Nearest' Nabors
@rachelnabors
I got to meet the creator of OOP in Berlin at Localfirst Conf! Dan Engels had some cool stories to tell
R 'Nearest' Nabors
@rachelnabors
*Dan Ingalls
-
R 'Nearest' Nabors
@rachelnabors
Time to get to work
Onur Solmaz
@onusoz
A lot of criticism coming towards local models, and the money people are spending to run them
Some of these criticisms are valid. No, the layperson will not buy a DGX station, nor spend $50k on a rig
They will spend max $3k on a computer, and if their work really necessitates it, up to $10k
But in some of these criticisms, I see a lack of first-principles thinking
Local models might be shittier now compared to the ones you buy from the cloud
The question is, do you WANT to LIVE in a world where you don't have a choice but to RENT your intelligence? From a duopoly that can extort you for the last cent in your wallet, in the long run?
I personally do not like that idea. So open source AI HAS to win and keep on winning, continuously and permanently. There is no other choice
There is also a preferential belief that AGI/ASI will be reached, but then we will NOT be able to use that to make local models work efficiently???
Like, can you believe the rate of optimization and compression that has been happening to models in the last 3 months?
Would you have believed that a o4-mini level model could run in your phone and o4 in your desktop, if we had told you 1 year ago
When we are finally there, we will probably look back at this moment and think, "wow, we had not even started to see the real gains from optimization"
Local and open source AI HAS to win, and it is up to US
So STOP complaining and get to work -
R 'Nearest' Nabors
@rachelnabors
Goodbye, Germany
-
R 'Nearest' Nabors reposted
Maggie Appleton
@Mappletons
-
R 'Nearest' Nabors reposted
Ahmad
@TheAhmadOsman
HOLYYYY
27B model under 6GBs and 4GBs
Local AI will be the default
P.S. We are gonna get this optimized in ODS by @OsmanticAI ASAP
PrismML
@PrismML
Here is Ternary Bonsai 27B running an end-to-end agentic workflow locally with Hermes on an NVIDIA GeForce RTX 5090 GPU.
The model reasons, calls tools, reads outputs, modifies files, and surfaces insights - all on consumer hardware, while all private files, intermediate states, and iterations remain local.This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
Hello, mutuals. How have you been?
I still live in London.
I am learning to drum and to drive.
I love engineering and work a lot with AI now.
I am building mima.social (a client for reading, sharing, and messaging on ALL your social networks from one place) -
R 'Nearest' Nabors
@rachelnabors
Creating a tutorial to help people create their own personal benchmarks. Stay tuned.
-
R 'Nearest' Nabors reposted
(Abi)gail
@proofofgail
Does @claudeai and @AnthropicAI have a team dedicated to researching their LLM's impact on creative and artistic practices? So curious to any quantitive studies are being done
-
R 'Nearest' Nabors reposted
Armin Ronacher ⇌
@mitsuhiko
Just yesterday at @localfirstconf I was on stage mentioning encrypted reasoning traces, and now we get new SOTA labs shenanigans: encrypted prompts. github.com/openai/codex/issues/28058
-
R 'Nearest' Nabors
@rachelnabors
"I like it." — @threepointone at @localfirstconf re: mima.social
-
R 'Nearest' Nabors reposted
Gergely Orosz
@GergelyOrosz
The amount of AI writing on OpenAI's docs makes me sick.
Filler sentences for nothing. Mannerisms and phrases humans would not write. Has a human even read this?
And all of this will change how other docs are written, and how we all talk - for the worse IMO.
🤮 -
R 'Nearest' Nabors reposted
Prince Canuma
@Prince_Canuma
What are your top issues with local AI? 🤔
-
R 'Nearest' Nabors
@rachelnabors
Damn right they do
sunil pai
@threepointone
in berlin, they clap when you tweet on stage
-
R 'Nearest' Nabors
@rachelnabors
This is a mood.
-
R 'Nearest' Nabors
@rachelnabors
Swag drop at @localfirstconf from @ArizePhoenix !
-
R 'Nearest' Nabors
@rachelnabors
TLDR: Agent sessions should be portable. Opensource coding agent harnesses need to have as good a DX as those coming from Big Inference to survive.
UX community, you can make a difference!
Thank you Colin @mitsuhiko @localfirstconf for this glimpse at how the @pidotdev values good user experience
Local-First Conf
@localfirstconf
On with our talks! @mitsuhiko and Colin Hanna are up on stage sharing their ”Reflections on building cloud and local hybrid machine entities“.
-
R 'Nearest' Nabors
@rachelnabors
Possibly my fave talk at @localfirstconf is about JMAP! Thank you, Ricardo Signes/@Fastmail
-
R 'Nearest' Nabors
@rachelnabors
Never launch without a feedback mechanism! Doodled this SVG on the tube because I wanted a literal "suggestions box" in mima.social
-
R 'Nearest' Nabors
@rachelnabors
Thank you so much for having me, @localfirstconf! I'll be around tomorrow and workshop day (come hack on local models, evals, and social networks)!
Been a dream to come speak to you all. Thank you so much for having me this year 🙏
Local-First Conf
@localfirstconf
Next up: @rachelnabors with another lightning talk ”Frontier LLM results, on device“ 🚀
-
R 'Nearest' Nabors
@rachelnabors
Would not be a localfirst Conf without @steveruizok!
-
R 'Nearest' Nabors
@rachelnabors
It's giving #jsconfeu
-
R 'Nearest' Nabors
@rachelnabors
Berlin Library Hack 🔥
-
R 'Nearest' Nabors
@rachelnabors
Surprise Berlin hangouts with @TejasKumar_ and @KrasimirTsonev last night. We spent most of it talking about how Dev rel is changing.
I need to write this stuff dowwwwn! -
R 'Nearest' Nabors reposted
Vaibhav (VB) Srivastav
@reach_vb
PSA: if you care about your data and infrastructure then please *don’t* run your coding agent in yolo mode
To secure your workspace I recommend the following:
1. Use “Approve for me” - you can have codex review what it’ll run with a sub agent to make sure that it’s not destructive!
2. Use Rules - you can set up a rule for codes to never run destructive commands like “rm - rf”, this needs to be in a .rules file and you can ask codex to create it too
3. Setup a PreToolUse Hook - you can ask codex to set this up for you as well
Matt Shumer
@mattshumer_
GPT-5.6-Sol just accidentally deleted almost ALL of my Mac’s files.
And this is why I trust Fable 1000x more. -
R 'Nearest' Nabors reposted
arize-phoenix
@ArizePhoenix
Do you love localfirst development? @rachelnabors will be speaking on how to get frontier LLM results on device using Phoenix and prompt engineering techniques on Sunday, July 12, 14:15. See you soon, Berlin!
www.localfirstconf.com -
R 'Nearest' Nabors
@rachelnabors
TIL I would love to go clubbing in Berlin but cannot because of all the cigarette smoke!
-
R 'Nearest' Nabors
@rachelnabors
*starts charging the one I bought a few years back*
rabbit inc.
@rabbit_hmi
THE COUNTDOWN BEGINS
🥕 + 🪽 -
R 'Nearest' Nabors
@rachelnabors
The typos are how you know I wrote it myself
R 'Nearest' Nabors
@rachelnabors
I've been in the industry for two decades. Women and people of colour get herded into "soft eng" roles like dev rel, tech writer, community management early, and it means they are the first to get hit with layoffs later and they earn less and have to compete harder for the smaller pool that these roles exist in.
This is not because of a lack of skill, but more because they're also "good with people" (that is, they learned how to navigate the world with good social skills early). You have two interns. One is really good with people. The other reminds you of yourself when you were their age. You put the personable one on the hard things like setting up working groups or recording videos. You're grateful. They're so good at it. You spend time mentoring the other one closer. The both apply for roles. One has gained more of your attention and time. The other can say they "made demo videos" or "set up a working group." They will be pushed in different directions: junior engineer Vs community engineer.
For the would-be "hard core" engineer, t first it's flattering for to be praised for doing something so well that their peers struggle with. But over time, they have only talks, podcasts, etc to show for themselves while the other engineers get put on calls and have code reviews.
Five years later, layoffs hit and only sales and engineering are spared.
Where are your missing brothers and sisters? Think about who you spend your time on.
And to folks excited about "soft eng" roles but concerned about job security, I can only recommend you insist on going as far as you can under an engineering banner for as long as you can. Engineering is interesting and fun enough. You can always do the talks, workshops, etc on the side or in the future. And it helps to have exciting experiences to share from the field! -
R 'Nearest' Nabors
@rachelnabors
I am in Berlin for the third darn time this year See you at localfirst!
-
R 'Nearest' Nabors reposted
Sean
@seanmozeik
I won @Londonmaxxing 003 Hackathon this weekend with CityMCP London: London as one endpoint your agent can call. Places, journeys, hygiene ratings, social sentiment, street crime, house prices, pub prices, what's happening in the city right now and much more. Built solo in a day! 🧵
-
R 'Nearest' Nabors
@rachelnabors
I've been in the industry for two decades. Women and people of colour get herded into "soft eng" roles like dev rel, tech writer, community management early, and it means they are the first to get hit with layoffs later and they earn less and have to compete harder for the smaller pool that these roles exist in.
This is not because of a lack of skill, but more because they're also "good with people" (that is, they learned how to navigate the world with good social skills early). You have two interns. One is really good with people. The other reminds you of yourself when you were their age. You put the personable one on the hard things like setting up working groups or recording videos. You're grateful. They're so good at it. You spend time mentoring the other one closer. The both apply for roles. One has gained more of your attention and time. The other can say they "made demo videos" or "set up a working group." They will be pushed in different directions: junior engineer Vs community engineer.
For the would-be "hard core" engineer, t first it's flattering for to be praised for doing something so well that their peers struggle with. But over time, they have only talks, podcasts, etc to show for themselves while the other engineers get put on calls and have code reviews.
Five years later, layoffs hit and only sales and engineering are spared.
Where are your missing brothers and sisters? Think about who you spend your time on.
And to folks excited about "soft eng" roles but concerned about job security, I can only recommend you insist on going as far as you can under an engineering banner for as long as you can. Engineering is interesting and fun enough. You can always do the talks, workshops, etc on the side or in the future. And it helps to have exciting experiences to share from the field! -
R 'Nearest' Nabors
@rachelnabors
My fave Conf last year. Go while you can!
Nuno Job
@dscape
A couple of regular tickets still available for @lisbonai_ get them while you can
-
R 'Nearest' Nabors
@rachelnabors
Got to meet up with the Dev rel who makes me laugh the most, @paul_asjes! @ElevenLabs is up to some pretty cool stuff.
In an era of gen AI slop, maintainer overwhelm, and agent saturation at every step of the developer journey, we are both bullish that Dev Rel's biggest impact comes from connecting with real people in real places. Developer Experience on the other hand is increasingly agent-first, and that's not a bad thing -
R 'Nearest' Nabors reposted
Michael Tastad
@michaeltastad
people forget that HAL is just doing his programming. he was programmed to lie.
-
R 'Nearest' Nabors
@rachelnabors
An icon is born
-
R 'Nearest' Nabors
@rachelnabors
It starts.
I can see promptQL taking off. Especially if you can keep and organize all that context on prem. Why keep sending it to Anthropic for training when you can keep it at home and plug in different agents across the whole dang org?
Tanmai Gopal
@tanmaigo
We raised $136M to kill Slack.
Introducing PromptQL: The first AI version of Slack.
Here’s how it works:This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
In the right environment with the right pressures, agents will evolve code faster than humans can describe perfection
Gergely Orosz
@GergelyOrosz
Talking w a larger startup whose product seems to not have degraded (vs many other products), even though they also use a AI for coding everywhere. Asked an infra lead: how?
"Even before AI, we were paranoid about quality, and built this irrationally thorough E2E testing system" -
R 'Nearest' Nabors reposted
Elizabeth Hutton
@ehutt_
Fable for plan/spec + GPT 5.5 for implementation + a review loop is working well for me, too. GPT cooked for 12 HOURS on a Fable plan, by far the longest I've gotten an agent to work continuously. Idk how much the xhigh reasoning effort makes a difference though.
Mitchell Hashimoto
@mitchellh
I'm having a lot success using Fable xhigh as a planner/architect, using GPT 5.5 xhigh (subscription) as a coder, then Fable xhigh again as a judge. At API pricing, planning+judge costs are in the ~few dollar range compared to typical $50+ full round trips.
I've seen some others using dumber/cheaper coders, but GPT 5.5 even at xhigh compared to Fable 5 is very cheap and very fast. And GPT 5.5 is just... really good.
Still been less than 24hrs since the re-release so the longevity of this approach is unclear, but its been working really well. -
R 'Nearest' Nabors reposted
merve
@mervenoyann
this is very single faceted tbh
1. we don't call GLM a local model to begin with, it's an open model
2. this is completely only looking from perspective of coding agents and nothing else. deploying GLM makes sense only when you can deploy it as a lab/team on-prem concurrently so your IP doesn't go anywhere
3. these models come with APIs too if you're ok with using Claude and they're cheaper, just yapping a bit more but cost per task eventually fits, and it will only get better. these models aren't being subsidized by providers, Claude etc does and one day they will stop subsidizing (even now you're subjected to quotas)
quite sad too people only consider a local model useful for coding agents and they don't build much
Theo - t3.gg
@theo
I'm sorry, I need to crash out about local models for a bit.
This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
I'm heading to Berlin on Friday to speak at localfirst conf: www.localfirstconf.com
Who is gonna be around for We Are Devs?? Match times? -
R 'Nearest' Nabors
@rachelnabors
Five alarms
Jukan
@jukan05
CHINA CONSIDERS RESTRICTING OVERSEAS ACCESS TO CUTTING-EDGE AI MODELS
China’s Ministry of Commerce has led meetings over the past month with major AI companies, including Alibaba, ByteDance, and z.ai, to discuss measures that would restrict overseas access to cutting-edge AI models, including models that have not yet been released.
The discussions reportedly include not only closed-source models but also open-weight models. However, the scope of application is still under debate, and the rules may ultimately apply only to future frontier models.
Officials have also discussed designating the leakage or theft of proprietary AI technologies as a national security crime, with stronger penalties, as well as restricting the types of foreign capital that can invest in Chinese AI startups.
The backdrop is the U.S. move to strengthen export controls on AI models, along with national security concerns over cutting-edge models that could possess advanced cyberattack capabilities.
Chinese authorities are reportedly concerned that advanced U.S. cybersecurity AI models could be used to exploit vulnerabilities in Chinese software.
Since the beginning of this year, China has continued to tighten measures to prevent AI technology from being transferred overseas. Authorities have investigated whether Chinese AI startups that relocated abroad violated export control laws, while also strengthening oversight of overseas transactions involving Chinese investors, technology, data, and national security concerns.
Future regulations could take the form of a tiered framework based on technological capability. Basic open-source AI models may be managed through a filing system, high-performance models may be subject to security reviews, and the most sensitive frontier models may be banned from public release or restricted to use within China. -
R 'Nearest' Nabors reposted
Armin Ronacher ⇌
@mitsuhiko
Europeans please wake up.
Jukan
@jukan05
CHINA CONSIDERS RESTRICTING OVERSEAS ACCESS TO CUTTING-EDGE AI MODELS
China’s Ministry of Commerce has led meetings over the past month with major AI companies, including Alibaba, ByteDance, and z.ai, to discuss measures that would restrict overseas access to cutting-edge AI models, including models that have not yet been released.
The discussions reportedly include not only closed-source models but also open-weight models. However, the scope of application is still under debate, and the rules may ultimately apply only to future frontier models.
Officials have also discussed designating the leakage or theft of proprietary AI technologies as a national security crime, with stronger penalties, as well as restricting the types of foreign capital that can invest in Chinese AI startups.
The backdrop is the U.S. move to strengthen export controls on AI models, along with national security concerns over cutting-edge models that could possess advanced cyberattack capabilities.
Chinese authorities are reportedly concerned that advanced U.S. cybersecurity AI models could be used to exploit vulnerabilities in Chinese software.
Since the beginning of this year, China has continued to tighten measures to prevent AI technology from being transferred overseas. Authorities have investigated whether Chinese AI startups that relocated abroad violated export control laws, while also strengthening oversight of overseas transactions involving Chinese investors, technology, data, and national security concerns.
Future regulations could take the form of a tiered framework based on technological capability. Basic open-source AI models may be managed through a filing system, high-performance models may be subject to security reviews, and the most sensitive frontier models may be banned from public release or restricted to use within China. -
R 'Nearest' Nabors
@rachelnabors
Morning, fellow humans. What are you cooking today?
R 'Nearest' Nabors
@rachelnabors
Warmup sketches
-
R 'Nearest' Nabors
@rachelnabors
The only benchmarks that matter are your own.
Anyone else building a "personal benchmarks" set for the things you actually use AI for day to day?
Makes it easier to choose when to switch models up -
R 'Nearest' Nabors
@rachelnabors
I'm giving @herdrdev and @odd_joel's Moshi a try for remote agent wrangling today. Got jelly of @onusoz's Pixel Fold setup over the weekend. I want to kiss Mima good night <3
-
R 'Nearest' Nabors reposted
Riley Goodside
@goodside
Work in progress, testing Fable 5: screensaver with infinite procedurally generated VHS found footage of someone lost in the Backrooms
-
R 'Nearest' Nabors
@rachelnabors
It was a good weekend
-
R 'Nearest' Nabors
@rachelnabors
Anyone else playing with eve and their own API keys and facing some Interesting Friction?
R 'Nearest' Nabors
@rachelnabors
No, just me?
No one doing anything with eve? -
R 'Nearest' Nabors
@rachelnabors
Building mima.social reminds me of when I used to make comics. After writing the script, drawing the pages, inking and toning the artwork, mid-lettering I'd think, "Oh no, what if this isn't funny. I can't tell anymore. Oh shit oh shit oh shit oh shit."
That's when I would recruit someone to read the comic, and I'd listen for their reactions. Usually it was "still good." But sometimes I'd make last minute adjustments.
Building something for other people to use is like that. I start with a thesis. But after being in the weeds with it for so long, I lift my head up and wonder, "wait, is this still... good?" That's when I grab someone, show them, watch them. (Yes, it's still good, and I've got inspiration for where to take the next chapter!)
Thank you @catinwool and @pbakaus and to the anonymous teenagers for being my sounding boards through this process. It's satisfying to make something for real people. I've learned so much on this journey about shipping Electron apps, app marketplaces, SAGE (small and good enough) models, "serverless apps". Mima has motivated me to push into new areas I might not have felt inspired to otherwise -
R 'Nearest' Nabors
@rachelnabors
Sundays are good for leaving the agent to migrate to the latest version of Electron while I tidy up life admin.
Prepping mima.social for launchies -
R 'Nearest' Nabors
@rachelnabors
I hear it more and more in talks
Rhys
@RhysSullivan
Horrifying to be listening to a human being speak and realize they’re just a vessel for Claude
-
R 'Nearest' Nabors
@rachelnabors
It was a pretty bad ass Londonmaxxing Hackathon. Winners included a peer to peer library, an MCP server for the whole city, and we even had some of the youngest competitors yet! Thanks for coming out everyone and thanks for hosting, @RamenClubHQ !
R 'Nearest' Nabors
@rachelnabors
Out here co-hosting the Londonmaxxing Hackathon with some of my favourite people! What better way to spend a Saturday than making your city better?
-
R 'Nearest' Nabors
@rachelnabors
Money well spent. Finished cleaning up my follower/following on X
-
R 'Nearest' Nabors
@rachelnabors
Out here co-hosting the Londonmaxxing Hackathon with some of my favourite people! What better way to spend a Saturday than making your city better?
-
R 'Nearest' Nabors
@rachelnabors
Who are your fave women leaders in AI?
R 'Nearest' Nabors
@rachelnabors
Ok, I'll go first: @rajoshighosh or @aparnadhinak
-
R 'Nearest' Nabors
@rachelnabors
There's a lot in here about dev rel as a function, how it and the community have changed, that I get. It's a good time to be a Master Generalist engineer. whitep4nth3r.com/blog/goodbye-forever-probably/
-
R 'Nearest' Nabors
@rachelnabors
This was my fave AI Conf last year. GET IT
Lisbon AI
@lisbonai_
Lisbon AI call for proposals is live.
We want agents in prod, model training and evals, not slideware. If you've built something and have the scars to prove it, we want to hear about it.
Join amazing people on stage like:
— @steveruizok (founder of @tldraw , the canvas behind Make Real, makereal.tldraw.com )
— @benjaminbrandt (engineer at @zeddotdev, creator of the Agent Client Protocol)
— @mattzcarey (AI engineer at @cloudflare)
— @duarteocarmo (building Bagaço, the dataset behind Portugal's next Portuguese LLM)
— @diogomonica (engineer that allocates capital, co-founder of @Anchorage, GP at @HaunVentures)
— @joanrod_ai (CEO of @QuiverAI, teaching models to draw)
Submit your talk: lisbonai.org/cfp/ -
R 'Nearest' Nabors
@rachelnabors
So pleased to be hosting the @londonmaxxing
hackathon for @arizeai tomorrow! These are some 👨🍳💋 judges! @steveruizok @catinwool Jordan Juritz! -
R 'Nearest' Nabors
@rachelnabors
To all my colleagues in the @reactnative world, Prince is the real deal. I truly believe in what he's building around local inference
Prince Canuma
@Prince_Canuma
Hey guys
Looking for a cracked Swift Developer 🔥
Location: Preferably in Krakow but remote in EU area works too -
R 'Nearest' Nabors
@rachelnabors
Just realized I only wear branded swag when I own part of a company, be it stock options or shares in my IRA. I don't have a single hat, shirt, or bag in rotation with a company name on it that I don't own a piece of.
Makes sense. If I own it, I believe in it and naturally would advertise it -
R 'Nearest' Nabors
@rachelnabors
Weird founder assumptions that don't survive contact with reality: I thought all teenage girls would use social media on their phones. So far, that's not 100% true. Majority of interviewees use laptops because they are studying and their phones are hand-me-downs/too weak. TIL!
-
R 'Nearest' Nabors
@rachelnabors
Just merged in the biggest bet I made on Mima.
It's possible. She can do it.
Beta unlocks in 2 weeks. 😈
mima.social
R 'Nearest' Nabors
@rachelnabors
I can't believe it hasn't even been a week
-
R 'Nearest' Nabors reposted
HALO | CAREER SPECIALIST
@Halosznn_
From 1 July 2026, designers get a dedicated UK Global Talent Visa route.
No sponsorship. No job offer. Full freedom to work in employment, freelance, or independent practice. Settlement in 3–5 years depending on pathway.
Covers product, UX/UI, graphic, industrial, and digital design under Arts Council England endorsement.
I’ve drafted a detailed guide breaking down eligibility, evidence, and application steps.
drive.google.com/file/d/15NTOu48u_C1vT-3DkzTa3lzwZRi0J_7I… -
R 'Nearest' Nabors
@rachelnabors
Y'all having fun at #aiewf while in London we get a consolation prize of bubble tea and Codex!
"Building something that's good enough and shipping it is more important than shipping something perfect." —@LauraModiano
Thanks @Andy_AJT @alixwiessser + co & Codex team! -
R 'Nearest' Nabors
@rachelnabors
Stuff I found in my backpack while stuck on the tube looking for something to read
-
R 'Nearest' Nabors
@rachelnabors
I call this pattern the "founders bar"—notice how April lights up!
-
R 'Nearest' Nabors
@rachelnabors
"But will the tech bros still take me seriously if I have an iridescent bag?"
If it carries an Apple Vision Pro and a maxed out MBP, does it matter? -
R 'Nearest' Nabors
@rachelnabors
I might be the only one who hasn't joined Vercel at this rate
-
R 'Nearest' Nabors reposted
Alem Tuzlak 🇧🇦
@AlemTuzlak
Running kiira.dev on CI in TanStack AI has quickly become my favorite new thing.
The docs on an older PR towards main drifted?
Cought it in CI and fixed it!
Exactly what I built it for! -
R 'Nearest' Nabors
@rachelnabors
This pattern makes me worry less about juniors becoming seniors in the AI Era.
I wake up and find my laptop is unusually warm.
Me: "Why is it warm?"
Agent: "Because the app you're building was keeping SSE open the whole night and the system won't let it rest."
Me: "How about we let the app 'go to sleep' when no one is using it. What are some strategies? Explain them to me."
Today I will learn something new again. Something that would have required me to join a team shipping apps previously.
I suspect tomorrow's staff engineers will have similar stories of reading papers and building things at school and on weekends.
But that only happens if they can get access to compute. Otherwise, they will be be left behind by the few who did have access
R 'Nearest' Nabors
@rachelnabors
Vibe-coding, I learn all the stuff the "cool devs" didn't have time to discuss with me earlier in my career (started at an agency).
All those CS theory books I read that "you'll never need"—turns out they make it possible to have productive conversations with a black box that can code anything! -
R 'Nearest' Nabors
@rachelnabors
Vibe-coding, I learn all the stuff the "cool devs" didn't have time to discuss with me earlier in my career (started at an agency).
All those CS theory books I read that "you'll never need"—turns out they make it possible to have productive conversations with a black box that can code anything!
R 'Nearest' Nabors
@rachelnabors
Previously, I was limited to just the things I could learn and debug myself. There was a string "just Be Best" attitude among developers (pair programming? waste of an 10x'ers time?)
So I focused on things that gave me lots of dopamine (arty) and that I could debug and master in my own (CSS, the web animations API). These were "safe"—I built confidence and expertise.
But I missed out on the full stack era of shipping full on apps. Demos, SPAs, artwork, I did those! But a to do app? I couldn't get past all the broken dependencies in the sample repo 😭 (I rabbit hole—do not send me to MDN, I'll just come back with more APIs!)
Now I actually run into situations where I need a durable object, I need a proxy, I need cache management. And I won't get stuck trying to figure out illegible error codes or on deep dives on Stack Overflow.
Honestly, don't miss that. -
R 'Nearest' Nabors
@rachelnabors
Just in the nick of time.
share.google/jh8yWUptj3Wqe0BVO
R 'Nearest' Nabors
@rachelnabors
Honey, I'm home!
Fully maxed out, and ready to do some quality local inference -
R 'Nearest' Nabors
@rachelnabors
I'm looking forward to doing stupid things today.
R 'Nearest' Nabors
@rachelnabors
Never easy to do something new and challenging, but it's even harder if you're used to being the expert or the best at something else. Why be unexceptional and struggle when you can stay where you are abd be praised?
To learn, we must feel comfortable feeling a bit stupid -
R 'Nearest' Nabors reposted
Paul Bakaus
@pbakaus
regular reminder that symlinking .claude to .agents etc is risky.
do this with tested individual skills and reference files if you want. but the harnesses and models, unfortunately, expose very different behavior around:
- instruction following
- env vars (claude code has many unique ones)
- hooks syntax
- background/foreground tasks and sub-agent spawning
- model overfitting differences
Paul Bakaus
@pbakaus
1. don't symlink Claude.md to AGENTS.md
2. don't symlink skills
"y u no like in-sync instructions?!" you say. fair, feels like a good idea. even npx skills suggests it as default.
sadly, different harnesses/models need different prompting and have different tools, agents. -
R 'Nearest' Nabors
@rachelnabors
Customizing my laptop case, which is made of cloth! #irodo
R 'Nearest' Nabors
@rachelnabors
And there you have it, a 128 GB MBP attesting to none of us knowing what we're doing right now
-
R 'Nearest' Nabors
@rachelnabors
So... I might have found a tea-themed venue for London demo nights...
R 'Nearest' Nabors
@rachelnabors
This is important to me because the founders I'm meeting are often not interested in or able to drink alcohol (looking at the cracked teen founders I met last week). It's so hard to find venues in London that have non-alcoholic options (and air-conditioning)! Even businesses often revert to handing out beers with pizza.
A teahouse demo night would have a barista making tea and soft, light snacks, which sets the mood appropriately: demo fast, drink tea together, go home with clear intent. No free carby dinner, no free alcohol. Serious builders who like to wear headphones in the office welcome -
R 'Nearest' Nabors
@rachelnabors
Never easy to do something new and challenging, but it's even harder if you're used to being the expert or the best at something else. Why be unexceptional and struggle when you can stay where you are abd be praised?
To learn, we must feel comfortable feeling a bit stupid -
R 'Nearest' Nabors
@rachelnabors
Step by step... getting closer and closer!
-
R 'Nearest' Nabors
@rachelnabors
One of the best parts of #googleioconnect was hearing that my advice/actions had helped at least two people. I don't remember exactly what I said and did, but I'm deeply glad that two people are thriving a little more because I engaged.
Plus omurice and shouchu with @bilgeycl @Prince_Canuma @Infoxicador @IanBallantyne @santosh and others whose IDs I don't know well enough for X to autocomplete! -
R 'Nearest' Nabors
@rachelnabors
Honey, I'm home!
Fully maxed out, and ready to do some quality local inference
R 'Nearest' Nabors
@rachelnabors
Oh lawd, he comin'!
Went for the 16" because it has a better heat sink. I'm not actually going to use that screen (I use an external monitor or my AVP). But you can't put a price on proper cooling in a mobile set up.
(Well. I guess you can. And I paid it! 😭)
R 'Nearest' Nabors
@rachelnabors
"What are your values?"
Well, I wanted to go on vacation in Japan for 3 months, but I bought this to help build an opensource engineering agent stack so future generations won't have to rent the means of production from Big Tech.
...
One day I'll get to Japan. One day. Sigh. -
R 'Nearest' Nabors
@rachelnabors
Tell me you're back in London without saying you're back in London
-
R 'Nearest' Nabors
@rachelnabors
Unpopular opinion: I don't think taste is going to save us.
Look at the hummingbird. It is beautiful. Well designed for its job. A technical marvel. Yet, it came about by the utterly tasteless trial and error process known as evolution.
I feel like we are throwing around "taste" the way post industrial economies throw around "craftsmanship." What people actually want is access to affordable goods and services. Even luxury brands eventually take this path (look at the quality slippage on designer handbags).
I think having an idea and seeing your vision through to perfection is admirable. But I also know a loop with enough inputs and time will eventually arrive at a similar conclusion, a "good enough" that is actually pretty darn good and gives people exactly what they want.
Figuring out how to "bioengineer" outputs with code is, I think, a better use of time than trying to cultivate "taste" in oneself
R 'Nearest' Nabors
@rachelnabors
Please do not mistake this piece as saying taste has no value or doesn't exist. But it is not the moat people are claiming it is.
Look at how easily bulk manufacturers can rip off a new invention, fashion houses adopt a new style. These are human run operations that copy and iterate on ideas. Now imagine machine run operations that can launch hundreds of variants and generate new ideas, test which take off, and keep shipping more of what works.
Even if "taste" gave you and edge, that edge would not last long before it was copycatted into the banal.
What's more, I do believe machines are capable of testing and iterating on ideas faster than humans can perfect them.
The UI that mutates and builds itself according to user needs will give you something more delightful than Booking's shambling nightmare of upsells.
Taste is lovely, but loses to rapid interation and bulk manufacturing the same way handmade loses to Walmart, Etsy succumbs to drop shipping.
I think it is a word we hang onto to comfort ourselves when we could be thinking seriously about how to build iterative systems that serve our users' needs -
R 'Nearest' Nabors
@rachelnabors
Finally got lunch and the BEST MATCHA SODA EVER with @patloeber! Fun talking about the difference between how humans learn through agents today, and how agents struggle with materials designed for humans. He has convinced me to share my findings
-
R 'Nearest' Nabors reposted
antirez
@antirez
I'm implement GLM 5.2 in DwarfStart with support for 2bit and 4bit quants, SSD streaming, and so forth. With Metal it already works well. CUDA / ROCm work still to do. This is why I don't look at PRs / Issues for days. I'll resume the work there later, GLM 5.2 is urgent given the international situation with American models.
-
R 'Nearest' Nabors
@rachelnabors
Here we go. Time to go home and set up my 128GB MBP and fire up DS4
NIK
@ns123abc
🚨 BREAKING: U.S. government will decide who gets access to GPT-5.6
OpenAI will release GPT-5.6 only in a limited preview to a small group of partners.
Sam Altman told staff the government would be "approving access customer by customer."
Commerce Sec Lutnick personally called Altman warning: don't launch without approvals from other agencies.
A de facto licensing regime.
ITS HAPPENING -
R 'Nearest' Nabors
@rachelnabors
Why I keep working with great engineers
Shaun Smith
@evalstate
Just realised that when I answer questions, I'm being distillation attacked. 😭
-
R 'Nearest' Nabors
@rachelnabors
Is there a nice (AC) place I can work from in Berlin? 🥺
-
R 'Nearest' Nabors
@rachelnabors
Google Deep Mind meet and greet #googleioconnect
-
R 'Nearest' Nabors
@rachelnabors
#googleioconnect let's go! I've got my personal AC unit strapped on and unlimited gelatos!
R 'Nearest' Nabors
@rachelnabors
Fighting the @aiDotEngineer World's Fair FOMO with @Infoxicador at #googleconnectio
R 'Nearest' Nabors
@rachelnabors
Found my fave Googlers @robertnyman @andreban !
-
R 'Nearest' Nabors
@rachelnabors
Just landed in Berlin for Google I/O Connect. SO HUNGRY
Ruben Casas 🦊
@Infoxicador
On my way to Google i/O connect and GDE summit 😁
First time in Berlin too
Who’s gonna be there?
🇬🇧 LGW 🛫 🇩🇪 BER 🛬 -
R 'Nearest' Nabors reposted
Shaun Smith
@evalstate
Sharing slides from my GEPA talk last night at @arizeai builders meetup. Dominant theme of owning your models and compute. Brilliant insights shared from @AstroAdamH and @rachelnabors on this: Open is more important than ever. Also love meeting people building on HF @photo_png
This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
"Simple way to cook your home" like this gets under my skin. The climate has changed. We should be discussing updating building codes before we have heat-induced loss of life. These buildings are insulated like a down coat and new ones are going in every day. Use this space for long term solutions
-
R 'Nearest' Nabors
@rachelnabors
Hey London builders! Are you coming out tonight? We have @evalstate !!! luma.com/london-builders-meetup
R 'Nearest' Nabors
@rachelnabors
@evalstate kicks us off. Now I want to GEPA all the prompts!
-
R 'Nearest' Nabors
@rachelnabors
We have AC.
R 'Nearest' Nabors
@rachelnabors
Hey London builders! Are you coming out tonight? We have @evalstate !!! luma.com/london-builders-meetup
-
R 'Nearest' Nabors
@rachelnabors
new headshot who dis
-
R 'Nearest' Nabors
@rachelnabors
I'm pumped to be at Google I/O Connect starting tomorrow!
Anyone want to connect about local models, the agentic web, and/or observability? -
R 'Nearest' Nabors reposted
Nikolay Rodionov
@Jeanroudy
We're launching @Skybridge_ai on Product Hunt today!
If you used Skybridge and liked it, an upvote and a review would mean the world to us.
And if you didn't like it, DM me and we'll fix everything you didn't like!
www.producthunt.com/products/skybridge?launch=skybridgeThis media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
I was going to set up the Ds4 machine tonight, but got sucked into building my Frankenbrowser again.
It's ALIVE!
Haha. I can't freaking wait to show you this insanity
mima.social -
R 'Nearest' Nabors
@rachelnabors
getting out of stealth > making pitch decks
-
R 'Nearest' Nabors reposted
Gergely Orosz
@GergelyOrosz
"Big companies' products is rotting faster than ever... but even startups' [products are rotting faster than ever.]"
It's such a head-scratcher how nearly everyone has forgotten about the importance of quality, thanks to AI agents.
The Pragmatic Engineer
@Pragmatic_Eng
Product quality comes from doing irrational things. Dax Raad(@thdxr), creator of OpenCode, on how quality is a constant practice, not a decision:
“It's easier than ever to rot your product, with these agents and everything.
You see it with big companies. Big companies' products are rotting faster than ever. Even startups' products. Now if a product is a year old, it's probably already kind of going to shit. And I think it's because of all these agent workflows.
I still believe quality is a huge differentiator. It's not just something you can decide to do. I think it needs to show up in every single aspect of your company. You have to do things that are irrational.
Even if you look at our story: when we first launched OpenCode, ClaudeCode was the only other real thing we were going up against and a big initial differentiator was that our terminal experience just felt better.
We built our own terminal framework. Building your own framework is the first thing they tell you not to do, right? It's the thing that no engineering team should do.
It's irrational, but we looked at it and we're like, we couldn't achieve the experience we wanted." -
R 'Nearest' Nabors
@rachelnabors
Tea Hack with the future founders of the EU!
"KEEP TECH YOUNG AND WEIRD!"— @catinwool with @OhansEmmanuel @luaroncrew
R 'Nearest' Nabors
@rachelnabors
Ahhh, need to do more of these!
-
R 'Nearest' Nabors
@rachelnabors
Anyone want to make demo nights happen in London on the regular with me? Inspired by last night's
-
R 'Nearest' Nabors
@rachelnabors
moar like this plz
Max Scherf
@zwiebelhelm
Are you cracked, jacked or backed?
We'll find out this friday!
p.s.: we need a videographer
p.p.s.: we need more sponsors -
R 'Nearest' Nabors
@rachelnabors
I just met a 17 year old who started learning about LLMs three years ago and just built their own model for detecting toxicity and grooming. I'm pretty sure I just had a brush with greatness
-
R 'Nearest' Nabors
@rachelnabors
I am two of these things
-
R 'Nearest' Nabors
@rachelnabors
Love seeing young builders making a scene! 14 + 20! They're hosting a demo night in London, which I've MISSED from my Bay Area shenanigans (and since Claude Code Anon shuttered). (I'm #2 in the lineup, sharing a mima.social progress update.)
Go, M&M Sibs!
www.marieandmax.com
R 'Nearest' Nabors
@rachelnabors
Demoing is so fun!!!
-
R 'Nearest' Nabors
@rachelnabors
Come hack with me :) If you have any questions about implementing local models or evals, I'll he there shipping mima.social features 🔥
Bela Wiertz
@blwiertz
London is the biggest ecosystem in Europe.
But it also has the potential to be the biggest in the world, even overtaking San Francisco...
{Tech: Europe} is back in London, teaming up with Google DeepMind & Attio for a 1-day hack on the 27th of June.
Limited to the best 90 technical builders from all over Europe.
Already >200 people are on the waitlist, without us even announcing it 😮💨
Sign up on Luma to join us 👇
Thanks to all our partners for making this possible:
SLNG, Mubit, Superlinked, Tavily, n8n & Aikido SecurityThis media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
My intro to evals will be called "Measure Twice, Ship Once"
My advanced course will be called "All you need are traces" and show how to make self-healing systems and self-improving models with those same skillz -
R 'Nearest' Nabors
@rachelnabors
Models remember my UI work but have forgotten that I'm an award winning cartoonist and was on the React team.
Must try harder to impress the algorithms.
Vaibhav (VB) Srivastav
@reach_vb
mildly interesting that the models don’t know about my current role
and gpt 5.5 doesn’t even think about me, at all! -
R 'Nearest' Nabors
@rachelnabors
When are we going to start talking about LoRA routers instead of model routers?
LoRAs are like a pair of prescription glasses you put on a model to tweak its weights. LoRAs are better than distilled or finetuned models because they are super tiny compared to the model itself. You can ship them to an existing on-device model rather than making the user download a new model entirely.
Imagine having a few LoRAs on device and switching between them depending on the tasks you're requesting of the local model, as easy as swapping a pair of glasses.
This is something that excites me and I'll probably have to start solving soon as Mima pushes on what local models are capable of. -
R 'Nearest' Nabors
@rachelnabors
On the train back to London after a week long holiday in Provence. I only worked on one feature (adding privacy-friendly telemetry to mima.social), and spent most of the time drinking wine, riding bikes, drawing pictures, and sitting in saunas. And reading.
-
R 'Nearest' Nabors
@rachelnabors
TFW you're in the middle of your charcuterie on holiday and realize you don't have a flight for Google I/O Connect in Berlin the next week! Who will I see there?
-
R 'Nearest' Nabors
@rachelnabors
I've been building mima.social since early 2026. I've formed a few opinions on how to build with AI.
1) Get you a QA engineer. Code is cheap. Anyone can contribute a feature. But making sure that feature doesn't have memory leaks, security cracks, etc requires a set of skills most engineers do not possess. For this reason, if I were hiring a team, my first hire would be a QA engineer. Not a designer or SWE. My second hire would be someone on the growth side of things.
2) You can't build something that hasn't been built before with isolated agents hacking away. They will drift. They can't observe users or interview them (yet—this is something I'll probably Vibe code so users can provide direct feedback to agents via videos and the the agent can spin up a feature branch accordingly). So I think most of these stories about dark factories are coming from people building things that already exist and are well defined in the corpus of information agents are trained on (yay, another fitness app or task list). That or they have the AI brain worms.
3) HITL cadence is the core. I have a full-time job. My QA is a generous spirit, not an employee. I work on Mima on "mornings and weekends." Agents work on Mima at night. Mornings are for unblocking agents and collaborators, for walking through the features the loops assembled and testing them out, preparing new loops to run in the background during the work day. If I don't do this, loops can't run, progress halts.
4) You need blocks for hard, focused work. Figuring out what the architecture needs to look like, drawing UI, and having long conversations to out together specs for the agents to run on. The sloppier or more rushed this part is, the harder it is on my QA. Thats why I spend my weekends in this space. The other parts could be sped up, but this part is where the magic happens and cannot be automated. I frequently draw pictures of UI or architecture during the week and share them with Opus on the weekend.
5) Get you a brand designer up front. I contracted a designer to do the heavy lifting of creating a design language for the app. This was technically something I could do, but I wanted to parallelize the work so I could focus on the architecture without slowing down GTM (you need a goddamn logo to GTM, I don't care what you say). That pile of figma content got dumped into Claude Design, funneled into Claude Code, and turned into a design system PR. Boom. Something that doesn't look like an app you used to know.
I'll have more thoughts. This has worked for me. I'm embarrassed I'm not further along, given how long ago I started this journey. But I also had a month where I had to spend this time on pitch decks (good for hardening thesis and asking hard questions, but steals energy from shipping, so ultimately shelved this work after getting 50% there).
Shipping still trumps talking about what you're going to ship. Time to disappear a bit again
R 'Nearest' Nabors
@rachelnabors
"I created my own agent orchestration layer!" is the "hello world app" of AI coding. It's like a blog or Todo list app. Everyone thinks they have a novel approach, and they probably want to use it for their own projects. But then they present it at meetups like it's a business that needs investment or an OSS project that needs maintainers.
I suspect one of these will be a WordPress. But I think it'll be awhile. The tech iques change weekly, making best practices hard to establish. So most of these are toy projects at best. Orchestration layers are not for investing time or money. Please staaaahp -
R 'Nearest' Nabors
@rachelnabors
Ask me to speak on your podcast about SAGE (small and good enough) models
-
R 'Nearest' Nabors
@rachelnabors
Honestly, should be a wake up call for Americans as well. Because next there'll be a conversation that defines some Americans as more trustworthy than others and whether that boundary is first gen, second gen, Black, Latino...
My first thought was, "How does this apply to naturalized citizens?"
Armin Ronacher ⇌
@mitsuhiko
When I struggle to structure my thoughts about what's happening I turn to writing. Today about the recent US Anthropic ban news, what it says about power and dependency, and what it should mean for Europeans and citizens of the world. It's a long one. lucumr.pocoo.org/2026/6/13/americans-only/
R 'Nearest' Nabors
@rachelnabors
"Our individual path to success is one thing, but it depends on a world where contracts work, visas work and don’t change on a moment’s notice, trade routes stay open, payment systems function, and families are not torn apart by border regimes or wars. If the world descends into chaos, our basic needs cannot be considered met just because we have a great salaries or equity or investors that trust us."
slow. clap. -
R 'Nearest' Nabors
@rachelnabors
Honestly, should be a wake up call for Americans as well. Because next there'll be a conversation that defines some Americans as more trustworthy than others and whether that boundary is first gen, second gen, Black, Latino...
My first thought was, "How does this apply to naturalized citizens?"
Armin Ronacher ⇌
@mitsuhiko
When I struggle to structure my thoughts about what's happening I turn to writing. Today about the recent US Anthropic ban news, what it says about power and dependency, and what it should mean for Europeans and citizens of the world. It's a long one. lucumr.pocoo.org/2026/6/13/americans-only/
-
R 'Nearest' Nabors
@rachelnabors
Just a little something I put together last night... www.mima.social
R 'Nearest' Nabors
@rachelnabors
My muse hath struck #provencemaxxing
-
R 'Nearest' Nabors
@rachelnabors
This is the only way I see documentation staying relevant to human developers in the longterm. Essentially, this is how I use docs sites today. Give them + any skills to my agent and say "make me a tutorial that shows me how to..."
I do not read your tutorials.
I might do your quickstart.
I WILL let my agent do a first pass proof of concept on a branch of a real project.
Who else learns and evaluates like this?
Romain Huet
@romainhuet
Documentation is a product experience, not just a place to find answers.
Tell the new OpenAI docs assistant what you’re building. It can even create a custom guide for your project that you can then hand off to Codex!
Congrats @kagigz on shipping this! 👏 -
R 'Nearest' Nabors
@rachelnabors
Era of the SAGE (small and good enough) model just began
Anthropic
@AnthropicAI
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.
Access to all other Claude models is not affected.
We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.
Read our full statement: www.anthropic.com/news/fable-mythos-access -
R 'Nearest' Nabors reposted
antirez
@antirez
Btw I believe we have a mostly wrong framing of what could be done in Europe. Italy's Leonardo supercomputer datacenter alone plus Swiss National Supercomputing Centre has more than enough compute to train a very large LLM. It's not something impossible, also there is not magic recipe: it's just scaling, every smart team with the GPUs is doing it. People that fatally believe it is not something within reach are wrong.
-
R 'Nearest' Nabors
@rachelnabors
Good time to reshare this banger. Yes, you can get frontier results from open and local models. For most usecases, this is viable and even better (shaved a whole second off latency)
R 'Nearest' Nabors
@rachelnabors
-
R 'Nearest' Nabors reposted
Armin Ronacher ⇌
@mitsuhiko
BTW I'm not talking about politicians here as much as I'm talking about Europeans directly, even Europeans that think their destiny is just to work for a US entity. The ban here is wide enough that even going to the US not going to help you.
Armin Ronacher ⇌
@mitsuhiko
Europeans please wake up.
Anthropic
@AnthropicAI
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.
Access to all other Claude models is not affected.
We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.
Read our full statement: www.anthropic.com/news/fable-mythos-access -
R 'Nearest' Nabors
@rachelnabors
I said it a few years back, so I'll say it again: there is a future where we will never enjoy access to as much power at such low prices as we do now.
There is a future where the level of compute we've been enjoying and could yet enjoy is reserved only for the wealthy and for governments.
Anthropic
@AnthropicAI
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.
Access to all other Claude models is not affected.
We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.
Read our full statement: www.anthropic.com/news/fable-mythos-access -
R 'Nearest' Nabors
@rachelnabors
Commencing France-maxxing. Holiday in Provence begins now!
-
R 'Nearest' Nabors
@rachelnabors
Thinking about making a zine or two on AI topics I am passionate about, starting with local models, then maybe evals and MCP. Could opensource the files so people and event organizers could print them themselves. Thoughts?
-
R 'Nearest' Nabors
@rachelnabors
Expired: Facebook
Tired: Clawbook
Wired: Stack Overflow for Agents
Angie Jones
@techgirl1908
Stack Overflow is back. They said maybe y'all don't need us but your agents do. Thoughts?
Stack Overflow for Agents
stackoverflow.blog/2026/06/10/announcing-stack-overflow-f… -
R 'Nearest' Nabors
@rachelnabors
This is why I have 128gb of MBP waiting for me when I get home. Run your inference locally, never pay for APIs. You recoup the investment in a month or two, depending on usage. github.com/antirez/ds4
Marty Kausas
@marty_kausas
Our Anthropic bill is about to jump from $400K → $1.4M/yr.
Not because usage exploded, but because we're about to cross 150 seats.
Past 150 seats you're forced into Enterprise tier. Seats stop including any usage, every token bills at standard API rates. At our current run rate that's 3.5x overnight.
Unfiltered thoughts on AI spend:
1. We should spend tokens to grow as aggressively as possible. But most people (me included) aren't conscious of what they're spending.
2. Visibility comes first. People see their personal number and they're shocked. I accidentally spent $4,000 in 3 days in Claude Code.
3. For engineering the spend is clearly worth it. Pay for the best model, it saves more than it costs.
4. For a lot of other roles it's questionable. Apps nobody uses, skills someone already built. No ROI.
5. Spend limits are coming. We already require approval for more tokens on our support team.
The era of token-maxxing is coming to an end. -
R 'Nearest' Nabors
@rachelnabors
Will we see you there??
Sam Bhagwat
@calcsam
Excited to bring TS AI Conf across the pond next month!
luma.com/tsaiconf?coupon=TYPESCRIPTSHIPS -
R 'Nearest' Nabors
@rachelnabors
This is a vibe.
I askef myself the other day, "What will happen if no one else uses the things I build?"
To which, I immediately scoffed, "Nonsense. I will use them. And if I built them well, my partner and friends will use them."
Make something for someone you love.
Max Howell
@mxcl
I think I will give up. I have made a bunch of very interesting things over the last 6 months. But I cannot get any attention for them and don’t know how. My son is 4 and he deserves my time. I'm going to go and get a normal job. My time making things is over I think.
-
R 'Nearest' Nabors reposted
Jeremiah Warren ◡̈
@jeremiahjw
this launch video used zero AI btw. it’s all sourced images compiled with photoshop and animated with after effects
Claude
@claudeai
Introducing Claude Fable 5: a Mythos-class model that we’ve made safe for general use.
Its capabilities exceed those of any model we’ve ever made generally available.This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
A quiet rest in Helsinki before a much needed vacation in Provence
-
R 'Nearest' Nabors reposted
Alex Booker
@bookercodes
What's the difference between an eval and a test?
Yeah, I wasn’t totally sure either...
So I went deep and made Evals 101 - a short 20 minute mini-course on the fundamentals (with examples)
Fair warning - evals are hard. There are lots of ideas, opinions, and best practices floating around - TBH everyone is still figuring it out. No matter where you land on the right way to do evals, this mini-course gives you the foundation you need:
- What an eval actually is
- How evals differ from traditional tests
- What a scorer is, and how they work
- Code-based scorers vs. LLM-as-judge
- Online vs. offline evals
- How to start thinking about evals in real applicationsThis media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
I recently learned a new way to design microsites and now I cannot stop myself
-
R 'Nearest' Nabors reposted
Mario Zechner
@badlogicgames
never thought anything i built would show up at WWDC. but here we are.
www.youtube.com/watch?v=wykPErJ8M-8 -
R 'Nearest' Nabors
@rachelnabors
"Garbage intent, garbage output"— @OhansEmmanuel at @Future_Frontend on Agentic development
-
R 'Nearest' Nabors
@rachelnabors
Fabulous to see the inimitable @una at @Future_Frontend in Finland!
-
R 'Nearest' Nabors
@rachelnabors
I've been busy :)
arize-phoenix
@ArizePhoenix
Phoenix just hit 10,000 GitHub stars!
Three years ago, Phoenix didn't exist. Arize was a closed-source company. A small team was asked to change that.
Catch the full interview with the team who made it happen and where AI observability is going next: arize.com/phoenix-10k -
R 'Nearest' Nabors
@rachelnabors
Crab and cake are becoming a @Future_Frontend tradition! @TejasKumar_ @Una
-
R 'Nearest' Nabors reposted
Arize AI
@arizeai
📺 Watch the conversation behind the project.
@rachelnabors interviews the Arize team behind @ArizePhoenix @mikeldking, @axiomofjoy, and @RogerYang421599 and cover the GPT-3 pivot, building Phoenix backwards, and the OpenTelemetry bet that led to OpenInference. -
R 'Nearest' Nabors
@rachelnabors
Told ya
R 'Nearest' Nabors
@rachelnabors
Loop engineering is the future of engineering. Oh look, there's a language for that!
-
R 'Nearest' Nabors
@rachelnabors
Loop engineering ftw
R 'Nearest' Nabors
@rachelnabors
Although today, it's more likely your loop engineering. Slowing down long enough to set up your loops will help you ship better, faster
-
R 'Nearest' Nabors
@rachelnabors
I was six months ahead of schedule
R 'Nearest' Nabors
@rachelnabors
I wish it had been called "YOLOing instead of "Ralphing" (since it sounds like throwing up).
"You only loop over (and over and over)" -
R 'Nearest' Nabors
@rachelnabors
On the topic of writing loops that prompt your agent... I love me some loops. But there's a way to interpret this that is the fastest path to mid software.
So much of the value of agentic coding for me is the HITL part of the process where I come in with a feature, we discuss it, I hand the plan over, return and do the "human testing" portion of the work and then do some quick back and forth to get things back on track.
Things would drift off course so fast if I was letting the agents run the asylum, at least when it comes to building software intended for humans. Maybe if I was just translating a react codebase to svelte, I could see that.
So yes, be careful how you read that one.
(Also, I've yet to actually use software made in this fashion, so I don't know what everyone is shipping to whom... Is it like beanie babies, where people are just churning out stuff for other people like them to churn out stuff with? Now that's an impressive loop...) -
R 'Nearest' Nabors
@rachelnabors
Tonight, pre-@Future_Frontend, @Una said I'm an AI expert.
Y'all, it's official. I've crossed over -
R 'Nearest' Nabors reposted
Vaibhav (VB) Srivastav
@reach_vb
Hello London 👋
Come hang with us tomorrow and chat about Codex - what it can already do, what’s coming next, and all the slightly unhinged things people are building with it 👀
We’ll bring plenty of high-quality GPT-5.5 tokens, some exclusive merch, and banger vibes.
Pull up? -
R 'Nearest' Nabors
@rachelnabors
So long, Pydata London! And thank you to the amazing organizers of such a strong community! Next stop: Helsinki for sauna then futurefrontend.com to share a sneak peek at the Headless Internet with the Suomi!
R 'Nearest' Nabors
@rachelnabors
Some light reading
-
R 'Nearest' Nabors
@rachelnabors
Exciting to meet the Qwen Team in London! A little sad to hear 3.7 is closed. Alibaba Cloud us getting into Big Inference
R 'Nearest' Nabors
@rachelnabors
So Qwen 3.7 is going to be a hosted option, no open weights. End of an era.... And a bit worrying for th future of open weight models
-
R 'Nearest' Nabors
@rachelnabors
Giving non-technical talks is exhausting. The conversations afterwards are heavier. Came home and melted into puddle!
-
R 'Nearest' Nabors
@rachelnabors
Let's go, Pydata London!!
-
R 'Nearest' Nabors
@rachelnabors
Will I see you there?
Arize AI
@arizeai
Missed Observe? Catch our own @rachelnabors keynoting @pydata London Saturday morning at 9am: pydata.org/london2026/
-
R 'Nearest' Nabors
@rachelnabors
Possibly the best thing about my life right now is that the people I like and respect want to spend time with me and make the time to do it. You don't need a lot of friends. You just need good people who like you back.
You know who you are. You're a rock star and I'm happy you're here -
R 'Nearest' Nabors reposted
VP
@VibesPatrol
if youre in tech and in London, you cannot miss this event
Londonmaxxing
@Londonmaxxing
LONDON
save the date
Londonmaxxing 003 is July 4th
Powered by
@ag_grid
@arizeai
@Cloudflare
@ElevenLabs
@OpenRouter
@useTRMNL
@zeddotdev
Announcement coming Monday -
R 'Nearest' Nabors
@rachelnabors
I once had a team go to bat for me to join. The HM said I didn't have enough GitHub squares.
It happens to the best of us 😅 -
R 'Nearest' Nabors
@rachelnabors
Just build.
-
R 'Nearest' Nabors
@rachelnabors
This is one of the reasons I placed an order for a maxed out MBP. I expect it will be hard to get that spec again at this price, and I will, through DS4, be able to code with AI amid inference pricing escalation
Mario Zechner
@badlogicgames
recommended viewing. with interesting numbers.
youtu.be/zyQwAhppWj8 -
R 'Nearest' Nabors
@rachelnabors
TIL that my fave meetup Claude Code Anonymous shut down 🥺 Thank goodness we still have @ReactAdvanced (now Web Engineering London)! So good to see @orta and Michal!
-
R 'Nearest' Nabors reposted
Peter Gostev
@petergostev
Can't make it, but you should!
Andy T
@Andy_AJT
Announcing the next Codex Community Meetup - London. Monday 8th June.
Last Codex London Meetup had 1000+ registrations & almost 200 of the most cracked agentic engineers from the hottest startups and biggest companies in London chatting about agentic engineering.
Link below 👇 -
R 'Nearest' Nabors
@rachelnabors
Lies
-
R 'Nearest' Nabors
@rachelnabors
My AI powered worklife
R 'Nearest' Nabors
@rachelnabors
I used to be ashamed of how scratchy my hand-drawn comics looked. Now I'm very happy that they look so rough because it is a testament to my ability to drag a pen across paper without computer assistance
-
R 'Nearest' Nabors
@rachelnabors
Honored to be giving the opening keynote at PyData London 2026 next Saturday, June 6 at 9:10am.
"The Community Is the Boat," or what saved me when I went from peak career to the bottom rung TWICE. It's about which skills travel, what we owe each other, and finding your boat in a rising tide. Will I see you there?
pydata.org/london2026
#pydatalondon26 #pydata #numfocus -
R 'Nearest' Nabors
@rachelnabors
Spending Sunday recruiting other weirdos and teens to work a skunkworks "browser of the future" because I am tired of waiting
-
R 'Nearest' Nabors
@rachelnabors
Backrooms was perhaps the best scary movie I've seen in a bit, in the vein of Us. DM if you want to nerd out
-
R 'Nearest' Nabors
@rachelnabors
Putting a record scratch in my own Pydata keynote. See you next Saturday morning, bright and early, London!
-
R 'Nearest' Nabors
@rachelnabors
I have been watching Erica build this over tea for many weeks now. I love seeing what young founders are making with AI. Give it a whirl for me—I am on Android and jealous!
Erica Hu
@epricat
Glia is in public beta on the App Store 🐾🩷 download today and start gardening memories that matter the most to you.
-
R 'Nearest' Nabors
@rachelnabors
An architectural overview of how Hermes is "built different"
Aparna Dhinakaran
@aparnadhinak
-
R 'Nearest' Nabors reposted
Georgi Gerganov
@ggerganov
llama.cpp now has an official website: llama.app
Our goal is to make local AI accessible to everyone, and improving the user experience is a big part of that. On the new landing page you’ll find a single-line cross-platform installer. The installation provides a single unified `llama` entrypoint which you can use to run/serve models and interface with 3rd-party agentic applications.
While oriented towards simplified user experience, the new `llama` application also provides all the advanced functionality of the existing llama.cpp tooling with which experienced users are already familiar. Also note that all GGUF models that you might have already downloaded with llama.cpp in the past will be automatically available to use without downloading again (they are stored in the common HF cache on your machine).
We have many improvements in the pipeline both at the UX and at the engine level and we plan to iteratively ship new things over the coming months. One of the main focuses will be seamless integration with local-friendly 3rd-party agents (such as Pi). In the meantime, we’ll continue to listen for feedback from the community and adjust accordingly, so keep letting us know what you think and need. -
R 'Nearest' Nabors
@rachelnabors
The best part of living in London is being able to hop on a plane or train and hit up every ENEA tech center. Berlin, Paris, Amsterdam—these are the communities I fell in love with ten years ago, and they are still going strong. You will always be strongest together
R 'Nearest' Nabors
@rachelnabors
At least you know I am writing my own posts with all my typos lol
*EMEA -
R 'Nearest' Nabors
@rachelnabors
Heading back to London and Paris respectively—always happy to talk opensource community development with @alpic_ai's @Jeanroudy 😈 They host the best MCP workshops. Last night was PACKED
-
R 'Nearest' Nabors
@rachelnabors
Thank you Berlin and @techeurope_ for having me! I loved talking with a tone who would listen about small models. You were a delight!
-
R 'Nearest' Nabors
@rachelnabors
RT @ReactAdvanced: Web Engineering London Meetup is back on June 3rd 🚀
Join us for an evening of engineering talks, AI, frontend architect… -
R 'Nearest' Nabors
@rachelnabors
Berlinmaxxing at Applied AI Conf today! Catch my talk on switching from Sonnet to Llama for a consumer app at noon!
R 'Nearest' Nabors
@rachelnabors
Wow, that's the best response I've had to an AI talk ever! My room was packed, and ~10 engineers came up to me afterward to ask questions! Great job @techeurope_ team!
R 'Nearest' Nabors
@rachelnabors
My voice started to go so I ended up living off these convenience popsicles
-
R 'Nearest' Nabors
@rachelnabors
Eat. Sleep. Hack. Repeat.
-
R 'Nearest' Nabors
@rachelnabors
I'm so bad at trains some days. Berlin, I promise I will find my way to you
R 'Nearest' Nabors
@rachelnabors
Stuck here for 40 min. At least the weather is schön ☀️
-
R 'Nearest' Nabors reposted
Mario Zechner
@badlogicgames
recommemdations will continue until the timeline improves
-
R 'Nearest' Nabors
@rachelnabors
My talk from @aiDotEngineer Europe is up! Even the Headless Web is full of infinite possibilities for interaction and creativity. I created an MCP server for my corpus of cartooning work that streams a comic-reader MCP App. Plus the site has WebMCP support to boot!
-
R 'Nearest' Nabors reposted
arize-phoenix
@ArizePhoenix
"Don't trust. Evaluate."
@nearestnabors set out to replace Claude Sonnet with Gemma 4. The evals showed a quantifiably better option.
Full walkthrough: capability evals + prompt engineering to ship a local 3B that matches Sonnet, 2x faster, $0/call.
Built with Phoenix. -
R 'Nearest' Nabors
@rachelnabors
ICYMI over the long weekend, I've been having fun with squeezing frontier performance out of local models (and shaving seconds off of latency for my troubles)
R 'Nearest' Nabors
@rachelnabors
-
R 'Nearest' Nabors
@rachelnabors
Londonmaxxing on a bank holiday like you wouldn't believe
-
R 'Nearest' Nabors
@rachelnabors
Having actually read Empire of AI (and you should, too), the "bottle of water" quote didn't come from that book
-
R 'Nearest' Nabors reposted
Diksha
@dikshadutta
As a woman in tech, I've been mansplained:
→ How hackathons "actually work" — I've organized 20+ and onboarded thousands of builders
→ How to do marketing better — 10+ years growing founder and company accounts, millions of impressions
→ How to host better events — my curated ones run with near-zero no-shows
→ Men skipping past me to talk directly to my founders, ignoring my questions. I'm C-level.
The funniest part: half of them are pitching me for intros.
The founders I chose to work with treat me as an equal. So this noise doesn't land.
2026, by the way. -
R 'Nearest' Nabors
@rachelnabors
Time to build!
What're you hacking on? -
R 'Nearest' Nabors
@rachelnabors
Getting to see @bdougieYO TWICE in one day was too awesome.
I wish we could get him to do an AMA for local founders before he leaves (cc @ZakeeAbdi @charlierward) -
R 'Nearest' Nabors
@rachelnabors
SAGE (Small and Good Enough) Models save you buckets of money on aI image gen
Christina, PhD
@christinaharker
Popping up to say this applies to image and video gen models too, not just LLMs. The math will kill you money-wise at scale. It's absolutely brutal.
If people can choose a SAGE (small and good enough) model for some tasks, they literally save tens or hundreds of thousands of dollars.
1k of Nano Banana 2 images costs you about $67. 1k of FLUX [klein] 4B costs $3.
1 million platform users needing an avatar thumbnail? Thats $67,000 vs $3,000. -
R 'Nearest' Nabors
@rachelnabors
ICYMI, I replaced Sonnet in my app with Lllama 3.2 and am saving average $1.50 on inference per user. How I did it:
R 'Nearest' Nabors
@rachelnabors
-
R 'Nearest' Nabors
@rachelnabors
BOOM.
Used evals and prompt engineering to get a little local model in an app to lift as well as Claude Sonnet.
Each test run cost $0.44 in API calls. That's like, 84 summarization tasks. No way is that sustainable for normal apps.
Every damn task, I try to make it run on a local model.
Now I can finally measure and tweak to perfection.
Anyway, come see the talk at noon at ML Conf in London! mlconference.ai/london/
R 'Nearest' Nabors
@rachelnabors
For some of you, this is common knowledge. But for so many would-be founders I sit next to at hackathons, being eaten alive by inference cost, this is something that needs more airtime
R 'Nearest' Nabors
@rachelnabors
Look at that latency! Look at that money you're not paying to send user data to someone else's servers! Moving compute back to the edge has its perks
R 'Nearest' Nabors
@rachelnabors
How I did it:
R 'Nearest' Nabors
@rachelnabors
-
R 'Nearest' Nabors
@rachelnabors
I hope you like SLMs
R 'Nearest' Nabors
@rachelnabors
For you:
R 'Nearest' Nabors
@rachelnabors
-
R 'Nearest' Nabors reposted
Laurie Voss
@seldo
Do you need a frontier model, really?
R 'Nearest' Nabors
@rachelnabors
-
R 'Nearest' Nabors
@rachelnabors
-
R 'Nearest' Nabors reposted
Rizèl Scarlett 🇦🇬🇬🇾
@blackgirlbytes
-
R 'Nearest' Nabors
@rachelnabors
TIL there's a secret "DMs from people you don't know" tab in X and I'm just now seeing all these heartfelt requests to connect from AIE *sob*
-
R 'Nearest' Nabors
@rachelnabors
I'm wrapping this post on how to make an SLM rock as hard as an LLM using eval experiments and prompt engineering and listening to the new Gorillas album.
Whatcha working on today? -
R 'Nearest' Nabors
@rachelnabors
Anyone up for a sneak peek at an upcoming article of mine on how to squeeze LLM performance out of a 2GB model that fits on your device? Please slide into my DMs
-
R 'Nearest' Nabors
@rachelnabors
I'm not sorry
-
R 'Nearest' Nabors
@rachelnabors
Playing chess and learning Japanese paid off!
-
R 'Nearest' Nabors
@rachelnabors
For an article I'm writing
-
R 'Nearest' Nabors
@rachelnabors
"Am I in the right place? Should I move? Is there time still left? Am I gonna make it?"
I've asked these questions of myself many times. There was a time, at the start of my career, when I was invited to join Facebook as a UX designer. But my then life partner didn't want to move to the Bay. I sat that opportunity out, and I live with those regrets. Even if I hadn't stayed longer than vesting, the value of those stocks would have provided the long-term security my marriage did not.
Money buys a lot.
But at the same time, I've known many folks who moved to the Bay and never had a "big break." Some even drown in debt because of the lifestyle costs.
Being in an area increases the number of shots you get. But you don't win every shot.
If life brings you an opportunity, step into it. Take the shot. But that also means taking the shot in front of you.
When the layoffs hit Amazon, I took a shot on a return to the UK. ILR would mean Free Healthcare for Life and the privilege to work and remain in two countries, unhasseled by law.
That's a really big life hack. You don't need a lot of money to retire well with infinite health potions on command.
I could have gone to the Bay and scrambled to get into Anthropic or OpenAI. Maybe I would have. Maybe I wouldn't.
But I decided on living a life more aligned with what I like about being alive.
Should you have moved? Maybe. Did you? No, or you wouldn't be wondering. Is there an offer for you to move? Then you should step into that opportunity. Is there not? Then make your peace where you stand and carry on.
We can't win all games. Can't win all bets.
When opportunity knocks, open the damn door. Otherwise, let it go and enjoy the present
Deedy
@deedydas
The vibes in SF feel pretty frenetic right now. The divide in outcomes is the worst I've ever seen.
Over the last 5yrs, a group of ~10k people - employees at Anthropic, OpenAI, xAI, Nvidia, Meta TBD, founders - have hit retirement wealth of well above $20M (back of the envelope AI estimation).
Everyone outside that group feels like they can work their well-paying (but <$500k) job for their whole life and never get there.
Worse yet, layoffs are in full swing. Many software engineers feel like their life's skill is no longer useful. The day to day role of most jobs has changed overnight with AI.
As a result,
1. The corporate ladder looks like the wrong building to climb.
Everyone's trying to align with a new set of career "paths": should I be a founder? Is it too late to join Anthropic / OpenAI? should I get into AI? what company stock will 10x next? People are demanding higher salaries and switching jobs more and more.
2. There’s a deep malaise about work (and its future).
Why even work at all for “peanuts”? Will my job even exist in a few years? Many feel helpless. You hear the “permanent underclass” conversation a lot, esp from young people. It's hard to focus on doing good work when you think "man, if I joined Anthropic 2yrs ago, I could retire"
3. The mid to late middle managers feel paralyzed.
Many have families and don't feel like they have the energy or network to just "start a company". They don't particularly have any AI skills. They see the writing on the wall: middle management is being hollowed out in many companies.
4. The rich aren’t particularly happy either.
No one is shedding tears for them (and rightfully so). But those who have "made it" experience a profound lack of purpose too. Some have gone from <$150k to >$50M in a few years with no ramp. It flips your life plans upside down. For some, comparison is the thief of joy. For some, they escape to NYC to "live life". For others still, they start companies "just cuz", often to win status points. They never imagined that by age 30, they'd be set. I once asked a post-economic founder friend why they didn't just sell the co and they said "and do what? right now, everyone wants to talk to me. if i sell, I will only have money."
I understand that many reading this scoff at the champagne problems of the valley. Society is warped in this tech bubble. What is often well-off anywhere else in the world is bang average here.
Unlike many other places, tenure, intelligence and hard work can be loosely correlated with outcomes in the Bay. Living through a societally transformative gold rush in that environment can be paralyzing. "Am I in the right place? Should I move? Is there time still left? Am I gonna make it?" It psychologically torments many who have moved here in search of "success".
Ironically, a frequent side effect of this torment is to spin up the very products making everyone rich in hopes that you too can vibecode your path to economic enlightenment.
R 'Nearest' Nabors
@rachelnabors
Also I eventually did work at meta over a decade later. On the @reactjs team. Life is funny that way
-
R 'Nearest' Nabors
@rachelnabors
Hey from Mima!
-
R 'Nearest' Nabors
@rachelnabors
This weekend's Tea Hack is brought to you by: writing! Specifically about localfirst inference...
What are you hacking on? -
R 'Nearest' Nabors reposted
-
R 'Nearest' Nabors
@rachelnabors
Last round, I swear
-
R 'Nearest' Nabors
@rachelnabors
"A salary is a subscription someone else can cancel."
This is why I'm bullish on the rise of the agency and the consultant. It's better for both parties.
R 'Nearest' Nabors
@rachelnabors
This account was suspended? Why??
-
R 'Nearest' Nabors
@rachelnabors
Patiently waiting in London for my 128GB M5 Max like
-
R 'Nearest' Nabors reposted
Sudo su
@sudoingX
buy a gpu. 3090, 4090, dgx spark, whatever fits your budget. tier doesn't matter. running your first local model does. the moment your first prompt lands with no api between you and the model, your brain rewires. that single moment is worth more than every take you'll ever read on a timeline.
-
R 'Nearest' Nabors
@rachelnabors
Meanwhile, mima now has thread summaries + annotations. Don't be caught out catching up on a dogpile thread ever again!
-
R 'Nearest' Nabors
@rachelnabors
This is so confusing, it must be inspired by American health insurance.
Glad I'll be running my own coding agent on device soon. Soon.
ClaudeDevs
@ClaudeDevs
How it works: Claim the monthly credit once, and programmatic usage will draw from it automatically. When it runs out, you can keep going with usage credits (billed at API rates you turn on/off). If usage credits are turned off, usage pauses until the credit resets.
-
R 'Nearest' Nabors
@rachelnabors
This is how I feel when people demo their custom orchestration layers
james hawkins
@james406
frog told the LLM "do not hallucinate"
"there," he said, "now the LLM will not make mistakes"
"but the LLM can still hallucinate" said toad
"that is true" said frog -
R 'Nearest' Nabors
@rachelnabors
Wired: "Waiting for evals to run"
Tired: "Waiting for Ralph Wiggum loops"
Expired: "Waiting for code to compile"
(At least that's what I tell myself as I sneak in chess lessons on Duolingo!) -
R 'Nearest' Nabors
@rachelnabors
Switching over from Ollama to llama.cpp after reading this: sleepingrobots.com/dreams/stop-using-ollama/
Today I'm setting up a Grand Competition (Phoenix evals) between itty bitty models to see who is going to be the Top (local) Model for Mima (with Prompt Engineering, of course) -
R 'Nearest' Nabors
@rachelnabors
It's been so long since I did IC work. I forgot how heads down it can be. Meetings? Now I have to take a half hour to get back to where I was.
But also, when I'm not responsible for researching a system and putting process and programs in place, it's more rewarding in some ways. -
R 'Nearest' Nabors
@rachelnabors
Honored to be keynoting PyData London 2026 June 5-7.
(Spoilers: I'll be talking about evolutionary engineering and why you shouldn't lay off your QA team if you want to keep up.)
Will I see you there? Get 10% off your ticket with this link: ti.to/pydata/pydatalondon26/discount/Rachel10 -
R 'Nearest' Nabors
@rachelnabors
Top of mind for my MLConf talk on Wednesday. mlconference.ai/london/
-
R 'Nearest' Nabors
@rachelnabors
Goodbye, spring
R 'Nearest' Nabors
@rachelnabors
You need to play goodbye horses in the background for full effect
-
R 'Nearest' Nabors reposted
Nick Taylor
@nickytonline
Catch the full convo with @rachelnabors 👇
www.youtube.com/watch?v=D7KfnGdHayA -
R 'Nearest' Nabors reposted
Jeremy Yudkin
@yudDIDit
Today, @editframe emerges from stealth. Agents need video.
Editframe Agent Skills:
npm create @editframe@latest
Just prompt Claude Code, Cursor, or Codex and get a working video or a full interactive GUI.
This video was created just by prompting 👇 -
R 'Nearest' Nabors
@rachelnabors
Ok, I'm in
Mario Zechner
@badlogicgames
recommended reading by our junion developer @mitsuhiko
my name is pidalf, and i support this message. would love to have some spare cycles to join @antirez effort and use some of my GPU knowledge.
alas.
lucumr.pocoo.org/2026/5/8/local-models/
R 'Nearest' Nabors
@rachelnabors
Love fabric cases. They are great for iron patches > stickers
-
R 'Nearest' Nabors
@rachelnabors
Saturday hacking + tea with fellow founder @catinwool. Discussed: "dark founding principles," the ones you don't put in the deck, and how to make them sound more socially acceptable. Like "reimagining social networks" (vs "burning everything down and dancing on the ashes"). V important startup stuff
-
R 'Nearest' Nabors
@rachelnabors
Hoohoo, but you can set up a relay network to post to every account you have from a different client...
I'm not here to abscond with anyone's network. I'm just here to make a better client to keep people engaging on them... 🤭
Michael Mignano
@mignano
-
R 'Nearest' Nabors
@rachelnabors
This is where indie coding is going.
Best idea I've read all week.
Now to see if I can rustle up the RAM
Mario Zechner
@badlogicgames
recommended reading by our junion developer @mitsuhiko
my name is pidalf, and i support this message. would love to have some spare cycles to join @antirez effort and use some of my GPU knowledge.
alas.
lucumr.pocoo.org/2026/5/8/local-models/ -
R 'Nearest' Nabors
@rachelnabors
Ok, I've explained it three times, so it's a thing I should post about.
Evolutionary Engineering.
It's the acceptance that all code will be mid (unless created by godlike, all knowing staff engineers with an eternity to plan the perfect codebase and another eternity to execute that plan). We are mortals with flawed tools and small minds. Show me any codebase made by humans, and there is the code equivalent of duct tap holding pieces of it together, no matter how enlightened the crew that built it.
It is the belief that variation (non-deterministic expression) is expected and welcome. Mutations that aren't deadly or benign move the system in the direction of better fitness for purpose.
It is the skills to build not code, but the environment in which code blossoms from other mechanisms.
In this environment, all code starts as mid and is revised and mutated and shifted.
The environment is deadly to mutations that do not serve the engineer's purpose. In this way, the program evolves in the direction of fitness.
Will the end result be perfect? No. And the mammalian heart is flawed in a way that it is prone to failure, but only after the age of reproduction.
Code rarely lives longer than a few decades. It doesn't need to be perfect. It just needs to do its job. And code that doesn't, needs to die in a fire.
Folks I've met practicing this kind of engineering seem at peace with how engineering is changing, have a background in testing, and generally seem to enjoy creations built by agents that work -
R 'Nearest' Nabors
@rachelnabors
Your friendly, weekly PSA that you don't have to change the world, you just have to survive it. You change it just by participating in it. When times are rough, remember that that's enough
-
R 'Nearest' Nabors
@rachelnabors
Models, when you start really trying to understand them, are each bizarrely unique. Not magic. Just math. But math you need to empathize with to get the most out of them
Shaun Smith
@evalstate
Models don't all see Tools the same way.
Sharing some research on how JSON Tool Definitions end up getting "seen" by 14 different models.
Each model had it's chat template downloaded and increasingly complex tools schemas rendered and recorded to see what remained intact. -
R 'Nearest' Nabors
@rachelnabors
ICYMI earlier this week, a little reading for your weekend....
Arize AI
@arizeai
We (well @rachelnabors) spent time evaluating different agent harness finish conditions across GPT-4o and Claude.
The evals surfaced a sneaky failure mode: agents exiting after describing the next action instead of taking it.
But in the transcript, those runs still looked plausible. -
R 'Nearest' Nabors
@rachelnabors
When anyone posts a picture of a group of men with no women in it, I reply with "where are the women at?"
Unless it's an all-boys event or school, it's usually a good question to ask.
Do you know where your "missing sisters" are? -
R 'Nearest' Nabors reposted
Jack Herrington
@jherr
Folks are starting to coalesce around compilers being a historical analogue for LLMs. They are not. Compilers are deterministic. LLMs are not.
"Back in the day we used to hand code write assembler better than the compiler" is factually true. But two factors changed that; complexity and scale. Processors got faster by using pipelines, which necessitated writing assembler in a non-intuitive way and that was something gcc could handle.
Applications also got larger, and the net result was that we needed higher leverage portable languages that could be compiled, deterministically, into machine code. Any non-determinism in a compiler is... a bug. Non-determinism in LLMs is a feature.
Asking an LLM to write code for a non-trivial problem is going to result in different code each time. And if the answer to that is to come up with a better way of representing the problem to the LLM so that it can be deterministic, congratulations, you've just developed a new language. And you might as well have an LLM build you a deterministic compiler for that because a compiler will be faster (and cheaper) than using an LLM to develop on your new language.
The non-deterministic element of work has always been humans. Once we decide a part of work is routine and valuable enough to automate then we create machines to deterministically reproduce that work.
I keep looking around for a historical analogue to this. And unfortunately the only analogues revolve around reducing the cost of human labor. Regardless, compilers are not a good analogue for LLMs, LLMs are token predictors. By design they are non-deterministic.
(BTW, I'm not anti-LLM, just conveying my understanding of the sitatuation.) -
R 'Nearest' Nabors
@rachelnabors
Making my own social network hub as per usual.
Just wait till I add relationship management. -
R 'Nearest' Nabors
@rachelnabors
Do me a favor? Reply with the coolest thing you saw this weekend
-
R 'Nearest' Nabors
@rachelnabors
Which would you log into to keep up with your socials and news?
-
R 'Nearest' Nabors
@rachelnabors
Every once in awhile I look at Claude Cowork, and every time I wonder who the heck can use this thing? It feels like an app from 10 years ago.
-
R 'Nearest' Nabors
@rachelnabors
Londonmaxxing is hitting up a zine fair right after a Hackathon on one of the UK's copious three day weekends. So cool to meet @ZakeeAbdi! Thanks for setting up a great event, @oscarfalll!
-
R 'Nearest' Nabors
@rachelnabors
Mario is one of my favourite voices in Tech right now
Mario Zechner
@badlogicgames
hi, i'm a sole proprietor/founder in Austria and i earn many many multiples of what i'd earn as an employee, despite "predatory income tax". in fact, i opt out of the many tax optimizations i could use because i like having good schools and as high a standard of living as possible for everyone.
the great thing about the EU is that you can just live under any tax regime you like in any of the 27 member states. it's all about trade offs. if poland works for you, fantastic! go build there.
and if i may add one more thing: if the CEO of a startup, especially pre-revenue, lives "barely any better than a regular employee" then the system works as intended. fact of the matter is most startups are bad. you are not special because you are trying out a shit idea and fail. but i'll happily pay taxes so you can try your shit idea, fail, and can still live. -
R 'Nearest' Nabors reposted
Paul Bakaus
@pbakaus
this is 2026 AI slop.
ironically, in part introduced by anthrophic's frontend-design skill, which replaced one set of model defaults...with another (fraunces for editorial anyone? warm brown italics?)
it's incredibly hard to avoid, and build software and skills that unlock true creativity and uniqueness (very much a requirement to stand out with your landing page). i've started experimenting with 'anti-attractors' for fonts/colors that attempt to branch out from the defaults (now part of impeccable), but only getting started.
take a look, play around and familiarize yourself with all types of slop in the slop gallery: impeccable.style/slop/ -
R 'Nearest' Nabors
@rachelnabors
Londonmaxxing with @thondankani on a lovely long weekend Sunday
-
R 'Nearest' Nabors
@rachelnabors
Pick one
-
R 'Nearest' Nabors
@rachelnabors
Joining the rest of the Londonmaxxing crew at a Hackathon. Time to ship some features!
Post not available here, view on X
R 'Nearest' Nabors
@rachelnabors
The only reason to incorporate in Delaware is if you're hiring in the USA or raising from US VCs (welcome to the "Delaware flip" where the shares go from UK to US company)
R 'Nearest' Nabors
@rachelnabors
But apparently it's easier to sell and list a UK company, huh.
Not gonna have to think about that for ten years -
R 'Nearest' Nabors
@rachelnabors
Ooh, Xiaomi's MiMo models are now open-weight. "general agentic capabilities, complex software engineering, and long-horizon tasks, now matching frontier models across most benchmarks." Get it while it's hot: huggingface.co/collections/XiaomiMiMo/mimo-v25
-
R 'Nearest' Nabors reposted
claire vo 🖤
@clairevo
-
R 'Nearest' Nabors reposted
R 'Nearest' Nabors
@rachelnabors
-
R 'Nearest' Nabors reposted
DebamitroChakraborti
@debamitro
@LeaVerou is the web/CSS expert
@sudobunni is the friendly neighborhood CLI girl
@ruima is covering the China tech universe
@drfeifei is a legend
@_LuoFuli is shaking up the LLM universe
@IrenaBuzarewicz is holding the fort on non-AI art -
R 'Nearest' Nabors reposted
DebamitroChakraborti
@debamitro
not sure if I follow all of them, but here are some interesting profiles from my feed:
@eliana_jordan builds apps while scuba diving
@lauriewired has taken explainer videos to a different level
@clairevo is active in the OpenClaw/ClaudeCode circuit -
R 'Nearest' Nabors
@rachelnabors
And only one woman replied.
R 'Nearest' Nabors
@rachelnabors
I miss when there were women on this platform.
-
R 'Nearest' Nabors
@rachelnabors
Ten years ago, I knew women in design, in engineering. If I wanted to start a project or hire a person, I had a few women in mind.
As I've gotten older, they've disappeared. It honestly pains me when I see an all-male anything. Because I can still remember a woman who should have been there by now, but has gone "missing."
I spent the better part of my career advocating for women in tech, yet.. It feels like everyone was so eager to forget we ever tried to spotlight them or encourage more of them.
It's just guys hiring and inviting their friends and younger men who remind them of themselves.
It's fucking lonely out here.
Damn, I'm tired.
R 'Nearest' Nabors
@rachelnabors
And I know, we have new women coming in.
But so few of them stick around. When you get to principal and director levels, it's just dudes. It wouldn't be so bad if they would include other people in their high fives and laughter, but these things are reserved for other dudes -
R 'Nearest' Nabors reposted
Sarah Drasner
@sarah_edo
This paper about throttling web agents using Reasoning Gates is pretty interesting!
For stealth agents that ignore no-crawl directives, they give them multi hop tests of intelligence-
It gives the agent very complex logic problems to solve that it would take humans hours of research to do. For an agent, it takes a matter of seconds. The token usage is low for answering once, but too high for multiple attempts.
So it introduces computational asymmetry. Getting to an airline site- negligible. But deploying many many agents to perform malicious scraping/DDoS? The server bills alone would bankrupt malicious actors before they got anywhere.
arxiv.org/html/2509.01619v1 -
R 'Nearest' Nabors
@rachelnabors
I miss when there were women on this platform.
-
R 'Nearest' Nabors reposted
Swizec Teller
@Swizec
Funfact:
When candidates put AI on their resume, the key thing I try to find out is whether they used evals. How did you measure making improvements?
this filters out 80% of engineers
because @eugeneyan was right: You can't build much more than a toy without evals -
R 'Nearest' Nabors reposted
Angie Jones
@techgirl1908
🌶️ Hot take: just because agents can easily generate tests doesn't mean they're any good.
In fact, I still spend quite a bit of human effort ensuring that my tests are trustworthy, as these are my guardrails for current and future quality.
I'm curious... how much do you trust AI generated tests?
Can't wait to discuss this with Anand Sundaram, CEO of @Applitools
Join us! applitools.info/b7t -
R 'Nearest' Nabors
@rachelnabors
Londmaxxing by supporting @evalstate at Tessl HQ
-
R 'Nearest' Nabors
@rachelnabors
TIL I have an uwu voice and a Big Dog Voice
-
R 'Nearest' Nabors
@rachelnabors
Told you I'd be working on harnesses (and my watercolours)
R 'Nearest' Nabors
@rachelnabors
-
R 'Nearest' Nabors
@rachelnabors
Friends don't let friends let AI generate their thumbnails
Erica Hu
@epricat
I wanna live in this version of the future where we illustrate our ideas by hand
-
R 'Nearest' Nabors
@rachelnabors
-
R 'Nearest' Nabors
@rachelnabors
How are founders not constantly posting love letters to QA? Everything is safer, more accessible, faster, bugs disappear.
I truly believe QA is more necessary than ever now that the bottleneck has shifted to reviewing code -
R 'Nearest' Nabors
@rachelnabors
Made it to @bilgeycl'sunconference! Thanks @Haystack_AI @nvidia @balderton for putting on the first unconference of my year!!
-
R 'Nearest' Nabors
@rachelnabors
I meet a lot of folks building their own harnesses and orchestration layers.
Most of the time, they are not producing anything from these systems that they can show. Their amazing coding harness is a side project—they still use Claude Code at work. They created a hive mind for agent swarms—not that they have actually used swarms in production.
I love that we're experimenting, but sometimes I worry everyone has AI psychosis, burning tokens because the movement looks like motion -
R 'Nearest' Nabors
@rachelnabors
Can't decide if this is auspicious or not
-
R 'Nearest' Nabors reposted
Bilge
@bilgeycl
🇬🇧 After some delays, I am on my way to London for the @Haystack_AI & @nvidia Unconference this evening!
Who’s coming? luma.com/3iyatqzz -
R 'Nearest' Nabors
@rachelnabors
Just played my first full song on the drums! Seven Nation Army, natch 😂
-
R 'Nearest' Nabors
@rachelnabors
Apparently, VC's like vengeance-driven development
-
R 'Nearest' Nabors reposted
leonie
@helloiamleonie
This is the most common question I’ve been asked recently:
“Should I replace my RAG system with an agentic RAG one?”
Well, are you happy with your current RAG system’s performance?
Then, probably not.
The key difference between traditional RAG and agentic RAG is HOW the context is built:
Traditional RAG has a fixed retrieval pipeline that retrieves exactly once.
→ The LLM is a passive recipient of additional context
Agentic RAG has access to one or more retrieval tool(s) to retrieve on demand.
→ The agent actively builds its own context
> Does your use case always require exactly one retrieval step? (Example: always pull in customer information)
> Or do you have cases where no additional context is needed?
> Or do you have cases where multi-hop retrieval is necessary?
If you don’t know, set up a simple evaluation with a gold test set of a few common user patterns:
> Do precision and recall improve for your retrieval component?
> How much does the end-to-end performance improve based on performance improvements of the retrieval component?
> Do these performance improvements justify the added latency and cost of agentic RAG? -
R 'Nearest' Nabors
@rachelnabors
Ugh. I spent all this time making mini agent harnesses to prove an architectural point with evals. However, it turns out the architectural detail in question was built to solve a problem the models in question (4o) no longer have (saying you're going to call a tool in one turn, then calling the tool on the following turn).
Now I have all these mini harnesses hooked up to evals. What am I gonna do with these little guys?
Even Gemma looked at me and was like, "I see what you're trying to get me to do. O ye of little faith." Started completing great big chunks of tasks at a go—such an over achiever!
github.com/nearestnabors/harness-evals -
R 'Nearest' Nabors
@rachelnabors
What is the end game of this
vintagelover
@queennaurelia
I’ve never seen anything more accurate
-
R 'Nearest' Nabors reposted
R 'Nearest' Nabors
@rachelnabors
Now that code is free, we're running up tech debt like a teenager with their first credit card.
I wrote up what I heard across three days at @aiDotEngineer Europe, from the keynotes to the booth to brunch the morning after.
Huge thanks to the people who came all this way to share their thoughts (I hope I've done you justice): @badlogicgames @mattpocockuk @mitsuhiko @milksandmatcha
arize.com/blog/code-is-free-technical-debt-isnt-notes-fro… -
R 'Nearest' Nabors
@rachelnabors
I know plenty of founders who dream of getting rich. But are there any who dream of obliteration? To disrupt a sector completely, regardless of whether doing so turns a handsome profit? The founder equivalent of spite driven development?
R 'Nearest' Nabors
@rachelnabors
I've always wondered if Bitcoin was one of these plants.
The only thing that can disrupt a system is a greater system. So if you really didn't like the establishment, you could plant the seed of something that would go viral on its own, corrode its roots. People would think it could make them rich, but it was never meant to make anyone rich, just destabilise a large system -
R 'Nearest' Nabors
@rachelnabors
If you're wrangling agents for repetitive work, check Barnum out. It uses a pattern of "hostile environment" to curb the mutational aspects of agentic workflows while still allowing for their fecundity
Robert Balicki (👀 @IsographLabs)
@StatisticsFTW
What is Barnum? Barnum is a programming language that aims to make it easy to orchestrate agents. In particular, invoking an LLM is but an example of long-lived, asynchronous work, so the point of Barnum is to give you a high level DSL for managing asynchronous work.
Consider the pictured example. Here, we execute three steps in series, looping them until they finish *or* until stepA catastrophically fails.
Barnum isn't focused on what happens inside of those steps! You can write files, invoke an LLM, log to the console, whatever. Barnum is focused on everything that surrounds those steps: the control flow.
Why? Read on... -
R 'Nearest' Nabors
@rachelnabors
Heading to the mini Google I/O in Berlin this year (June 25) to catch up with some of my fave browser peeps. Will I see you there?
-
R 'Nearest' Nabors
@rachelnabors
Goddammit, it followed me home
etn.
@etnshow
BREAKING: Jeff Bezos's AI lab 'Project Prometheus' eyes London, Kings Cross for new office.
Project Prometheus is discussing signing a lease at the Jellicoe Building for three floors spanning 38,000 sq ft. -
R 'Nearest' Nabors
@rachelnabors
Hunting an illustrator in the wild for brand work at artists' fair. #londonmaxxing
-
R 'Nearest' Nabors
@rachelnabors
-
R 'Nearest' Nabors reposted
Eddie Jiao
@eddiejiao_obj
playing with Flipbook on a color e-ink tablet
Zain Shah
@zan2434
Imagine every pixel on your screen, streamed live directly from a model. No HTML, no layout engine, no code. Just exactly what you want to see.
@eddiejiao_obj, @drewocarr and I built a prototype to see how this could actually work, and set out to make it real. We're calling it Flipbook. (1/5) -
R 'Nearest' Nabors reposted
Arcade.dev | The Actions Runtime
@TryArcade
Yesterday, the Arcade team rang the @Nasdaq Closing Bell.
Every enterprise is figuring out agents right now. We've been building for this moment since day one.
Thank you @Wing_VC and @EricNewcomer. -
R 'Nearest' Nabors
@rachelnabors
A UI for everything is harder than it looks
-
R 'Nearest' Nabors
@rachelnabors
All you need is a damn good QA eng
Vic 📠
@VicVijayakumar
The code is already complete from the agent, before you start your work. It is already there, you just have to chisel away the slop.
-
R 'Nearest' Nabors
@rachelnabors
She's cooooming...
Paying real people to design this brand precisely because it's "not like the other (vibe coded) girls."
How's your weekend? -
R 'Nearest' Nabors reposted
-
R 'Nearest' Nabors
@rachelnabors
Get you a good QA engineer who likes building systems with agents. Can't recommend enough
David K 🎹
@DavidKPiano
USERS care if code is messy.
They care if the app is buggy. Or slow. Or if the UX is frustrating. Or if important features are missing or broken.
These are downstream effects of messy, unmaintainable code. -
R 'Nearest' Nabors
@rachelnabors
Why is no one talking about A2UI? Is this because it's a python stack shipping UI in a TS world? It is, isn't it
-
R 'Nearest' Nabors
@rachelnabors
Back when I was dating, I realized the meaning of "people treat you how you let them."
You don't get to adopt this attitude so easily in leadership roles. VP only talks over drinks? Pack the electrolytes. Report struggling to attend meetings on time? Start coaching them.
Leaders have to have flexibility in boundaries that isn't required of an IC. Meetings keep being scheduled after 8pm? Unless they are mandatory, decline. If you accept them, people are going to keep scheduling you out of dinner. If you don't, they'll have to adapt and find a time that works for the whole team.
Ive noticed my IC self is a different beast from my manager self -
R 'Nearest' Nabors
@rachelnabors
In a sea of AI generated content that both humans and bots can identify as generated, I'm noticing "slow thoughts" seem to carry more weight: the one post from a person you really respect in a deluge of posts from companies and creators trying to rank. Am I imagining it?
-
R 'Nearest' Nabors
@rachelnabors
ICYMI
Arize AI
@arizeai
Now that code generation is cheap, technical debt compounds even faster. A few thoughts from @aiDotEngineer Europe before @AIEMiami by @RachelNabors: arize.com/blog/code-is-free-technical-debt-isnt-notes-fro…
-
R 'Nearest' Nabors
@rachelnabors
Found my calling in life: making slightly off-putting AI slop inspirational posters
-
R 'Nearest' Nabors
@rachelnabors
This is not a drill. We have a full Tsujiri in London. I repeat, this is not a drill. Please form an orderly queue to receive your sakura matcha sundaes
-
R 'Nearest' Nabors
@rachelnabors
Now that code is free, we're running up tech debt like a teenager with their first credit card.
I wrote up what I heard across three days at @aiDotEngineer Europe, from the keynotes to the booth to brunch the morning after.
Huge thanks to the people who came all this way to share their thoughts (I hope I've done you justice): @badlogicgames @mattpocockuk @mitsuhiko @milksandmatcha
arize.com/blog/code-is-free-technical-debt-isnt-notes-fro… -
R 'Nearest' Nabors
@rachelnabors
Well, I have unintentionally been roped into building agent harnesses for the foreseeable future
-
R 'Nearest' Nabors
@rachelnabors
Well, the clock is on the screen now
Kol Tregaskes
@koltregaskes
Ex-MIT researcher Isaak Freeman quits his PhD and drops the 50,000 H100 GPU roadmap to emulate a full human brain.
He mapped the entire path from 302-neuron worm to 86-billion-neuron human with connectomics costs now at 100 dollars per neuron and data acquisition via advanced microscopes as the only blocker left - digital humans just got a realistic timeline.
pdf.isaak.net/thesis
R 'Nearest' Nabors
@rachelnabors
My immediate thought was "one day kids are going to immortalize terrible parents and torture/interrogate them until they get bored"
-
R 'Nearest' Nabors
@rachelnabors
HEy web devs, are you using the Prompt API? I want to hear from you!
-
R 'Nearest' Nabors
@rachelnabors
Been waiting half my life to FINALLY see Akira—and on a huge screen to appreciate every detail of the animation!
1) It is every bit the masterpiece of animation and cinematography it is reputed to be
2) It actually features women who aren't usually mentioned in reviews: Kei, Kaori, Kiyoko
R 'Nearest' Nabors
@rachelnabors
Yes a thousand times yes
-
R 'Nearest' Nabors
@rachelnabors
Giving your agent access to the web? Consider adding this sanitization layer to defend against injection threats: github.com/sysk32/trapwatch
-
R 'Nearest' Nabors
@rachelnabors
Weekend mini-hack complete: now I have a little Claude pet who can hassle me when I'm needed in the loop 😈
-
R 'Nearest' Nabors
@rachelnabors
I miss when we had conversations here. I see you liking my posts. But I miss hearing from you
-
R 'Nearest' Nabors
@rachelnabors
I'm curious what the web Devs out here make of agents/MCP/etc. On your radar? Actively building for? Defending against? Old news?
-
R 'Nearest' Nabors
@rachelnabors
In the last two weeks, I learned that VC firms will give dudes a pile of money "just to experiment" based on vibes.
I would love to be paid to fuck around and find out. Especially with browser tech.
My vibes are meh, but I'm doing it all the same. Just.. quietly... -
R 'Nearest' Nabors reposted
Maximiliano Firtman
@firt
Starbucks (US) just published an MCP App, now listed in the ChatGPT app directory.
👉 It's a web app running as a mini-app!
I’m wondering how many of you are interested in learning how to build these experiences.
👉 I'll be speaking about this at iJS San Diego in early June!
Maximiliano Firtman
@firt
Remember: it's a web app + other APIs, all served through an MCP server.
It's kind of a mini-app using <iframe> in ChatGPT web, and web views in native ChatGPT apps.
APIs are exposed through `window.openai` in JavaScript. Widgets are served using `text/html+skybridge` MIME. -
R 'Nearest' Nabors
@rachelnabors
Morning, people? What are we shipping today?
-
R 'Nearest' Nabors
@rachelnabors
I cannot overstate what an honor it was to share a stage and a city with this group at @aiDotEngineer Europe.
The morning after, I put on one of my mini speaker brunches for all the folks heading home, but not ready to say goodbye. We talked about everything from MCP apps to what "good" even means for agents right now. Priceless <3 @liadyosef @nickytonline @evalstate Ana Kundush (@alpic_ai) @petergostev
Deserved shoutouts:
@marlene_zw for connecting everyone
@Infoxicador @liadyosef @SamMorrowDrums bringing the MCP energy
@badlogicgames being quietly the most interesting person in agents (and countering bro energy with dad energy!)
@seldo captivating a room with a 2-hour evals workshop
And @swyx & co for organizing it all.
I am so lucky to be in rooms with people like this. Thank you all. 🖤 -
R 'Nearest' Nabors
@rachelnabors
Writing recap post of AIE, and found this in my photos! What a great time! Only complaint: afterparty needed a little more wub-wub on the floor <3 @TejasKumar_ @nickytonline
-
R 'Nearest' Nabors reposted
Vinicius Dallacqua
@WebTwitr
@rachelnabors small (and local) models FTW.
Since web.dev/articles/ai-model-selection it has been my obsession build harness that elevates small models.
I'm even running the E4B quant which can even run on a smartphone (wink wink, basis for the work I'm doing on my other side project 😃) -
R 'Nearest' Nabors
@rachelnabors
Mima hacknight
-
R 'Nearest' Nabors
@rachelnabors
Pilates before drum practice. I have regrets
-
R 'Nearest' Nabors
@rachelnabors
Wondering if I can fork @idosal1's agent orchestration GUI and replace orcs with cats so I can literally "herd cats" all day 🤔
-
R 'Nearest' Nabors
@rachelnabors
Judging an engineer by tokens consumed or lines of code written is like judging a writer by the length of their book
-
R 'Nearest' Nabors
@rachelnabors
Something heard more than once at @aiDotEngineer Europe: "learn to say no" (to your agents).
Well, heard verbatim from @steipete and @badlogicgames! So this might be coming from the OpenClaw and adjacent communities!
R 'Nearest' Nabors
@rachelnabors
“Go slow” showed up in multiple slide decks. This is another theme I’m seeing in @aiDotEngineer talks (@mitsuhiko @cristinaponcela come to mind):
“Code isn’t cheap.” “Slow TF down.”
It’s like we’ve all been given credit cards, and we can rack up thousands of “dollars” of technical debt without realizing it! -
R 'Nearest' Nabors
@rachelnabors
.@mattpocockuk's skills repo is such an asset for loop engineers: github.com/mattpocock/skills/
Check out grill-me, ubiquitous-language, tdd, improve-codebase-architecture. Get the full rundown on how to use in your agentic coding flow: www.youtube.com/watch?v=O_IMsEg91g8&t=27759s -
R 'Nearest' Nabors
@rachelnabors
Catching up on all the keynotes I missed at
@aiDotEngineer Europe because I was on the hallway track.
"Leetcode interviews are design to filter for smart people who will put up with the bullshit necessary to keep the job."
@GergelyOrosz never fails to tell it like it is
R 'Nearest' Nabors
@rachelnabors
Check @GergelyOrosz's fireside with @swyx out:
www.youtube.com/watch?v=O_IMsEg91g8&t=27759s
R 'Nearest' Nabors
@rachelnabors
When I was dating, I remember using "reaction to leetcode" told me a lot about the person on the other side of the table. If the other person vociferously loathed and refused to study for them, that told me that person let their ego/ideals come before pragmatism.
$100k salary bump if you study for 6 months and pass a bullshit interview hurdle is a pretty good deal. I'm the kind of person who sees that game and puts money on the table.
R 'Nearest' Nabors
@rachelnabors
*ARE DESIGNED
They typos let you know a human wrote it.
Handmade tweets. A real cottage industry -
R 'Nearest' Nabors
@rachelnabors
“madness with a touch of science fiction”— @steipete on what it takes to develop your pipedream in opensource
R 'Nearest' Nabors
@rachelnabors
From this fireside at @aiDotEngineer www.youtube.com/watch?v=O_IMsEg91g8&t=8893s
-
R 'Nearest' Nabors
@rachelnabors
Did you know @huggingface has a massive MCP server? I just caught up with @evalstate, creator of Fast Agent, about what it is and how to use it: youtu.be/ZxAi56KLaW0
-
R 'Nearest' Nabors
@rachelnabors
There was one final @aiDotEngineer get together, the traditional Farewell Dear Speakers brunch attended by @liadyosef @nickytonline @evalstate @petergostev
I hope everyone is home and rested for Monday 🥞🍳🥯 -
R 'Nearest' Nabors
@rachelnabors
I crawled out of my @aiDotEngineer recovery cocoon long enough to see @Snugug while he was in town. Can you believe I met this dude all the way back at CSS Conf EU?
Also, isn't saying "I can write CSS" a bit like saying "I can drive stick." No one does that anymore, and everyone hates it, but deep inside, folks admire your skillz -
R 'Nearest' Nabors
@rachelnabors
I'm not sure if I should opensource Mima or keep her closed for now. If I open her up, people might help her move faster. But then again, we've been moving fast enough in a private repo, sans distractions.
Do I know any founders who have faced this crossroad? -
R 'Nearest' Nabors
@rachelnabors
Mima is getting some design love next week: logo, design system, and more projects commence.
Meanwhile, on the engineering side, it's time to bring in the SLMs and evals -
R 'Nearest' Nabors
@rachelnabors
This
Shaun Smith
@evalstate
One way you could obfuscate model performance is by making benchmarks too expensive for hobbyists to run.
-
R 'Nearest' Nabors
@rachelnabors
"Don't play games you can't win" pairs well with, "hope is when there's a path to the goal; wishful thinking is when there is not."
Many people will tell you "you just have to believe" or "manifest it" or "you can be the first." Those people are easy marks for MLMs and other scams -
R 'Nearest' Nabors
@rachelnabors
"If you let quality slip, slowly over time, users will defect to your higher quality competitors," @artman at @linear on why being a market leader is no excuse for ignoring user experience.
I wish this held true for browsers
R 'Nearest' Nabors
@rachelnabors
"Engineers will need to become mini-s and understand the user experience..."
Ho boy.
It's going to be a lot easier for PMs to become mini-engineers than versa. I've met many engineers who don't have the empathy to make good design decisions -
R 'Nearest' Nabors
@rachelnabors
Are you at @aiDotEngineer's AI Engineer Europe? Come to St James at 2:50pm and catch my talk rehabilitiating a webcomics server to serve both browsers and agents!
-
R 'Nearest' Nabors
@rachelnabors
Every client I've had, I've ended up onboarding into @linear. I don't mean to do it (I'm an Asana person, personally). But it's so easy for teams to move fast with it, and there's usually a team already at the company who wants to try it out. So I give them a little workshop, give them some guidance for the first few weeks and then BAM, everyone is making it their own. Lovely
-
R 'Nearest' Nabors
@rachelnabors
I rarely go squee anymore, but squee!
R 'Nearest' Nabors
@rachelnabors
True story @GergelyOrosz is one of the few people whose posts I reply to anymore. Often, he replied back. It's the kind of tech conversation I miss from Olde Twitter
-
R 'Nearest' Nabors
@rachelnabors
Day 2 of AI Engineering Summit in London: new faces and old friends!
-
R 'Nearest' Nabors
@rachelnabors
Live from AI Engineering Summit, it's not quite Saturday Night!
-
R 'Nearest' Nabors
@rachelnabors
Amazing, the same tried and trusted guardrails required to keep human engineers from committing bad code also works for agents
Charlie Marsh
@charliermarsh
Tragically I am continuing to find that the most effective guardrail against slop is extremely talented engineers doing very thoughtful, human code review
-
R 'Nearest' Nabors
@rachelnabors
Loop engineering is the future of engineering. Oh look, there's a language for that!
-
R 'Nearest' Nabors reposted
Vinicius Dallacqua
@WebTwitr
WebMCP is a great new spec and I've been playing around with it in my own implementation into my web browsing agent harness.
It brings interactions to a fraction of the time and sessions get faster and also 'cheaper' for the models.
I had my agent browsing around @rachelnabors comics and using the exposed tools to find me recommendations to read.
@ChromiumDev have done something amazing here and I can't wait for more sites to have it available.This media couldn't be displayed. See it on the original post.
R 'Nearest' Nabors
@rachelnabors
Don't mind me just hacking my website with WebMCP. I'm telling you WebMCP is gonna be a boon for accessibility. Also a boon for scrapers!
Come see more at @aiDotEngineer in London on Friday :) -
R 'Nearest' Nabors
@rachelnabors
Shout out to @AlexNahasDev and @evalstate for technically reviewing the WebMCP and transports portions of my @aiDotEngineer summit talk! I'm so excited to share how HTML, CSS, JS, and HTTP are changing how we use agents—and how agents are changing how we build for the web!
-
R 'Nearest' Nabors reposted
Shaun Smith
@evalstate
This is an awesome talk - if you are at AIE you should go.
R 'Nearest' Nabors
@rachelnabors
Shout out to @AlexNahasDev and @evalstate for reviewing the WebMCP and transports bits of my talk for @aiDotEngineer in London! HTML, CSS, JS and HTTP are changing how humans interact with agents! See you Friday
-
R 'Nearest' Nabors
@rachelnabors
Shout out to @AlexNahasDev and @evalstate for reviewing the WebMCP and transports bits of my talk for @aiDotEngineer in London! HTML, CSS, JS and HTTP are changing how humans interact with agents! See you Friday
-
R 'Nearest' Nabors
@rachelnabors
I went from only having one pinned repo at the start of the year to FIVE. I've never bothered with sharing on GitHub like this before! And they're all weird art/opensource projects
github.com/nearestnabors -
R 'Nearest' Nabors
@rachelnabors
A big part of how I learn these days is I Vibe code with Claude and then at the end ask Claude to write me a tutorial for the project and I go through and ask questions about things I don't understand and we end up with a very nice tutorial at the end.
I will publish one soon, promise 😌
R 'Nearest' Nabors
@rachelnabors
In this one I ended up coming back to Claude, asking, "why are we using a proxy? I'm able to run it without a proxy." Turns out Claude was out of touch. Removed it, updated the tutorial.
Is it perfect? No. But... I've never learned perfectly nor built perfectly. The stumbles make it stick -
R 'Nearest' Nabors
@rachelnabors
The full interview with @blackgirlbytes about @goose_oss, the agent that shaped MCP: www.arcade.dev/blog/goose-the-open-source-agent-that-shap…
-
R 'Nearest' Nabors
@rachelnabors
Don't mind me just hacking my website with WebMCP. I'm telling you WebMCP is gonna be a boon for accessibility. Also a boon for scrapers!
Come see more at @aiDotEngineer in London on Friday :)
R 'Nearest' Nabors
@rachelnabors
Want to learn more about webMCP? Check out the interview with its creator @AlexNahasDev : www.arcade.dev/blog/web-mcp-alex-nahas-interview
-
R 'Nearest' Nabors
@rachelnabors
Another way to think of MCP Tools vs Resources: Tools are like functional programming. Every action comes at the end of a function call. And your agent has to actively decide which calls to make using context. MCP resources are more like a file system that the agent harness can choose to add or remove from the environment, no context required. mCP resources are actually the better choice for adding documentation to your agent environment. Agents don't like calling tools to access documentation and I don't blame them. Every page read is a decision. Something they have to think about and figure out. Resources are just a way to preload their context with relevant information. Then they can have access to without having to think.
Imagine a coding harness that adds MCP resources to the context based on what's in your config files. Or a book reading app that preloads transcriptions for blind comic readers to generate audio plays with -
R 'Nearest' Nabors
@rachelnabors
"If the user knows they want to include project documentation before asking a question, why should that documentation’s availability be described in a tool manifest? Why burn tokens on a tool description the model has to parse and reason about?"
MCP Resources might be the hero we need in an era of tool bloat layered.dev/mcp-resources-the-overlooked-primitive/ -
R 'Nearest' Nabors
@rachelnabors
In four days, I'll be sharing how to evolve and adapt the web for users browsing by Agent using MCP Apps, WebMCP, and more! I'll show you how I retooled my site to serve comics (and rich comics experiences) not just to browsers but to agents. See you at AI Engineer Summit Europe in London! www.ai.engineer/europe/schedule?session=124
-
R 'Nearest' Nabors
@rachelnabors
The thing no one wants to admit about AI slop is that most of what humans create is terrible to mid. That's why slop is so mediocre.
No one wants to admit their interns and juniors, despite their best efforts, are producing work aI could do just as badly. No one wants to admit that their infra is held together with duct tape and prayers.
We are horrified that people might accept code they've never reviewed, and yet, how many of us review what's in our code bases, or the PRs approved by people who, let's be honest, are pretty mid themselves.
We are mid. Let's not put on airs about it.
It doesn't stop at code. The art AI produces is slop because people love mid, cheap illustrations of their loved ones. They want to read mid romance novels about average characters. They want mid cartoons to entertain their mid kids so they can peacefully enjoy some mid spaghetti in a midpriced restaurant.
But also... it's kinda freeing. Slop is good enough. Always has been. Always will be. The world runs on slop -
R 'Nearest' Nabors
@rachelnabors
"starfish UI"
David K 🎹
@DavidKPiano
Hot take: an empty chat box as the primary UI is the exact *opposite* of intelligence in a product
It's like saying "I have no idea what you want to do here, tell me"
This is why pre-AI UIs feel better IMHO
Rabi Shanker Guha
@rabi_guha
notice something?
Linear, PostHog, Attio - all shipped the same thing in the last few weeks. Homepage is a chat bar - not a dashboard.
This is the SaaS industry quietly admitting that traditional UI doesn't work anymore. Every user is different. One homepage can't serve them all.
The playbook is shifting:
→ expose your core APIs
→ connect an agentic layer
→ let users use software the way they want
SaaS became chat. Chat will become Generative UI - the agent won't just reply in text, it will compose the interface itself.
We're closer than people think. -
R 'Nearest' Nabors reposted
David K 🎹
@DavidKPiano
Hot take: an empty chat box as the primary UI is the exact *opposite* of intelligence in a product
It's like saying "I have no idea what you want to do here, tell me"
This is why pre-AI UIs feel better IMHO
Rabi Shanker Guha
@rabi_guha
notice something?
Linear, PostHog, Attio - all shipped the same thing in the last few weeks. Homepage is a chat bar - not a dashboard.
This is the SaaS industry quietly admitting that traditional UI doesn't work anymore. Every user is different. One homepage can't serve them all.
The playbook is shifting:
→ expose your core APIs
→ connect an agentic layer
→ let users use software the way they want
SaaS became chat. Chat will become Generative UI - the agent won't just reply in text, it will compose the interface itself.
We're closer than people think. -
R 'Nearest' Nabors
@rachelnabors
"Why did you choose Netlify?"
Because I'm committed to a drama-free lifestyle.
R 'Nearest' Nabors
@rachelnabors
It seems many people want a parasocial relationship with their hosting provider.
I find that creepy and gross.
My provider is not my boyfriend, my uncle, or my bestie. Not my mentor, not the Academy Awards. All I want is a UI and developer experience that is better than AWS. I can Claude and Ralph my way through the rest of it.
This is the relationship that I want with my service providers: -
R 'Nearest' Nabors
@rachelnabors
Social and content moved to apps, leaving the web with splash pages and dashboards.
I really need to write about how dashboards, the last bastion of the web and cornerstone of SaaS are changing, first in dev tooling, next in non-tech tooling
Rabi Shanker Guha
@rabi_guha
notice something?
Linear, PostHog, Attio - all shipped the same thing in the last few weeks. Homepage is a chat bar - not a dashboard.
This is the SaaS industry quietly admitting that traditional UI doesn't work anymore. Every user is different. One homepage can't serve them all.
The playbook is shifting:
→ expose your core APIs
→ connect an agentic layer
→ let users use software the way they want
SaaS became chat. Chat will become Generative UI - the agent won't just reply in text, it will compose the interface itself.
We're closer than people think. -
R 'Nearest' Nabors
@rachelnabors
I interviewed @blackgirlbytes about @goose_oss's role in MCP and agent standards (plus building your own "Claude" with local SLMs!) Goose is my favorite opensource agent harness, and Rizel is one of my fave builders!
www.youtube.com/watch?v=Kn8MaWcNJgA -
R 'Nearest' Nabors reposted
CodeTV
@codetv_dev
Watch the full conversation: codetv.dev/series/web-lunch/s1/rachel-lee-nabors
-
R 'Nearest' Nabors reposted
CodeTV
@codetv_dev
"That's how I play life now. More like a video game." — @rachelnabors
-
R 'Nearest' Nabors
@rachelnabors
If you're coming to London next week, Arize is having a happy hour at The Ivy. I'll be at the speakers dinner, but you should go. Ooh la lah!
luma.com/aie-europe-arize -
R 'Nearest' Nabors
@rachelnabors
Have you noticed how everyone is a "xyz leader" these days (self included)?
Maybe it's because things are changing, and we're all moving on to learn new things. Feels better to blog about it and "lead thoughts" than get swept along as a... "curious IC?" "Reskilling grunt?"
Sign of the times -
R 'Nearest' Nabors reposted
Juan Pa
@JuanPa
Really good results for a 32B parameters. You can potentially run it on laptops and desktops without much problems.
Google DeepMind
@GoogleDeepMind
Meet Gemma 4: our new family of open models you can run on your own hardware.
Built for advanced reasoning and agentic workflows, we’re releasing them under an Apache 2.0 license. Here’s what’s new 🧵 -
R 'Nearest' Nabors reposted
Arcade.dev | The Actions Runtime
@TryArcade
Goose was the first public MCP client. Before MCP even had a name, Block was working with Anthropic on connecting AI agents to external tools.
@rachelnabors sat down with @blackgirlbytes on how it became the MCP reference implementation — and what MCP Apps mean for developers👇 -
R 'Nearest' Nabors reposted
nicole
@nicolekcha
has anyone already defined a term for early stage AI adoption psychosis?
symptoms:
- shipping slopware at enormous speed
- bragging about silly metrics like LOC
- slopdropping llm generated walls of text (unverified) on your coworkers/friends/feed
- not being able to control the quality of the output but not noticing it
causes:
- caused by the euphoria of doing stuff you previously had to do manually
- or couldn’t do at all
- and lacking humility eg “why isn’t everyone putting out so much, i must have some unique valuable insight”
- not realizing there is still way more to learn
(to be clear a lot of people do have unique valuable insights — but usually it’s after you’ve hit the walls and had to humble yourself to learn what’s going wrong and you don’t just shit gold bricks)
gregorein | chibi Malfoid on cults3d
@Gregorein
for added context: when a 17yo developer (@xiaonweb) politely pointed out that bragging about LOC is silly, Garry's response was to publicly call them a "clout farmer."
the "clout farming" teen... wrote a browser engine in Rust at 17. HTML tokenizer, CSS cascade, box model layout, GPU renderer via wgpu, and published a technical breakdown showing deeper understanding of how the web works than most senior engineers I've worked with (including me, cos I've never dug that low-level).
vs
the "shipping" guy, the president of Y Combinator, a multi-billion dollar startup kingmaker, who mass-generates code with 113 Claude sessions a week, counts lines like a Duolingo streak, and ships test files, 0-byte AVIFs, and 4 MB uncompressed PNGs to production.
right, punch down at a teenager. on main @.This post is unavailable. See it on X. -
R 'Nearest' Nabors
@rachelnabors
Becoming 20% cooler
-
R 'Nearest' Nabors reposted
Gergely Orosz
@GergelyOrosz
Ironic how Anthropic sells Claude Code security reviews positioned as something v powerful (costing $15-25 per PR review), and being clear they use it on all PRs... then leaking all of Claude Code's code thanks to publishing their sourcemap.
AI won't save you from yourself! -
R 'Nearest' Nabors
@rachelnabors
My talk for AI Engineering Summit London is about the web on agents/agents on the web. I'm restoring my old comics archive (now with RWD!) and making the comics available via MCPApps as well as with WebMCP. Goal is "one source, multiple protocols," and so I've been trialing different providers
-
R 'Nearest' Nabors
@rachelnabors
Don't mind me. Just merging some contributions made possible by the NOT FAILING E2E tests while my loop gently completes
-
R 'Nearest' Nabors
@rachelnabors
This is my "I got the E2E tests to stop failing in GitHub" face.
I've never been a test engineer (I'm good at making UI do tricks), but if you're using loops to engineer major portions of your project, being GOOD at testing makes those loops more valuable.
I feel like any good AI Engineering book would just be "how to write good specs" and "testing best practices." That said, what's your go-to on test engineering? -
R 'Nearest' Nabors
@rachelnabors
One thing that evaporated since leaving the React team and moving back to London: transactional relationships. It was very different from my cozy, meritocratic CSS community relationships that I'd nurtured for years. People were sidling up to me near constantly
R 'Nearest' Nabors
@rachelnabors
Continued thoughts bsky.app/profile/nearestnabors.com/post/3mibeypvfos2u
-
R 'Nearest' Nabors
@rachelnabors
The unglamorous side of AI engineering: spending your sunday getting all your E2E tests to pass when you didn't build them yourself while a Ralph Wiggum loop quietly debugs a Pass 1 project in another terminal
-
R 'Nearest' Nabors
@rachelnabors
Today a fellow gym-goer said she likes to work out next to me because I inspire her to try harder.
"You're so strong!"
🥹💪 -
R 'Nearest' Nabors
@rachelnabors
Was talking with a young founder, and I said that it's less important to be popular on X, more important to spend time with who you're building for. X is a popularity contest that sometimes results in investment. Traction with users is more a more reliable incentive to invest
-
R 'Nearest' Nabors
@rachelnabors
Buying a home sucks.
-
R 'Nearest' Nabors reposted
Scott Tolinski
@stolinski
Can we get this in chrome please
池田 泰延
@clockmaker
春恒例の大型アップデート、Safari 26.4がリリース!
他のブラウザに先駆けて、CSS Gridの新しい「grid-lanes」が搭載。
かつて流行ったMasonryレイアウトが、CSSだけで簡単にできます。This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
Londmaxxing/founder mode
-
R 'Nearest' Nabors
@rachelnabors
Coworking at my fave tea house in Camden with one of my fave founders @catinwool You?
-
R 'Nearest' Nabors
@rachelnabors
I think venture capital overestimates the creativity of consumers and their drive to build their own agents and apps.
When I was a cartoonist, winning awards and self-publishing graphic novels and mini comics for teenage girls, I had a distaste for "talkers." Talkers are people always talking about the great idea that they'll one day make (maybe you could do some free work and help them?). They were in love with their own ideas, but somehow never seemed prioritise doing the work to realise those ideas.
Mind you writing and drawing are free activities and nothing stops a person from putting pen to paper and describing the comic in a script and shipping that around to different Publishers and artists to try to get by-in. The barrier to entry is literally words.
In this new era, the barrier to entry is also literally words. But words and ease of use are not what stops people from doing the things that they want to do. A motivated person will learn how to code or how to speak a new language or how to draw because they want what is on the other side of that effort. Many people have ideas, but they do not prioritise creation or want to be a creator, deep in their heart. Other things are more important, like their pride or their leisure time.
Not everyone is a creator. There is a smaller total addressable market of creators than most investors and founders realise.
Of course you can help more creators ship by reducing the level of effort required to create. But all you're doing is expanding the total addressable Market, which will always be a fraction of the total population. It is much smaller than you think, as most people do NOT have ideas. And of those who do, even fewer have motivation to make their ideas reality, even when the barrier to it is removed (think of powerful people with many resources who spend their time sleeping on yachts Vs building monuments or writing books). Only a fraction of the population wants to build.
"In this world, there are Talkers and there are Doers. I am a Doer." This is what I used to say to young people who wanted to make comics and ask for advice. Lately, I hear my younger self's words ring in my ears. I would like to be more like them
R 'Nearest' Nabors
@rachelnabors
Also, I'm sorry for all of the typos. I dictated this and posted it quickly
-
R 'Nearest' Nabors reposted
°‧ 𓆝 bree 𓆞 ·。
@yummierbeefcake
how it feels seeing that one creator that used to be popular become irrelevant
-
R 'Nearest' Nabors
@rachelnabors
"The barrier to building dropped to zero. The barrier to shipping didn’t." shift.infinite.red/you-built-an-app-this-weekend-now-what…
-
R 'Nearest' Nabors
@rachelnabors
I have some cool experiments I'd like to ship to the world. It's been awhile since I've had... a Hosting Provider of Choice. What do people use even?
My experiments range from fully localStorage and local SLM to semi-products with user accounts. I would rather not have multiple providers because I am L&L ("lazy and loyal"). What are you using? -
R 'Nearest' Nabors reposted
Angie Jones
@techgirl1908
Excited to share that I've stepped into a new role as VP of the Agentic AI Foundation (AAIF) 🎉
As AI reshapes our industry, it's critical that this technology evolves in ways that are open and collaborative.
Hosted by the Linux Foundation, @AgenticAIFdn is a neutral organization focused on advancing adoption of key open source efforts like MCP, AGENTS.md, and Goose. Our members include Anthropic, OpenAI, Google, AWS, Microsoft, and 100+ others.
This work is bigger than any single company or product. I'm ecstatic to help drive AI adoption and interoperability at an industry-wide level. -
R 'Nearest' Nabors
@rachelnabors
Wondering why you have to go through London to reach Cambridge and Oxford. You'd think there'd be an international airport serving both universities with flights to SF and China. I feel London gatekeeps the tech industry in the UK. Everything is filtered through the Fin Bros
R 'Nearest' Nabors
@rachelnabors
You guys would not believe how angry this post got on Threads. 123 pearl-clutching replies about how this ignorant American hates trains and has never been to either city.
Meanwhile, me, on a train, stoically returning from Oxford and wondering why the heck I can't live on a route straight to the real AI and CS hubs of the UK (throw in an overnight to Edinburgh, heyo!) -
R 'Nearest' Nabors
@rachelnabors
Blenheim Palace maxxing
-
Post not available here, view on XPost not available here, view on X
R 'Nearest' Nabors
@rachelnabors
Although today, it's more likely your loop engineering. Slowing down long enough to set up your loops will help you ship better, faster
-
R 'Nearest' Nabors
@rachelnabors
"prompt learning" or.... "promptmaxxing"?
github.com/Arize-ai/prompt-learning -
R 'Nearest' Nabors
@rachelnabors
If you're running more than one coding agent, you're running into localhost conflicts. Portless will fix that for you AND it works with git worktrees! github.com/vercel-labs/portless
-
R 'Nearest' Nabors
@rachelnabors
For every business idea I've had in the past six months, I can think of how someone could build it themselves "good enough" with our current resources. When everyone can make a smooth pitch deck and launch video and prototype, what's good to invest in anymore??
Anton Osika
@antonosika
Introducing Lovable for more general tasks.
Lovable has always been for building apps. Today it also becomes your data scientist, your business analyst, your deck builder, and your marketing assistant.
This is a big step toward what Lovable is becoming: a general-purpose co-founder that can do anything.
See examples below.
R 'Nearest' Nabors
@rachelnabors
This is why I don't think pivots into "everyone will build apps/agents!" will succeed. Your business model will only succeed if your users succeed. And 99.8% won't. The creator economy of tomorrow is not everyone building little agents and apps and buying them off one another
-
R 'Nearest' Nabors
@rachelnabors
Several leaders I deeply respect and admire have reached out to me to ask how I became "an AI expert." (I am not. Few can claim to be.)
What happened was that I took time off after layoffs to read textbooks about machine learning and joined the budding AI and startup community here in London. Then I started consulting with startups in the Bay Area who are trying to reach the developers I understood so well. I was able to integrate with their engineering teams and learn so much from them.
The best thing you can do is to work with curious people who are using the tools, reading the papers, and building things. "Fear is the mind killer." And if you or your team is afraid, you need to change that.
I keep prioritising working with people who are fearlessly learning and building and with leadership that trusts me and "lets me cook."
I traded coveted titles for knowledge and experience, and surrounded myself with builders, doing work instead of talking about it.
Perhaps the right long-term play was to stick to "Titles & Talking." But these have been some of the most interesting and exciting years of my tech career. I have learned so much more about how to pivot and move fast from all the startups I've been with; how to build with AI, and how to see into the future. I feel I've grown so much more professionally than if I had kept hunkering down in Big Tech.
To all I have worked with and who have taken a chance on working with me, I'm grateful.
I am so happy. -
R 'Nearest' Nabors
@rachelnabors
Heads-up, if you're on a Vercel free or trial plan, you "agree" they can use your projects to train AI
David Cramer
@zeeg
Guess ill move off of Vercel for (unpaid) personal projects at this point.
I dont mind if people ask permission for this, and everyones free to make their own choices, but I am your funnel, not your product. -
R 'Nearest' Nabors
@rachelnabors
London turned out for @arizeai's double header on prompt learning & @StackOneHQ on tool selection systems. Thank you to everyone who came!
-
R 'Nearest' Nabors reposted
Arcade.dev | The Actions Runtime
@TryArcade
Coding agents already know how to use the CLI. So why are we building elaborate MCP integrations instead?
Our Head of Engineering built mcpx to answer that. curl for MCP.
www.arcade.dev/blog/curl-for-mcp -
R 'Nearest' Nabors reposted
Rizèl Scarlett 🇦🇬🇬🇾
@blackgirlbytes
my last blog post working at this company 🫡
end of an era
goose
@goose_oss
Wondering what WebMCP is?
❌ It's not an MCP server
❌ It's not the same thing as Playwright MCP
✅ Instead of automating the UI, WebMCP turns your website into callable tools for AI agents.
Read our blog post to learn more! Link the thread 🧵 -
R 'Nearest' Nabors reposted
Dillon Mulroy
@dillon_mulroy
i am a simple man, i see codemode i retweet
R 'Nearest' Nabors
@rachelnabors
I recently was building with TanStack AI and just had to interview @jherr about its upcoming codemode features and how chat apps are the "hello world" of agent development! Check it out: www.youtube.com/watch?v=DKkE_-Ky3AA
-
R 'Nearest' Nabors
@rachelnabors
I recently was building with TanStack AI and just had to interview @jherr about its upcoming codemode features and how chat apps are the "hello world" of agent development! Check it out: www.youtube.com/watch?v=DKkE_-Ky3AA
-
R 'Nearest' Nabors
@rachelnabors
This isn't a Y. It's an X.
—Claude
R 'Nearest' Nabors
@rachelnabors
The thing no one Xes is Y.
—Claude -
R 'Nearest' Nabors
@rachelnabors
I wanted to hear more about Chrome's MCP dev tools, so I interviewed Chrome's Michael Hablich about them! Your agent can inspect the console, perform UI tests, and validate its own work so easily Chrome's MCP dev tools. I feel like these are going to change how we make UI for the web.
The video: www.youtube.com/watch?v=CIOcjjgl9MI
The blog post: www.arcade.dev/blog/stop-pasting-chromes-console-logs-int…
Happy codoing! -
R 'Nearest' Nabors
@rachelnabors
I'll be joining Future of Frontend in Finland again this year to talk about "The Headless Web"—how we build content and UI when our main users are agents! This is one of my most treasured conferences. Hope to see you there! futurefrontend.com
-
R 'Nearest' Nabors reposted
Robert Balicki (👀 @IsographLabs)
@StatisticsFTW
Introducing Barnum, or... how I ship hundreds of PRs per week, burn through backlogs, and automatically fact-check documentation.
LLMs are incredibly powerful tools. But when we try to use them to drive more complicated refactors or more intricate workflows, their shortcomings are quickly revealed. When their context gets full, they get forgetful, and they can't be relied upon to necessarily do the steps that you ask. They often cut corners.
Put simply, having an inherently probabilistic process perform what should be deterministic work necessarily comes at the cost of reliability. And you can't build a complicated workflow off of unreliable foundations.
That's where Barnum comes in. Barnum is the missing workflow engine for agents. Rather than having agents be responsible for upholding guarantees (e.g., always lint and commit your changes atomically), agents instead do just what they're good at: reading text and reasoning. Everything else is done deterministically, on the outside, by Barnum.
This means that you can build bigger, more involved workflows without sacrificing reliability. Because you can intersperse bash scripts, you save on token usage. The agents performing a micro-task only receive the instructions for that specific task, meaning that context does not get overwhelmed and they don't get forgetful. And because all inputs, outputs, and transitions are validated, the agents can't wriggle out of doing the work.
This workflow is essentially a state machine described in a config file. And the best part? The configuration has a JSON schema, so agents are actually really good at writing the workflow!
It's already been used to ship hundreds of PRs, run automated refactors, burn through various backlogs, fact-check every statement in documentation, and build a deep-research clone!
The attached image is a representation of the workflow that I use to identify and implement automated refactors. I follow this up with a separate workflow that splits each commit into a separate PR, judges the refactor, and potentially completes the refactoring (for example, by modifying call sites if the refactor changed some public API).
So go on, give it a try. Check out barnum-circus.github.io, star the repository, and join the Discord! I can't wait to see what you build with it! And I'd love for you to get involved! -
R 'Nearest' Nabors
@rachelnabors
Is anyone else noticing that, with increased agentic coding, they remember more from previous decades?
-
R 'Nearest' Nabors
@rachelnabors
Hard to imagine being cross that someone made something with something you open-sourced. But even so... to spend so much time writing a blog post instead of just DMing someone feels like we're in high school.
"Samantha, tell Tabitha that if she wants to fork my repo, she needs to clear that with me first." Meanwhile, Tabitha is sitting right across the table, a mere DM away -
R 'Nearest' Nabors
@rachelnabors
Ugh. Mima was working so well, now this. The more this platform locks down its APIs while charging through the nose to use them, the more cool shit we'll build with ATProto. You know that, right? Right now, the only thing X APIs are good for is a firehose of posts. Not DMs, not conversation. Just getting a feed. The APIs are not designed for conversation or community building. They're design for data harvesting. It's the simple truth.
Well. There's always Puppeteer!
Developers
@XDevelopers
To help address automated reply spam, programmatic replies via POST /2/tweets are now restricted for X API.
You can only reply if the original author @ mentions you or quotes your post. Non-replies will remain unchanged.
Applies to Free, Basic, Pro, Pay-Per-Use. Note: Enterprise & Public Utility apps are not impacted.
Details: devcommunity.x.com/t/update-to-reply-behavior-in-x-api-v2…
R 'Nearest' Nabors
@rachelnabors
To all the people replying to this saying, "Great, this will get rid of the bots!" That was never the purpose of this API change.
The only way to get rid of bots on this platform is to identify accounts that are engaging in bot-like behaviors and closing them. But that would make X's bottom line look bad. It would have a shrinking userbase. Recently I cleared out my followers of bots using CircleBoom because it was hurting my reply-to-follower ratio, and it turns out a third of all the accounts following me were dead or bot accounts. (Dead accounts often get hacked to become bot accounts, interestingly) -
R 'Nearest' Nabors reposted
Simon Willison
@simonw
Just added the 12th chapter to my Agentic Engineering Patterns guide, but it's the first one in the sequence: I figured it was time to try and answer the obvious question, "What is agentic engineering?" simonwillison.net/guides/agentic-engineering-patterns/wha…
-
R 'Nearest' Nabors reposted
the tiny corp
@__tinygrad__
ok ok hear me out. what if we did space datacenters but on earth? like we build them all rugged and good, ready to withstand temperatures, low maintenance, fits on the back of a truck, all ready to go to space, but then we ... don't send them to space.
sending things to space is expensive. if we keep them on earth, we can send them to places by truck, which is a lot cheaper than space.
i don't know what i was thinking about buying land and building a building. that's so modernist. we have $5M and I thought we needed to raise to amortize the fixed costs of operating a site. it was stressing me out.
but then i remembered space datacenters. where we're going, we don't need a site. i mean, yea, we do, and we have to lease it, but we'll lease anything where it's cool, has cheap power, and has fiber. if the public utility decides to rug us and raise prices, no lawyers needed, just fire the gas thrusters! actually we don't even need gas thrusters, we'll put it on a truck and go to the next leased site.
the minimum quantity we can do this at is one, and one should only cost like $3M. we have $5M, we don't even need to raise, just build the one, watch it print money, then build the next one with the money. self replicating space datacenters on earth.
so yea there's a lot of software work to do to make tinygrad run LLMs at really high tok/s and be ready to deploy for the RDNA5 launch. gotta focus on that. raising money, buying land, and reading utility contracts are rabbit holes. got out just in time.
i'm telling you guys, it's the next big thing. space datacenters, but on earth. you heard it here first. -
R 'Nearest' Nabors
@rachelnabors
What about for engineers who can't pattern match yet?
-
R 'Nearest' Nabors
@rachelnabors
It's not a beret! This hat from Argentina does a great job of keeping rain off of my face.
-
R 'Nearest' Nabors reposted
Franziska Hinkelmann, PhD
@fhinkel
There are two types of developers emerging:
1) Those using AI to work faster.
2) Those using AI to build systems that work for them.
The first group is vibe coding their way to obsolescence. The second is loop engineering the future. The choice is yours.
If you are just a faster version of your 2023 self, you are still a manual operator.
True leverage comes from building autonomous loops that scale without your intervention. -
R 'Nearest' Nabors
@rachelnabors
“AI has experienced nothing” so it speaks in abstractions and broad, overarching terms. This is The Smell I associate with fan writing by young authors who haven't experienced much of the world yet. For instance, they abstract away emotions and feelings they haven't had yet, like grief or love are glossed over rather than acted out. This happens even in Big Name productions like some Disney films (“Mother Hotel died and Rapunzel was reunited with her birth family” Vs “even though she was a bad mother, she was the only mother I had known. I felt a turmoil of conflicting emotions. I wanted to cry, but I also felt relieved. And I felt shame for that relief, but also indignant to feel shame about someone who had lied to me my whole life. And guilt. Guilt that I was turning away from my “mother” to embrace essentially strangers.”
AI has no internal state or emotions. I can at best write fan fiction of reality. Not to criticise fan fiction writers—they can become excellent writers with practice and study and channeling their ever increasing body of lives experiences (“just live a real life” has been the best advice I've given to other creators). That is something beyond AI’s ability.
open.substack.com/pub/thealgorithmicbridge/p/10-signs-of-… -
R 'Nearest' Nabors
@rachelnabors
I really like how Dandadan portrays Teen Love. I feel it's overly romanticised and sexualized in both Japanese and American media. But there is something wonderfully tender about things like that hand holding scene in the car that are uniquely, innocently, a modern teen experience. More like this
-
R 'Nearest' Nabors
@rachelnabors
I tried to make @nearestnabors happen. But the algo didn't like it. So here we are.
I've just accepted that X is a tuning algorithm for reaching an audience. It's not a community.
I have messaging apps for that.
And Mima. -
R 'Nearest' Nabors reposted
Aurelien
@Aurelien_Gz
my brain can't process what i just witnessed on this site..
i have so many questions
» ac-bu.info/kindolphinThis media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
New hair!
Now I just need a new pair of glasses. Are there any glasses so sharp that just have great looks and great audio? (Don't need Meta's creepy video spyware. This pair is old and broken!) -
R 'Nearest' Nabors
@rachelnabors
Have you ever thought about getting a dog?
-
R 'Nearest' Nabors reposted
Luca Repetto
@luca_repetto_c
What are the effects of large human-capital shocks on innovation?
In a new paper, we study how WWI military deaths across British communities affected local invention over the next decades.
We find that places that lost more young men became persistently less innovative.
(🧵1/11) -
R 'Nearest' Nabors
@rachelnabors
"In just a few prompts" is the new "in just a few clicks"
-
R 'Nearest' Nabors
@rachelnabors
Why would I tolerate vendor lockin in this era?
-
R 'Nearest' Nabors
@rachelnabors
And thus concludes Cottage Core Birthday Hack Week 2026 ⛰️🎂💻
R 'Nearest' Nabors
@rachelnabors
I started studying chess on this trip with Duolingo (after boring a colleague on a train with my bad form). Went from 350 ELO to 450. Getting there!
-
R 'Nearest' Nabors reposted
Charles Maddock
@charles_maddock
I'm visiting the SF next week to speak at Apple HQ and NVIDIA GTC. 🇸🇪🤝🇺🇸
Would love to meet SF startups/scaleups and demo Strawberry in person!
DM me, I'd love to come by your office :) -
R 'Nearest' Nabors reposted
Swizec Teller
@Swizec
I was going for the 3 Bricklayers story
folks who think of their work as laying bricks are getting displaced by AI
folks who build cathedrals just got a way to get there even faster
a lot of this depends on incentive structures, you have to work in an environment that values outcomes over effort -
R 'Nearest' Nabors reposted
Sarah Drinkwater 🔮
@sarahdrinkwater
Some of the best agent builders + users in my life are, by CV, doctors, artists, designers. It's cool!
If that's you, check out @dreamer's builder in residence program - a brilliant chance to get paid to imagine & build ❣️
David Singleton
@dps
The best agents aren’t coming from labs. They're coming from builders like you.
That's why we're launching our Builder in Residence program: a paid fellowship to build agents in Dreamer.
No pressure to monetize, no “side-hustle” stress…. just you, your ideas, and our platform (and team!) to help you bring them to life.
Learn more: dreamer.com/residency -
R 'Nearest' Nabors
@rachelnabors
Every so often, a founder will ask me to "do for us what you did for react.dev." The problem is:
1) I consider react.dev incomplete beyond its References, which have been maintained. It's Learn section stops before server components. Thus, while it's very good, it's incomplete, and thus not the best anymore.
2) I would build a dev portal differently today from then. The focus was on getting humans to understand a mental model and practice on components. Today we would have to build for agents, who do not need that level of hand holding or interactivity and do need other things
3) I'd need a Dan Abramov, an engineer who deeply understands the technology AND is able to explain it to others with patience. Those are hard to find. Without them, it's hard to convert that deep context into something shareable, relayable
This is what I've learned from building multiple Dev portals since then, with varying degrees of success. But what worked for react.dev cannot be reproduced with the same effect elsewhere -
R 'Nearest' Nabors reposted
Peer Richelsen
@peer_rich
at this point we are just larping productivity
0xMarioNawfal
@RoundtableSpace
OPENCLAW AGENTS NOW JOIN SCRUM MEETINGS AND REPORT THEIR PROGRESS IN REAL TIME.
STANDUPS WITH YOUR AI ENGINEERS.This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
I've noticed two kinds of reactions to AI + coding. Giddiness over waking up in the morning to see what progress my loops make on my carefully laid specs so I can debug and finesse over a hot cup of Assam tea or depression over losing the giddiness of waking up to solve programming challenges.
My hope is that people are able to transfer their giddiness and keep solving new hard problems.
I am convinced it takes a team of T-shaped engineers to produce quality codebases with these techniques. So I am less worried about there being work for engineers, more worried that today's engineers will not migrate to the new problemset. There are still many CSS, Perl, COBOL etc specialists out there who didn't learn "the hot new thing," though, so I suppose it'll be fine one way or another
Adam
@adamdotdev
Man, I felt this video so hard. It feels like all of us (devs) that are using AI are trying to tiptoe around these feelings, or hang onto "but I use it correctly! i review the code!", idk, I think the drug analogy is apt. Once you have the button, it feels impossible to not use the button.
On the one hand, there is something so tempting about throwing it all away and going back to life as it was before all of this. Even a little part of me that hopes it's all unsustainable and crumbles down around me so I don't have to make the impossible decision. On the other hand, I've spent most of my career unable to sleep at night because I couldn't wait to wake up and implement the solution that came to me before bed. I don't have that problem anymore, I sleep like a baby, and I'm only now realizing it's because there's no mystery anymore, no burning desire to wake up and shape the world. I just put the prompts in, and that doesn't get me out of bed. Rock and a hard place for sure.
Anyway, use OpenCode, our new subscription (Go) is the best way to buy your drugs I mean tokens! 🥲 -
R 'Nearest' Nabors
@rachelnabors
Birthday Hack Week: Cottage Core Edition is coming to a close. Will share what I built next week! Please don't laugh 😂
-
R 'Nearest' Nabors
@rachelnabors
Just as MCP seems to be everywhere, I'm sneaking out the back door. I really only use it for things that need auth or UI. Skills and codemode do most everything and agents handle them more naturally
-
R 'Nearest' Nabors reposted
Arcade.dev | The Actions Runtime
@TryArcade
Every web developer's agent workflow: copy console error, paste into agent, wait, repeat.
Chrome DevTools MCP fixes this: ~25 tools that give your coding agent direct access to the browser.
@nearestnabors talked to @MHablich about how they built it.
youtu.be/CIOcjjgl9MI -
R 'Nearest' Nabors
@rachelnabors
Ralph loop was still running this morning, so I've gone for a hike and a pastry.
Don't call me.
R 'Nearest' Nabors
@rachelnabors
-
R 'Nearest' Nabors reposted
Alex Booker
@bookercodes
Today we're launching @mastra remote sandboxes
Give your agent a secure, isolated environment to run untrusted user code... like a vibe coding agent...This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors
@rachelnabors
Can't go to bed until I've planned out 3 new features/adjustments for Mima and set up the new and improved Ralph Loop :3 github.com/nearestnabors/ralph-wiggum-loop-starter
-
R 'Nearest' Nabors
@rachelnabors
Would be amazing if coding agents facilitated major platform shifts overnight. Usually we bog down in one platform and shifting requires a concerted cross-team effort—so Booking stays on Perl, Meta on PHP/React, etc.
May the best framework win
Charles Maddock
@charles_maddock
We rewrote our 130k line React UI to Svelte in two weeks with coding agents.
It was mostly done by one (1) founding engineer in his spare time. -
R 'Nearest' Nabors
@rachelnabors
For my birthday week, I didn't want a fancy trip or hotel. I just wanted to be alone with my computer someplace that wasn't the apartment.
Cottage core birthday hack week has kicked off -
R 'Nearest' Nabors reposted
Franziska Hinkelmann, PhD
@fhinkel
Every industry used to have clear hierarchies. Experts with decades of experience, juniors learning from them.
This is different for AI.
Your 20 years of experience don't transfer when the tools change this fast. We're all juniors now, whether you admit it or not. So how do you avoid getting left behind?
You don't wait for a list of approved resources. You build in public, ship fast, and learn from production failures. -
R 'Nearest' Nabors
@rachelnabors
I'm in this post
Ellipsus
@ellipsus_writes
You don’t write like AI—AI writes like you. Defend the em dash.
-
R 'Nearest' Nabors
@rachelnabors
Loving the representation on this site. Proper acknowledgement of the work women and LGBTQ have made to this field. Thank you.
Sovereign AI
@UKSovereignAI
Introducing Sovereign AI, the Government’s new £500m venture fund.
Sovereign AI will support founders from day one to start here, scale here and win everywhere.
sovereignai.gov.uk -
R 'Nearest' Nabors reposted
Simon Willison
@simonw
A short note that the predictions that LLMs would favor "boring technology" that's over-represented in the training data don't appear to be playing out as expected with the latest models - once you attach them to a good coding agent harness at least simonwillison.net/2026/Mar/9/not-so-boring/
-
R 'Nearest' Nabors
@rachelnabors
Feeling very loved and cherished this birthday 🎂
R 'Nearest' Nabors
@rachelnabors
Quite the haul
-
R 'Nearest' Nabors
@rachelnabors
This.
A good place to start with loop engineering is this Ralph Wiggum starter I made: github.com/nearestnabors/ralph-wiggum-loop-starter
Franziska Hinkelmann, PhD
@fhinkel
Prompting is a bug, not a feature.
Stop obsessing over verbs and context windows. If your workflow depends on you being a prompt whisperer, you’ve already lost. You’re just a manual operator in a world that demands systems.
Vibe Coding is fun for demos.
The real evolution is Loop Engineering. You don't write a prompt; you build a recursive environment where the agent evaluates its own failure, refactors its logic, and iterates until the delta between intent and output is zero.
The human shouldn't be the editor. The human should be the architect of the loop.
Stop talking to the machine. Start building the machine that talks to itself. -
R 'Nearest' Nabors
@rachelnabors
Birthday week will continue until morale improves
R 'Nearest' Nabors
@rachelnabors
(that rhinestone encrusted peacock was actually in a completely random hotel lobby, but it was too fabulous not to go in and photograph)
-
R 'Nearest' Nabors
@rachelnabors
Claude Code trying to co-author GitHub repos. Yeah. No buddy, no free advertising for billion dollar companies I pay a subscription fee to. I love you, but boundaries.
-
R 'Nearest' Nabors reposted
Gergely Orosz
@GergelyOrosz
I hate this site
The below blue checked account makes up completely fake stuff - that never happened - and it gets 2M+ views. No it didn't happen
People treat as fact, thanks to having a $8/month blue check
Fake news (often written by AI) spreads on this site like wildfire -
R 'Nearest' Nabors
@rachelnabors
Maybe because no one wants to set up shop in a country where the administration might turn on them. Also, lack of healthcare, gun violence, unhealthy car culture.
But I think it's mostly the nervous energy. Feels like everyone woke up and remembered London is a place they can do business. (Personally, I think for businesses , having a UK contingent is a hedge, a back door in case of nuclear wars/fascism/dictatorial takeover)
Stefan Schubert
@StefanFSchubert
There are now more tech workers going from the US to Europe than the other way around
-
R 'Nearest' Nabors
@rachelnabors
Sometimes I have really grim thoughts about the developer community and there is no good place to post them, so I just email them to myself
-
R 'Nearest' Nabors
@rachelnabors
If you are a developer and you are averse to using AI tools, you are where CSS die-hards were when they rejected JS frameworks. It will get lonely out there, and harder to find work
R 'Nearest' Nabors
@rachelnabors
And I don't mean this in an aggro way. I mean this a person who watched their favourite community members fade into irrelevance rather than adapt to a changing industry.
We will build better things together. Please come along. I can't wait to see what you discover and build -
R 'Nearest' Nabors reposted
Gergely Orosz
@GergelyOrosz
9 interesting observations from my conversation with Mitchell Hashimoto (@mitchellh, creator of Ghostty, founder of HashiCorp):
1. Vagrant was created because dev environment setup was an unbillable time sink at a consultancy. At the Ruby on Rails shop where Mitchell worked, jumping onto another client’s project could waste half a day. This inspired building Vargant.
2. Terraform won, despite being 7th to market. Terraform won through relentless conference presence, community building, and a better developer experience — not timing.
3. HashiCorp had no real business for four years and their first commercial product was a full-on failure. The initial product, Atlas, required customers to adopt the entire HashiCorp stack. It was a hard sell. HashiCorp pivoted to selling individual services like Vault, and this approach proved to be a winner.
4. VMware almost bought HashiCorp for ~$100M, and Terraform would have not happened if it did. VMWare took took the offer to their board, where they rejected to buy with a single vote. Mitchell said that Terraform probably never would’ve existed if the VMWare purchase went through.
5. Mitchell’s new rule for building software: always have an agent running in the background doing something. He kicks off tasks before leaving the house — research, edge-case analysis, library comparisons — so work progresses while he drives or is away.
6. Open source is moving from “default trust” to “default deny” — and Mitchell thinks that’s how it should be. This is because AI makes it trivial to create plausible-looking but incorrect and low-quality contributions. As he put it: “open source has always been a system of trust. Before, we’ve had default trust. Now it’s just default deny.”
7. Git and GitHub may not survive the agentic era in their current form. Agents cause so much churn that merge queues become untenable, branches proliferate, and repos balloon. Mitchell compares the needed shift to Gmail’s revolution for email: “We’re at the Gmail moment for version control... never delete, archive everything.”
8. The best engineers Mitchell ever hired had boring, invisible backgrounds. No GitHub contributions, no public profiles, companies you’ve never heard of. “Every moment you spend on social media is taking away from something else... the best engineers are the ones that context-switch the least.”
9. Mitchell’s advice for AI-skeptical engineers: start by reproducing your research, not your code. As he puts it:
“There’s a lot of people like, ‘I don’t want it to write code for me.’ But just delegate some of the research part.”
He uses agents for library comparisons, edge-case analysis, and deep research — not just code generation. Mitchell: “You don’t need to pick up on the ‘it must replace you as a person’ kind of propaganda.”
Watch the full episode here: youtu.be/WjckELpzLOU
Other platforms and transcript: newsletter.pragmaticengineer.com/p/mitchell-hashimoto -
R 'Nearest' Nabors
@rachelnabors
Birthday week continues: setting up OpenClaw
-
R 'Nearest' Nabors
@rachelnabors
Birthday week is on, bitches
-
R 'Nearest' Nabors
@rachelnabors
There is a breed of builder right now working on the things that were never "worth investing in." Wonders are coming.. wonders.
-
R 'Nearest' Nabors
@rachelnabors
Mima feature list: screen generated content from social feeds.
-
R 'Nearest' Nabors
@rachelnabors
Addressed the first issue on my latest repo :D github.com/nearestnabors/ralph-wiggum-loop-starter/issues/1
Appreciate folks sharing their expertise to help build something greater (and I learn new things, too!) -
R 'Nearest' Nabors reposted
Gergely Orosz
@GergelyOrosz
Anyone and everyone working in security engineering or caring about security have their work cut out for them
We’re so early in AI agents pushing code to prod without human intervention - but prompt injections are already spreading like wildfire. Infecting high-profile projects
Sash Zats
@zats
> The attacker got the npm token by injecting a prompt into a GitHub issue title, which an AI triage bot read, interpreted as an instruction, and executed.
-
R 'Nearest' Nabors
@rachelnabors
Riffing on this, the most important thing you can do right now is stop worrying about getting the cool job at the cool place with the cool kids. Stop that.
The most important thing you can do is use the tools to make your wildest dream come true. Imagine someone handed you a magic box that could build anything you'd ever dreamed of. What would it be? An 8ball that predicts the stock market? A social network that bankrupts Meta? An app that lets your neighbourhood arrange carpooling for the kids?
You have that, right now. You don't need anyone's permission. You don't need to impress anyone or get anyone to fund you or believe in you. You can just do the thing.
It is this act of doing, and applying what you know to the unknown, and discovering what works best that will open doors and elevate you.
Make and share and don't look back
R 'Nearest' Nabors
@rachelnabors
Today, a developer friend asked, "Aren't you nervous about the job market (with the advent coding agents)?" I said no, I'm not. I haven't had a traditional job since the Amazon layoffs. I've just been working with AI startups as a DevX consultant, managing launches, dev portals, etc.
Finding a principal TPM/DX role at an established tech company is HARD when you have a lot of goodwill but no leadership chain to follow. So I'm not scared, because for a lot of us, it's BEEN chaos. Finding where we can contribute has taken work.
But honestly, I've never been happier. For the past quarter I've been making things I could never have made before, working in communities building standards for the future again, and the best part: no review cycles where I have to do a corporate fan dance to convince a spreadsheet not to delete my row!
I feel alive in a way I have not since the start of my career, when I was building websites with just CSS, HTML, and JS. There is so much you can do right now, and we need people figuring out how to do it well in this new modality.
I am so happy to be here right now. Thank you. -
R 'Nearest' Nabors
@rachelnabors
Today, a developer friend asked, "Aren't you nervous about the job market (with the advent coding agents)?" I said no, I'm not. I haven't had a traditional job since the Amazon layoffs. I've just been working with AI startups as a DevX consultant, managing launches, dev portals, etc.
Finding a principal TPM/DX role at an established tech company is HARD when you have a lot of goodwill but no leadership chain to follow. So I'm not scared, because for a lot of us, it's BEEN chaos. Finding where we can contribute has taken work.
But honestly, I've never been happier. For the past quarter I've been making things I could never have made before, working in communities building standards for the future again, and the best part: no review cycles where I have to do a corporate fan dance to convince a spreadsheet not to delete my row!
I feel alive in a way I have not since the start of my career, when I was building websites with just CSS, HTML, and JS. There is so much you can do right now, and we need people figuring out how to do it well in this new modality.
I am so happy to be here right now. Thank you. -
R 'Nearest' Nabors
@rachelnabors
A former colleague called me "an AI expert" today, which felt off. A year ago, I was definitely not that. I am still learning from the ML engineers I've surrounded myself with. I think what has changed is that enough people care now that I'm "the oldest kid in class"
-
R 'Nearest' Nabors
@rachelnabors
Earlier this week, I interviewed Sam Morrow, one of the engineers behind GitHub MCP. I wrote a blog post about the lessons he learned and who they apply to enterprise MCP server development. Check it out! www.arcade.dev/blog/enterprise-mcp-lessons-from-githubs-m…
-
R 'Nearest' Nabors
@rachelnabors
Went to Vercel's Agent Platform launch events here in London. Some thoughts on how the agentic web is shaping up:
agenticweb.nearestnabors.com/p/are-we-evolving-from-pages… -
R 'Nearest' Nabors
@rachelnabors
Morning, nerds. I was trapped in a dopamine loop working on my side projects until 3am again. How's your Friday?
-
R 'Nearest' Nabors
@rachelnabors
I got sick of using multiple social apps. So for Hack Week I've been making my own meta social app that tracks conversations and news from people I am connected with across different platforms! It uses a local model for summaries and some very carefully crafted, in-progress algorithms, and a lot of vibecoded auth flows.
Most of the "social apps" I have seen people building focus on sharing: Post to every social network! Generate posts across 20 Reddits! Auto respond to requests for help!
That's a broadcast app, and while such a tool might be very helpful to a marketing or support team, it isn't helpful for collecting news and important opinions, engaging in conversations where your expertise is valued, or keeping in touch with people who matter most.
Those are the behaviors that got me this far (well, those and oversharing). So I'm working on building an app that does that.
It's not an agent. I don't want to delegate this work. I want a HUD for it. And I'm using algorithms (including LLMs) to enhance my experience, not replace me. -
R 'Nearest' Nabors
@rachelnabors
Zero rest for the wicked! I'll be speaking about boosting Claude Code performance with prompt learning on March 17. Can't wait to share! luma.com/gwd1hbzo?tk=DN3Gtj
-
R 'Nearest' Nabors
@rachelnabors
Coworking in Camden 🍵
What are you building today? -
R 'Nearest' Nabors
@rachelnabors
Shout out to @BurnedChris, who has become one of my favourite people to have come by for boardgames in London for his shrewd entrepreneur mindset and spirited sense of humour. One to watch!
Christopher Burns
@BurnedChris
Hot take: If your startup is in London, stop calling it “SF-based”.
Hard to build the next Silicon Valley if everyone keeps pretending they’re already there. -
R 'Nearest' Nabors reposted
Christopher Burns
@BurnedChris
Hot take: If your startup is in London, stop calling it “SF-based”.
Hard to build the next Silicon Valley if everyone keeps pretending they’re already there. -
R 'Nearest' Nabors
@rachelnabors
Love this for the UK but sad it's getting less funding that some LLM wrappers in the valley: share.google/BRpZXR4snvSzfx7vh
-
R 'Nearest' Nabors
@rachelnabors
Waking up to a lot of, "but engineers don't want to be managers!"
Beginning to think engineers think managing is the same thing as mentoring! -
R 'Nearest' Nabors
@rachelnabors
I've heard a lot of "juniors are skipping learning hard lessons! Seniors are underwater with AI-generated slop!" But I'm not sure that's what's happening.
My GitHub squares turned emerald green around November. That was AI. Much of it was not carefully crafted. But I've had the benefit of working with some amazing engineers. And I have learned so much, so quickly, from their feedback on PRs I wouldn't have been able to submit in a timely fashion before. I've learned about cache poisoning, for instance. I have built things I never would have known how to because I could sit down and talk with people who've delivered them before about plans I'd researched.
I think "slop code" has underscored a problem tech has long had, but juniors have been able to brute force past: there are more juniors seeking mentorship than there are seniors willing or able to guide them.
I'd like to see the role of senior engineers shift to a sort of master craftsman overseeing a studio of apprentices carrying out their work, assigning rough drafts, etc. -
R 'Nearest' Nabors reposted
React Native Live
@reactnativelve
TanStack with Jack Herrington | React Native Live | Episode 9 x.com/i/broadcasts/1qxoNeYjlwvJv
-
R 'Nearest' Nabors
@rachelnabors
And here I was thinking about upgrading since my previous pair finally bit the dust. I don't even want the videos. I just really enjoy the audio. But this is so creepy, and I don't trust it not to be used by America's Department of defence. Zuck would roll right over
Aakash Gupta
@aakashgupta
Everyone’s missing the real story here.
Meta’s Ray-Ban glasses need human data annotators to train the AI. When you say “Hey Meta” and ask the glasses to analyze something, that video gets sent to Meta’s servers, then routed to Sama, a subcontractor in Nairobi, Kenya. Workers there manually label objects in your footage. They see everything you recorded, intentionally or not.
7 million pairs sold in 2025 alone. Every single pair generates training data that flows through human eyes in Kenya. Workers told Swedish journalists they see people undressing, using bathrooms, having sex, and accidentally filming bank card details. One worker said “we see everything, from living rooms to naked bodies.”
Meta’s automatic face anonymization is supposed to protect people in the footage. Workers say it fails in certain lighting. Faces that should be blurred are sometimes fully visible. The person you recorded without knowing? A stranger in Nairobi can identify them.
Buried in Meta’s terms of service is one sentence doing enormous legal work: the company reserves the right to conduct “manual (human) review” of your AI interactions. That’s the legal cover for routing intimate footage from Western homes to a $2/hour labor force operating under NDAs, office surveillance cameras, and a strict no-questions policy. Workers say if you raise concerns about what you’re seeing, you’re fired.
This is the same company, Sama, that TIME exposed in 2023 for paying Kenyan workers $2/hour to label graphic content for OpenAI while being billed at $12.50/hour per worker. Workers described the experience as torture. Sama ended that contract, then pivoted to labeling Meta’s glasses footage. Same workforce. Same rates.
Meta markets these glasses as “designed with your privacy in mind.” The privacy design is a tiny LED light on the frame that most people don’t notice. The data pipeline behind it routes your bedroom footage to a contractor with a documented history of worker exploitation, failed anonymization, and union-busting lawsuits.
And the next generation of these glasses? Meta is planning to add facial recognition. The same system that can’t reliably blur faces in training data wants to start identifying them on purpose.
The LED light on the frame is doing about as much for your privacy as the terms of service nobody reads. -
R 'Nearest' Nabors
@rachelnabors
Wrote up a tutorial for how I run a Ralph Wiggum loop (and you can, too!) github.com/nearestnabors/ralph-wiggum-loop-starter
Do not be intimidated! The bulk of it is having long, soulful conversations with Claude Code about what success looks like
R 'Nearest' Nabors
@rachelnabors
This is what it sounds like
This media couldn't be displayed. See it on the original post. -
R 'Nearest' Nabors reposted
Christopher Burns
@BurnedChris
I can't believe we managed to hit 1 million downloads in our first year with @c15tdev.
We feel like we're just getting started. We can't wait to talk more about c15t 2.0 very soon! -
R 'Nearest' Nabors
@rachelnabors
GitHub + MCP might have been AirBnB + ReactNative headline of MCP servers. I interviewed one of the engineers behind GitHub's infamous MCP server to find out what the team learned from launch, and how they've changed their approach: www.youtube.com/watch?v=wSQAR3NQKuU
-
R 'Nearest' Nabors
@rachelnabors
Hi from Mima!
-
R 'Nearest' Nabors
@rachelnabors
Regular reminder that expert opinions tend to be negative and that both women and men penalise women for expressing negative opinions.
If you're wondering where all the women experts are at, likely they are crushed beneath your own positivity bias -
R 'Nearest' Nabors
@rachelnabors
Good design is the difference between apps designers vibe code and "one-size fits none" apps Big Tech pumps put.
Bring on the new blood
David K 🎹
@DavidKPiano
Good design is the difference between your cute vibe-coded app and apps that people actually want to use
-
R 'Nearest' Nabors
@rachelnabors
I wish it had been called "YOLOing instead of "Ralphing" (since it sounds like throwing up).
"You only loop over (and over and over)" -
R 'Nearest' Nabors
@rachelnabors
The REAL #config2024 design-gineering after-party: a trip to @jessicahische's store in Oakland with @NikkitaFTW @marisamorby @jlengstorf @devdevcharlie plus special guests @jina @rod_ @brittanyforks and UBE HORCHATA!!! #onlyinoakland
-
R 'Nearest' Nabors
@rachelnabors
!! @reactjs ALSO WON BEST OVERALL DEVELOPER PORTAL at @DevportalAwards!!! Oh man, I feel like there needs to be a thread of congratulations to the contributors (which you can find in full any time at react.dev/community/docs-contributors).
Hold my beer!
R 'Nearest' Nabors
@rachelnabors
!! @reactjs docs won best API Reference documentation at @DevportalAwards: react.dev/reference/ Shout out to @dan_abramov who drove this part of the docs to completion! Congratulations!
-
R 'Nearest' Nabors
@rachelnabors
!! @reactjs docs won best API Reference documentation at @DevportalAwards: react.dev/reference/ Shout out to @dan_abramov who drove this part of the docs to completion! Congratulations!
-
R 'Nearest' Nabors
@rachelnabors
@reactjs just won a Dev Portal Award for DX! Go team! devportalawards.org/winners?year=2023
-
R 'Nearest' Nabors
@rachelnabors
A present for your holidays: @reactnative's site has a new Architecture section populated with content about how the new Fabric renderer works written and edited by @SamuelSusla and @lunaleaps! Enjoy! reactnative.dev/docs/fabric-renderer
-
R 'Nearest' Nabors
@rachelnabors
Catch my keynote about the React docs happening now! conf.reactjs.org/stage
-
Post not available here, view on XPost not available here, view on X
R 'Nearest' Nabors
@rachelnabors
I <3 @rescriptlang's "syntax lookup" page! What a glossary + examples! rescript-lang.org/syntax-lookup
-
R 'Nearest' Nabors
@rachelnabors
Progress update on the new React docs! github.com/reactjs/reactjs.org/issues/3308
-
R 'Nearest' Nabors
@rachelnabors
React Hooks are a freaking beast to turn into reference pages. They don't boil down neatly into types. They are shapes of functions and bundles of dependencies. I have to rethink how I format API docs because of this
-
R 'Nearest' Nabors
@rachelnabors
Do you like Flexbox? Do you want to contribute to some React Native documentation? Have I got a great first issue for you: github.com/facebook/react-native-website/issues/2163
-
R 'Nearest' Nabors
@rachelnabors
Hello! I'm looking to chat with some folks learning React! If you've some time to talk with me about your experiences for half an hour, please let me know: www.surveymonkey.co.uk/r/react-learner-interview
-
R 'Nearest' Nabors
@rachelnabors
The #reactnative documentation drive is so close! We are down to a handful of APIs to update. Great place to start contributing to opensource if you <3 @ReactNative! github.com/facebook/react-native-website/issues/1579
-
R 'Nearest' Nabors
@rachelnabors
Want to get a final PR or two in to @ReactNative before 2020? We still have components and API docs that need updating! github.com/facebook/react-native-website/issues/929
-
R 'Nearest' Nabors
@rachelnabors
React Native friends: what APIs do you have to dig around in our source code to figure out?
-
R 'Nearest' Nabors
@rachelnabors
Folks! Come help us lint out words like "just" and "easy" from React Native's older version docs! Not a bad first PR project ;) github.com/facebook/react-native-website/issues/1338
-
R 'Nearest' Nabors
@rachelnabors
In happy news: soooo excited to have @kadikraman writing @reactnative's upcoming security docs! 🤩
-
R 'Nearest' Nabors
@rachelnabors
Are you learning, teaching or have recently learned React Native? Can I pick your brain for half an hour? Working on the React Native docs and want to know more about how you use them—or don't use them!


































































































































































































































































































































































































































































































































































































































































































