Parallel🤝@vercel
Parallel Decisions API, a.k.a. WebJev, is free for the next 10 days on @vercel AI Gateway.
Happy building!
Politicians, athletes, journalists, NGOs, companies... They all use corporation-owned social networks to communicate with us. Citizens shouldn't have to hand over their data to companies to read content published by public entities. Public Feeds changes that by making posts of popular accounts freely available. Read more.
Parag Agrawal (Hindi pronunciation: [pəˈɾaːɡ əɡɾəˈʋaːl]; born 21 May 1984) is an Indian software engineer and businessman who was the CEO of Twitter, Inc. from November 2021 to October 2022. More on Wikipedia
Parallel🤝@vercel
Parallel Decisions API, a.k.a. WebJev, is free for the next 10 days on @vercel AI Gateway.
Happy building!
Consistently the highest quality, and on the quality-cost Pareto frontier
Parallel Search now leads the @ValsAI Web Search Index, delivering the best mix of quality score, cost, and speed for legal and finance research tasks.
Check out the full report: www.vals.ai/benchmarks/web_search
Parag Agrawal reposted
Gemini 4 Hydrogen is now widely available on
@googlecloud. We’re thrilled to be day-one launch partners.
Use Parallel to ground Gemini 4 Carbon natively and bring accurate, fast, and agent-ready web data to your production AI.
Get started: docs.cloud.google.com/gemini-enterprise-agent-platform/mo…
Parallel on AWS - big unlock for so many of our mutual customers.
Parallel Search is now available in the AWS Marketplace with committed spend drawdown eligibility.
Teams building on AWS can add high-quality, low-cost web search to agents running any Amazon Bedrock model and easily configure it in Amazon Bedrock AgentCore.
Thanks to @ZetaGlobal for joining us for this launch. Read more and check out our listing in the comments.
Parag Agrawal reposted
Fable 5.2 (@AnthropicAI) has reclaimed the #1 spot on Parallel's Search Capability Leaderboard, followed by Opus 5.5 and Hy4 (@TencentHunyuan), the highest-ranking open model we've tested.
When this leaderboard launched, Fable 5.1 led with a score of 73.4, which models like Sonnet 5.5 now beat through intelligent use of web search.
Parag Agrawal reposted
Announcing Parallel for Startups:
Free credits for teams at the frontier of the agentic web.
Available today to founders and teams backed by @ycombinator, @speedrun, @neo, @joinef, @fdotinc, or @nullfellows.
Not in a partner program? Apply at parallel.ai/startups
Parag Agrawal reposted
there's a lil' sidekick waiting for you on the Parallel for Startups webpage. his name's Clicky ꒰১🖱️໒꒱
Announcing Parallel for Startups:
Free credits for teams at the frontier of the agentic web.
Available today to founders and teams backed by @ycombinator, @speedrun, @neo, @joinef, @fdotinc, or @nullfellows.
Not in a partner program? Apply at parallel.ai/startups
Parag Agrawal reposted
come check out my talk about webjev at @aiDotEngineer !
We've got a big week ahead at @aiDotEngineer Code Summit in SF:
• Tonight @ 5:30 pm: the welcome reception, co-hosted with our friends at @AnthropicAI
• Tomorrow @ 6:30 pm: happy hour with @factory (link in comments to request a spot)
• Thursday, 2:20 pm: @devingupta_ presents "The first Grounded Decision Model" on WebJev, our new Grounded Decisions API that answers questions using evidence from the live web
• All week: stop by our booth to say hello
Parag Agrawal reposted
Parag Agrawal reposted
parallel is hiring across every team
- research + mts (search ranking, model training, compute scaling)
- deployed engineers
- design (brand, product, design eng)
- gtm
- marketing
- finance, legal + ops
sf, palo alto, nyc, london. if you want to build the web for its second user, come join us / tag them below
Agents doing legal work have really scaled, and some of the best ones use Parallel to for their web search. The reason is simple: @p0 is most accurate, cost-effective solution.
Announcing Parallel for Legal: upgrades to our index and ranking that put Parallel on the cost-accuracy frontier for legal AI search.
- On our private PWS-Legal benchmark, Parallel leads in accuracy and holds the entire Pareto frontier, beating every other provider at up to 2.3x lower cost.
- On Legal Hallucinations, Parallel delivers frontier accuracy at up to 2.6x lower cost to comparable providers.
- On LegalBench Citation, Parallel beats every other agentic search API on accuracy, at the same or lower cost.
Legal AI is only as good as the sources behind it. The evidence lives in court dockets, state codes, agency guidance, and international tribunals, and much of it sits on sites that general web search reaches poorly.
Just ask @Harvey, @GCAI, and @finchlegal, who already run Parallel in production.
Parag Agrawal reposted
Just in time for the weekend, happy building
What's better than the highest accuracy web-grounded decisions model? How about a week of free requests?
Try Parallel Decisions to get fast structured answers about the live web with probabilities and citations, free for 7 days exclusively for OpenRouter users!
Parallel Decisions is now on OpenRouter. Parallel Decisions delivers the highest accuracy on web-grounded decisions for System One models like Jev. Get web-grounded answers 13× faster at 1/16 the cost of OpenAI’s GPT Luna with native web search.
Ask questions about the live web and get structured answers with probabilities and citations in under a second.
Parag Agrawal reposted
sub-second grounded decisions over the live web,
completely free
with an openrouter api key
right now!
go try it out!
What's better than the highest accuracy web-grounded decisions model? How about a week of free requests?
Try Parallel Decisions to get fast structured answers about the live web with probabilities and citations, free for 7 days exclusively for OpenRouter users!
Parallel Decisions is now on OpenRouter. Parallel Decisions delivers the highest accuracy on web-grounded decisions for System One models like Jev. Get web-grounded answers 13× faster at 1/16 the cost of OpenAI’s GPT Luna with native web search.
Ask questions about the live web and get structured answers with probabilities and citations in under a second.
Parag Agrawal reposted
What's better than the highest accuracy web-grounded decisions model? How about a week of free requests?
Try Parallel Decisions to get fast structured answers about the live web with probabilities and citations, free for 7 days exclusively for OpenRouter users!
Parallel Decisions is now on OpenRouter. Parallel Decisions delivers the highest accuracy on web-grounded decisions for System One models like Jev. Get web-grounded answers 13× faster at 1/16 the cost of OpenAI’s GPT Luna with native web search.
Ask questions about the live web and get structured answers with probabilities and citations in under a second.
Parag Agrawal reposted
Parallel Decisions is now on @OpenRouter
Free for the next 7 days.
It delivers the highest accuracy for web-grounded decisions, with answers in under a second. Get web-grounded answers 13× faster at 1/16 the cost of OpenAI’s GPT Luna with native web search.
Select Parallel: Decisions Fast in the model list, or set your model ID to parallel/decisions-fast. We can’t wait to see what you build.
Learn more: parallel.ai/blog/decisions-openrouter
Parallel Decisions is now on OpenRouter. Parallel Decisions delivers the highest accuracy on web-grounded decisions for System One models like Jev. Get web-grounded answers 13× faster at 1/16 the cost of OpenAI’s GPT Luna with native web search.
Ask questions about the live web and get structured answers with probabilities and citations in under a second.
Parag Agrawal reposted
Parag Agrawal reposted
Parallel Decisions is now on OpenRouter. Parallel Decisions delivers the highest accuracy on web-grounded decisions for System One models like Jev. Get web-grounded answers 13× faster at 1/16 the cost of OpenAI’s GPT Luna with native web search.
Ask questions about the live web and get structured answers with probabilities and citations in under a second.
big leap on webjev quality, free, on open router
Parallel Decisions is now on OpenRouter. Parallel Decisions delivers the highest accuracy on web-grounded decisions for System One models like Jev. Get web-grounded answers 13× faster at 1/16 the cost of OpenAI’s GPT Luna with native web search.
Ask questions about the live web and get structured answers with probabilities and citations in under a second.
Upgrade your agent with a single prompt
Parallel MCP 2.0 is live.
It's the most powerful web search and retrieval MCP, launching today with four tools for agentic search:
- Web search
- Page extract
- Web-grounded decisions
- Deep research
Parallel MCP is free and keyless by default, with auth for advanced features and higher limits.
It's also the most configurable web data MCP, designed for admins to define web research standards across an entire fleet of agents.
Add it to any MCP client: mcp.parallel.ai/mcp
Learn more: parallel.ai/mcp
Parag Agrawal reposted
Parag Agrawal reposted
@p0 is now hiring many research/MTS roles in nyc! parallel.ai/careers this is a uniquely high talent density team and a win for the nyc tech community 🗽
Parag Agrawal reposted
Parallel's Monitor API is the perfect shape for the coming wave of proactive agents. Excited for this partnership with @browserbase
Personal Agents like BELL should always know what's happening, without you having to ask.
We partnered with @browserbase to add Parallel's Monitor API to BELL.
Monitor is like a web search that's always on. Now your personal assistant can keep track of the web and alert you when something changes 🔔
BELL agents are always watching.
We partnered with @p0 to let BELL monitor the web 24/7 for new information. Get competitive intelligence, market signals, or track products.
Fork our repo and build your own custom assistant that's actually personal.
Parag Agrawal reposted
Today, your agent asks the web for what it needs.
@paraga expects a lot of that to flip in three years, with systems like @p0's pushing information that wakes an agent up to start work.
Agents on CoreWeave can enable Parallel's search with a single setting.
⇢
Parag Agrawal reposted
Parallel is thrilled to deliver best-in-class web search for agents built in the @FireworksAI_HQ Frontier Ecosystem, announced today at Forge.
fireworks.ai/partners
The future of AI is specialized intelligence with models tuned to your data, your domain and your customers.
Today, we're proud to launch live from Forge the Fireworks Frontier Ecosystem with 60+ partners across the AI stack, helping you build and ship faster.
Check it out:
Parag Agrawal reposted
BELL agents are always watching.
We partnered with @p0 to let BELL monitor the web 24/7 for new information. Get competitive intelligence, market signals, or track products.
Fork our repo and build your own custom assistant that's actually personal.
Parag Agrawal reposted
‼️ high quality customer-oriented decisions benchmark!
grounding improves decision performance significantly for real customer traffic
Parallel Decisions beats Perplexity by 1.5x and OpenAI by nearly 2x on decisions that depend on the live web.
We’re open-sourcing the benchmark: PWS-Decisions, 300 questions whose answers change with the world. Run it on your own model.
Full results: parallel.ai/blog/pws-decisions-benchmark
Parag Agrawal reposted
Confused by what grounded decision model can be used for? Use our widget to find 20 use cases specific to your company in under 30 seconds.
webjev-use-cases.paralleldemos.com
open-source decisions api benchmark with questions whose answers change with the world
Parallel Decisions beats Perplexity by 1.5x and OpenAI by nearly 2x on decisions that depend on the live web.
We’re open-sourcing the benchmark: PWS-Decisions, 300 questions whose answers change with the world. Run it on your own model.
Full results: parallel.ai/blog/pws-decisions-benchmark
Monitor the web for things that change - now in slack
Introducing Parallel in Slack.
Tell Parallel what to monitor on the web, and the agent posts in the channel when there's a hit.
Or tag it with a question to get a cited answer based on the context in your thread.
Available today on the Slack Marketplace.
Parag Agrawal reposted
We've been quietly working on a Parallel Slack agent that does one thing incredibly well:
Always-on monitor the web for events and changes you care about.
It doesn't make animations. It doesn't negotiate your phone bill. But if you care about competitor launches, vendor policy changes, or relevant news on X, there is no better way to stay in the loop.
slack.com/marketplace/A0AP6UAHQ4A-parallel
Parag Agrawal reposted
You may not need an agent after all.
We rewrote five agents into decision workflows using @typesafeai's Jev and @p0's Decisions API. The 20-document run took about 20 seconds, cost about two cents and got 19 right.
A few takeaways:
• If the steps are known, use a workflow.
• Decision models are fast, cheap and easy to test.
• Static rules can be expressed in code.
• With Agentrun, you can get the best of two worlds: the agent can write the workflow and handle the cases the workflow can't decide.
Read:
Parag Agrawal reposted
Introducing Parallel in Slack.
Tell Parallel what to monitor on the web, and the agent posts in the channel when there's a hit.
Or tag it with a question to get a cited answer based on the context in your thread.
Available today on the Slack Marketplace.
Parag Agrawal reposted
Parallel for Finance is the new state of the art for financial agents.
On FinSearchComp T3 (chart below), Parallel Advanced is the most accurate search API at 1.3x-3x lower cost than every competitor, and Fast beats all of them at 13x+ lower cost. Top score on FrontierFinance and PWS-Finance too.
Live today across all @p0 APIs, no new plan.
Announcing Parallel for Finance: universal upgrades that make us the unmatched SOTA web search for Finance AI on the cost-accuracy frontier.
- On FinSearchComp's hardest tier (T3), Parallel is the most accurate search API at up to 2x lower cost than Exa and Perplexity. Parallel Fast comes second, at 10x lower cost than the others.
- On @samaya_AI's FrontierFinance, Parallel ties for the top score at 38% lower cost than Exa, and beats OpenAI, Perplexity, and Tavily on both accuracy and cost.
- On our new private PWS-Finance benchmark, Parallel leads, beating OpenAI's native search at 30% lower cost.
Accuracy, cost, latency. Irrespective of what you are optimizing for, you should be using our web search for your agents.
Announcing Parallel for Finance: universal upgrades that make us the unmatched SOTA web search for Finance AI on the cost-accuracy frontier.
- On FinSearchComp's hardest tier (T3), Parallel is the most accurate search API at up to 2x lower cost than Exa and Perplexity. Parallel Fast comes second, at 10x lower cost than the others.
- On @samaya_AI's FrontierFinance, Parallel ties for the top score at 38% lower cost than Exa, and beats OpenAI, Perplexity, and Tavily on both accuracy and cost.
- On our new private PWS-Finance benchmark, Parallel leads, beating OpenAI's native search at 30% lower cost.
Parag Agrawal reposted
Announcing Parallel for Finance: universal upgrades that make us the unmatched SOTA web search for Finance AI on the cost-accuracy frontier.
- On FinSearchComp's hardest tier (T3), Parallel is the most accurate search API at up to 2x lower cost than Exa and Perplexity. Parallel Fast comes second, at 10x lower cost than the others.
- On @samaya_AI's FrontierFinance, Parallel ties for the top score at 38% lower cost than Exa, and beats OpenAI, Perplexity, and Tavily on both accuracy and cost.
- On our new private PWS-Finance benchmark, Parallel leads, beating OpenAI's native search at 30% lower cost.
Parag Agrawal reposted
everytable.paralleldemos.com
i've seen $0.0024/lead and $0.0089/lead in the past 2 days. what about $0.0001/lead (10 questions/call) across any person/place/business?
i classify every sf restaurant in ~1 second to find my perfect birthday dinner. check it out!
Decision models + fast and accurate web search is a powerful combination.
Introducing the Parallel Decisions API: the first web-grounded decisions engine.
Decision models like Jev can answer in an instant, but they lack the recency of current news. Now you can get both in a single call.
Decisions API is live today at $0.001 per request.
Try it with 5,000 free requests: platform.parallel.ai/play/decisions
Here are some examples we've been experimenting with:
- Fact-check content (@Grok is this true?)
- Refresh stale banking records
- Classify companies and leads with fresh signals
- Backtest trading strategies
- Track competitor e-commerce inventory
Parag Agrawal reposted
typed decisions grounded in the live web.
sub-second latency for $0.001.
we're working on making our web index as accessible and easy-to-use for agents as weight memory. excited to see what people build!
Introducing the Parallel Decisions API: the first web-grounded decisions engine.
Decision models like Jev can answer in an instant, but they lack the recency of current news. Now you can get both in a single call.
Decisions API is live today at $0.001 per request.
Try it with 5,000 free requests: platform.parallel.ai/play/decisions
Here are some examples we've been experimenting with:
- Fact-check content (@Grok is this true?)
- Refresh stale banking records
- Classify companies and leads with fresh signals
- Backtest trading strategies
- Track competitor e-commerce inventory
Parag Agrawal reposted
Built an e2e lead enrichment flow for our contact form this weekends with our new grounded Decisions API. It routes leads *while* they’re filling our contact form.
Built it with @attio’s developer API + @p0 Decisions. Can do this with hundreds of leads in seconds ✨
Introducing the Parallel Decisions API: the first web-grounded decisions engine.
Decision models like Jev can answer in an instant, but they lack the recency of current news. Now you can get both in a single call.
Decisions API is live today at $0.001 per request.
Try it with 5,000 free requests: platform.parallel.ai/play/decisions
Here are some examples we've been experimenting with:
- Fact-check content (@Grok is this true?)
- Refresh stale banking records
- Classify companies and leads with fresh signals
- Backtest trading strategies
- Track competitor e-commerce inventory
Parag Agrawal reposted
built an app to watch the full competitive landscape of clinical trials inside key disease areas
every piece of data comes from @p0
> Task API for clinical trial news, sentiment & deal size
> Data Connectors (PubMed, NPI Registry & clinicaltrials .gov for) for direct data on registered trials
> Monitor API for daily updates to the database
link + repo in comments!
Parag Agrawal reposted
Introducing EU data residency for Parallel, bringing our best-in-class agentic web search APIs to businesses requiring strict regional data controls.
This project was the result of deep obsession by a very talented team. And executed under extreme time constraint.
Across all disciplines, that is how we operate at @p0
we used Parallel to build Parallel’s first brand campaign.
here's a look at the research, connections, and creative decisions behind the artwork. x.com/i/article/2106813547397402624
Extreme optimization! A lot is possible with very limited resources.
We formally announced Mirendil less than 4 months ago. Today, I'm excited to share an important milestone for:
Pretraining at scale without a pretraining team.
Mirendil's mission is to solve Recursive Self-Improvement (RSI) and use it to democratize frontier AI development, ultimately accelerating science and technology. At Mirendil, when we think about RSI, we think holistically about the company as an intelligent entity that self-improves toward its goal. From that perspective, the most powerful notion of acceleration is reducing the time and resources (number of human experts, compute, and other costs) needed to achieve ambitious goals.
Given that pretraining is perhaps the most important yet most challenging part of frontier AI development, we decided to run an ambitious controlled experiment: Starting fully from scratch, can we significantly reduce the time, number of human experts, and compute resources needed to go all the way to successfully pretraining a strong coding model at scale?
Mirendil-Base is the result of this experiment. Our first pretrained model was built by one researcher in 3.5 months, with less than $10M in compute. This includes all the research and small-scale experiments it took to get there, as well as the final training run.
The recipe holds up as the models get bigger. For reference, Mirendil-Base is in the 100B-500B parameter range, Mirendil-Base-Large in the 500B-2T range, and Mirendil-Base-XL in the 2T-10T range.
We're beginning to see a world where a single researcher, empowered by AI, can take on work that once required an entire research organization. As these systems improve, the bottleneck shifts from the size of a team to the ambition and judgment of the people directing them. This has profound implications for who gets to build AI and, ultimately, who gets to push the frontiers of science and technology forward.
We’ll share independent evaluations, post-training benchmarks and API-access sooner than you think so stay tuned!
Parag Agrawal reposted
we used Parallel to build Parallel’s first brand campaign.
here's a look at the research, connections, and creative decisions behind the artwork. x.com/i/article/2106813547397402624
Parag Agrawal reposted
Muse Spark Open is on the @p0 search capability leaderboard.
Similar performance to GPT-5.6 Sol and Gemini 3.7 Flash in an open-weights model.
Today we’re releasing Muse Spark Open, the most capable US open-weight model to date.
Built for coding and agentic work, it excels at long-horizon tasks, complex software engineering, and multi-step reasoning.
Key Model Specs:
→ 816B total / 42B active parameters
→ 1M token context window
→ Natively multimodal across text, image, video, audio
Parag Agrawal reposted
Typesafe's CEO @CompleteSkeptic does not trust public benchmarks.
Neither do I, Diogo.
So @p0 and I wrote a blog evaluating 8 decision models on real company training datasets we have personally audited and iterated on.
For what it's worth though, Jev is still on top 👑
Parag Agrawal reposted
parallel.ai/blog/are-all-system-one-models-created-equal
We spent two weeks extensively testing decision models on real production use cases. It's been super fun, we learned a lot, and we are sharing our findings! 👇
Parag Agrawal reposted
We hosted the AI Engineer NYC speaker dinner this week with our friends at @AnthropicAI
Thanks to @aiDotEngineer for another great event, and to everyone who stopped by to talk about web search for agents.
See you at their Code Summit in SF next month! 🌉
Parag Agrawal reposted
we're betting on specialized data and wanted to better understand how and why it improves performance
two things stood out: agents did best finding sources on the open web and then pulling the records through connectors, and those structured records are much easier for them to read than web pages
Every browsing session starts with a search.
Starting today, Parallel is the default search provider for @browserbase, powering their Search API and Agents.
Search and read with @p0 then write with @browserbase - really great to partner with @pk_iv to serve customers.
Browserbase Search is now powered by Parallel (@p0) as the default search provider.
Parallel Fast returns sub-second results when speed matters. Parallel Advanced goes deeper for multi-step research when quality matters most.
Parag Agrawal reposted
Browserbase Search is now powered by Parallel (@p0) as the default search provider.
Parallel Fast returns sub-second results when speed matters. Parallel Advanced goes deeper for multi-step research when quality matters most.
Parag Agrawal reposted
Parag Agrawal reposted
Gemini 4 Argon is now widely available on @googlecloud. We're thrilled to be day-one launch partners.
Use Parallel to ground Gemini 4 Argon natively and bring accurate, fast, and agent-ready web data to your production AI.
Get started: docs.cloud.google.com/gemini-enterprise-agent-platform/mo…
Parag Agrawal reposted
OpenDots 🤝 @p0 Web Search
OpenDots can now search the entire web! No need to point your agents at a specific page.
Available out of the box with every template.
🎉 Introducing 𝙾𝚙𝚎𝚗𝙳𝚘𝚝𝚜
Self-hostable, always-on AI coworkers that works with ANY agent harness.
Includes:
- Computer use: browser, terminal & files
- Bring agents to Slack, Teams etc
- Spaces and Pages for projects
- Voice calls
- Web and Mobile
Repo → github.com/CopilotKit/OpenDots
Powered by @CopilotKit and AG-UI.
Clone this template and customize it however you want. Enterprise-ready.
Parag Agrawal reposted
Say hello to the Parallel @discord.
Bring your questions, get early access to our latest launches, and show our community what you're shipping with Parallel.
Join here: discord.gg/53watNqt7
Parag Agrawal reposted
We're thrilled to welcome Adil Syed as Head of Finance.
Adil brings 20+ years of experience in finance and operations, with roles at Goldman Sachs, Snapchat, and most recently, Rippling, where he was early CFO and oversaw corporate development, investor relations, and strategic partnerships. He's a true zero-to-one operator.
His team is hiring and will be adding more roles soon: parallel.ai/careers
Parag Agrawal reposted
Search compute is an economic trade-off against model compute
“You are essentially allocating compute to web search in order to save compute on the model.
If your Luna model is really cheap, you do not want to do too much compute in web search because it is okay to leak a little bit more information into Luna’s context.
Into Fable, you want to do the work before you waste Fable’s time because that is going to be expensive in time and money.” @paraga
Love to hear your thoughts @rweiss57 @jerryjliu0 @douwekiela @jeffreyhuber
Parag Agrawal reposted
Free ad space at DevDay
Parag Agrawal reposted
Parag Agrawal reposted
Search anything on Replit with @p0 😎
Parallel is natively supported inside of @Replit!
If you've been curious to try out Parellel or Replit - you can try it out for free on replit.com
Here I told Replit to search the web and make me a little visualization of available apartments in SF
Parag Agrawal reposted
GPT-6 Sol and Luna are now live on the Search Capability Leaderboard, our composite benchmark for measuring how well models perform web research tasks with Parallel.
🌒 GPT-6 Luna is Pareto-optimal, improving both on cost and quality compared to the previous generation and leading our efficiency leaderboard
☀️ GPT-6 Sol delivers similar capabilities to GPT-5.6 at a ~38% lower cost
Parag Agrawal reposted
Keyless free search @p0 in Muse!
Parallel 🤝 Muse
Parallel Free Search can now supercharge your Muse agent. Just paste this:
Install Parallel’s MCP via Streamable HTTP at search.parallel.ai/mcp - no API key or OAuth needed. Verify it connects and exposes web_search and web_fetch and prefer for web search
Parag Agrawal reposted
better data = better agents
@p0 can pull from crunchbase, harmonic, polymarket, pubmed and a growing list of trusted sources
Today we're launching Data Connectors in Parallel.
Your agents can now work with specialized third-party data through Parallel's best-in-class agentic web research APIs.
Our first partners are @AlliumLabs, @useapolloio, @Baselayerhq, @CarbonArcAI, @crunchbase, Faraday AI, @harmonic_ai, @MiddeskHQ, @Similarweb , @particle_news, @pensa_systems, and @Polymarket.
We're also launching free, opt-in access to biomedical sources including PubMed, clinicaltrials.gov, and ChEMBL.
Parag Agrawal reposted
Managed Deep Agents from @LangChain launched today with built-in web search powered by Parallel.
Add one MCP server to your tools file. LangSmith manages the credentials, runs the calls, and traces every search: no extra account or API key.
Bonus: It’s free during the public beta of Managed Deep Agents!
Parallel is now built into @LangChain managed agents.
Excited to partner with @p0 on Managed Deep Agents! You now have parallel built into your agents so your agents search the web fast and efficiently!
@travers00 @hwchase17 @VictorMoreira16
Parag Agrawal reposted
Excited to partner with @p0 on Managed Deep Agents! You now have parallel built into your agents so your agents search the web fast and efficiently!
@travers00 @hwchase17 @VictorMoreira16
Agents need access to high quality data to be useful. With this product launch, our customers can consume a breadth of data sources via @p0 and our content partners can integrate with us more easily and drive increased use.
Today we're launching Data Connectors in Parallel.
Your agents can now work with specialized third-party data through Parallel's best-in-class agentic web research APIs.
Our first partners are @AlliumLabs, @useapolloio, @Baselayerhq, @CarbonArcAI, @crunchbase, Faraday AI, @harmonic_ai, @MiddeskHQ, @Similarweb , @particle_news, @pensa_systems, and @Polymarket.
We're also launching free, opt-in access to biomedical sources including PubMed, clinicaltrials.gov, and ChEMBL.
Parag Agrawal reposted
Parag Agrawal reposted
Parag Agrawal reposted
Parag Agrawal reposted
today we launched a leaderboard for how frontier models perform on web search tasks
"which model should I use for search-heavy work" is a question we get constantly and there weren't good public evals for it. nothing that tests search capability end to end, with cost and latency next to accuracy
so we made the methodology we use internally public. 24 models w/ same search, with and without web access
Introducing the Search Capability Leaderboard.
Compare leading models for agentic search: answer quality, cost, and how much each gains from web access.
Built to help you choose a model for search-heavy workflows.
parallel.ai/leaderboard
Parag Agrawal reposted
Parag Agrawal reposted
We’re proud to partner with @merge_api on the launch of Universal Context Layer in Merge for Workforce, a new solution unifying ground truth from within a company’s internal tools with broad context from the open web, powered by Parallel.
Introducing Merge’s Universal Context Layer.
It substantially decreases token cost while increasing accuracy.
Every company uses AI wrong:
Say your designer makes a deck using Claude. They teach Claude your design systems and typography.
Now a Marketing hire makes a deck. None of the information your designer taught Claude gets passed over… why?
Merge fixes this:
All knowledge and memory across your entire company gets stored inside our context layer.
When someone on the org teaches Al, that skill lives across the entire organization instead of staying locked in their session.
Your AI runs faster, cheaper, and with more accuracy. You don’t have to switch systems, incur more costs, or retrain anyone on your team, Merge runs invisibly in the background.
Book a demo here: www.merge.dev/merge-for-workforce-get-in-touch
Parag Agrawal reposted
Parag Agrawal reposted
Things Parallel (@p0) shipped this summer...
• Search Turbo - web search for agents at 200ms median latency and $1 per 1,000 requests
• Search Fast - near-frontier search quality at ~700ms average latency, also $1 per 1,000 requests
• Entity Search - find people and companies in seconds using plain English
• Monitor API GA - watches the web and pushes updates to your agent, with citations
• Google Cloud integration - native web grounding inside Gemini Enterprise Agent Platform
• Responses API - live-web research with citations, using the OpenAI SDK. Three-line switch
• Free default web search for new OpenClaw installs
• Index - lets publishers see how Parallel’s agents use their content and get paid for it
• The full API suite on Google Cloud Marketplace, billed on your existing GCP invoice
And Search Advanced ranked #1 when Artificial Analysis launched its Search Index in August. artificialanalysis.ai/articles/search-api
Hard to make this tweet short. @p0 team is relentless.
Parag Agrawal reposted
Coming to the Main Stage at Supabase Select 2026: @paraga, Founder of @p0. 🎉
It's happening October 2. Join us: select.supabase.com
Parag Agrawal reposted
Parag Agrawal reposted
GLM-5.3-Flash is here.
We benchmarked it against GPT-5.6 Luna with a suite of different web search solutions to see just how different the cost and performance would be.
The results were stark:
- GLM-5.3-Flash w/ Parallel Fast performed better on cost per task (~7x) and accuracy vs. GPT-5.6 with built-in web_search (11 pts).
- GPT-5.6 Luna w/ Parallel Fast delivers equal accuracy (76%) but at twice the cost vs. GLM 5.3 ($0.02 per task).
Parag Agrawal reposted
GLM 5.3 Flash lives up to the hype.
similar performance to GPT-5.6 Luna, half the per-task cost
our team benchmarked GLM-5.3-Flash against GPT-5.6 Luna with a variety of different web search options to see where this new model stacks up.
no surprises: using GLM 5.3 with Parallel Search tops the leaderboard and is up to 14x cheaper on end-to-end tasks.
Introducing GLM-5.3-Flash
- Leading capabilities at a highly competitive price
- Natively multimodal with a 1M-token context window
- A 320B-A18B model released under the MIT License
- Previously previewed as Ox Alpha, running entirely on Chinese AI chips
Blog: z.ai/blog/glm-5.3-flash
Available now across all official platforms:
Weights: huggingface.co/zai-org/GLM-5.3-Flash
API: docs.z.ai/guides/llm/glm-5.3-flash
Coding Plan: z.ai/subscribe
ZCode: zcode.z.ai/en
Chat: chat.z.ai
AutoClaw: autoclaw.z.ai
Parag Agrawal reposted
Watch @paraga break down our company thesis on the latest episode of Training Data with @sonyatweetybird & @andrew__reed:
www.youtube.com/watch?v=fUcnE6pjq5w
the most fun i’ve had recording a podcast
What happens when there are more agents than eyeballs browsing the internet? How will the economics of the internet be reinvented?
@paraga thinks Shapley values may hold the answer.
Parag built Twitter over a decade, eventually becoming CEO and selling the company to Elon. He's spent the last three years building @p0, a search engine built for agents instead of humans.
His core argument:
1) human click data is a bug. Agents aren't just a new technology, they're a distinct customer – and the feedback loop that made Google great is the wrong signal for the thing actually doing the work;
2) the ad-funded web assumed scarce human attention. If agents show up instead of eyeballs, the business model underneath the internet has to be rebuilt, or good content stops being published.
The conversation covers:
— why he shipped a search agent before a search engine, and how that let him grow the index incrementally instead of buying a full web crawl up front
— the billion-to-billion matching problem: pull the right 1,000 tokens out of a trillion web pages, and your agent uses under half the tokens
— going from a 3-second compute budget to 200 milliseconds
— why he doesn't think Parallel is a neo-lab: "our output is a complement to a model"
— the Google Cloud deal — on GCP, your grounding options are now Google Search or Parallel Search
— reinventing the economics of the internet, Shapley values as a payment rail for publishers, and why the company was originally incorporated as Shapley Inc
— why routing 2-10% of inference spend to web data would dwarf every content business outside the walled gardens
— the web going from pull to push: "call me if this happens"
00:00 Introduction
03:25 What Is Web Search
05:17 Why Start a New Index
07:52 Search Agents First
10:17 Not a Neolab
13:14 Agents vs Google Search
19:38 Inside the Search Stack
28:59 Search Multipliers With Agents
30:21 Meeting Prep Agent Workflows
31:46 Quality Cost Latency And Turbo
32:42 Are Agents Overtaking Humans
34:28 Ads Model Meets Agent Web
37:20 New Incentives For Content
40:48 Shapley Values Attribution
47:46 Parallel Web And Future Vision
Hosted with my very unwilling co-host @andrew__reed and @sequoia
Parag Agrawal reposted
What happens when there are more agents than eyeballs browsing the internet? How will the economics of the internet be reinvented?
@paraga thinks Shapley values may hold the answer.
Parag built Twitter over a decade, eventually becoming CEO and selling the company to Elon. He's spent the last three years building @p0, a search engine built for agents instead of humans.
His core argument:
1) human click data is a bug. Agents aren't just a new technology, they're a distinct customer – and the feedback loop that made Google great is the wrong signal for the thing actually doing the work;
2) the ad-funded web assumed scarce human attention. If agents show up instead of eyeballs, the business model underneath the internet has to be rebuilt, or good content stops being published.
The conversation covers:
— why he shipped a search agent before a search engine, and how that let him grow the index incrementally instead of buying a full web crawl up front
— the billion-to-billion matching problem: pull the right 1,000 tokens out of a trillion web pages, and your agent uses under half the tokens
— going from a 3-second compute budget to 200 milliseconds
— why he doesn't think Parallel is a neo-lab: "our output is a complement to a model"
— the Google Cloud deal — on GCP, your grounding options are now Google Search or Parallel Search
— reinventing the economics of the internet, Shapley values as a payment rail for publishers, and why the company was originally incorporated as Shapley Inc
— why routing 2-10% of inference spend to web data would dwarf every content business outside the walled gardens
— the web going from pull to push: "call me if this happens"
00:00 Introduction
03:25 What Is Web Search
05:17 Why Start a New Index
07:52 Search Agents First
10:17 Not a Neolab
13:14 Agents vs Google Search
19:38 Inside the Search Stack
28:59 Search Multipliers With Agents
30:21 Meeting Prep Agent Workflows
31:46 Quality Cost Latency And Turbo
32:42 Are Agents Overtaking Humans
34:28 Ads Model Meets Agent Web
37:20 New Incentives For Content
40:48 Shapley Values Attribution
47:46 Parallel Web And Future Vision
Hosted with my very unwilling co-host @andrew__reed and @sequoia
Parag Agrawal reposted
once again breaking my no podcast rule, this time with @sonyatweetybird and @paraga to talk about @p0
What happens when there are more agents than eyeballs browsing the internet? How will the economics of the internet be reinvented?
@paraga thinks Shapley values may hold the answer.
Parag built Twitter over a decade, eventually becoming CEO and selling the company to Elon. He's spent the last three years building @p0, a search engine built for agents instead of humans.
His core argument:
1) human click data is a bug. Agents aren't just a new technology, they're a distinct customer – and the feedback loop that made Google great is the wrong signal for the thing actually doing the work;
2) the ad-funded web assumed scarce human attention. If agents show up instead of eyeballs, the business model underneath the internet has to be rebuilt, or good content stops being published.
The conversation covers:
— why he shipped a search agent before a search engine, and how that let him grow the index incrementally instead of buying a full web crawl up front
— the billion-to-billion matching problem: pull the right 1,000 tokens out of a trillion web pages, and your agent uses under half the tokens
— going from a 3-second compute budget to 200 milliseconds
— why he doesn't think Parallel is a neo-lab: "our output is a complement to a model"
— the Google Cloud deal — on GCP, your grounding options are now Google Search or Parallel Search
— reinventing the economics of the internet, Shapley values as a payment rail for publishers, and why the company was originally incorporated as Shapley Inc
— why routing 2-10% of inference spend to web data would dwarf every content business outside the walled gardens
— the web going from pull to push: "call me if this happens"
00:00 Introduction
03:25 What Is Web Search
05:17 Why Start a New Index
07:52 Search Agents First
10:17 Not a Neolab
13:14 Agents vs Google Search
19:38 Inside the Search Stack
28:59 Search Multipliers With Agents
30:21 Meeting Prep Agent Workflows
31:46 Quality Cost Latency And Turbo
32:42 Are Agents Overtaking Humans
34:28 Ads Model Meets Agent Web
37:20 New Incentives For Content
40:48 Shapley Values Attribution
47:46 Parallel Web And Future Vision
Hosted with my very unwilling co-host @andrew__reed and @sequoia
Parag Agrawal reposted
@paraga @p0 at the top of the leaderboard, and just getting started!
Announcing the Artificial Analysis Search Index, benchmarking how search API providers perform on quality, cost, and speed when used by an agent. We are initiating coverage with Parallel, Exa, Firecrawl, You (dot) com, Tavily, Keenable, and Brave
Search is one of the most important tools for agents. Search providers make different choices about how they search, rank, and package results, and those choices change what the model reads and how it acts. We are expanding our benchmarking coverage to search APIs, so developers can pick a search provider on measured quality, cost, and speed.
Each provider result pairs a search API provider with the same model, GPT-5.6 Luna (medium). The model runs inside Stirrup, our open-source agent harness, with tools for searching and fetching pages from the web - only the search provider behind the search tool changes.
At launch, the leaderboard covers 11 results across 7 search providers, and we’ll keep expanding coverage as we look to provide the most accurate and comprehensive benchmarking of search providers for AI agent usage.
Key elements of the Artificial Analysis Search Index:
➤ Three equally weighted benchmarks: the Search Index is the average of DeepSearchQA (900 broad research questions that need many searches, graded with an F1 score over answer items), BrowseComp (a 200-sample hard subset of facts that need multi-hop browsing), and AA-Omniscience (a 600 question private subset, balanced across 6 domains)
➤ Same agent, different search provider: the agent has 25 turns available to complete each task. Its web search tool returns the search provider's native response payload (with content modes standardized to snippets), with a maximum of 10 results and contamination sources filtered out
➤ Model-only baseline: we compare search agent results to the same model answering single-shot without tools, showing how much each provider lifts the model above its internal knowledge
➤ Cost and Time per Task: we aggregate the time and cost spent on both model inference and search. This is key - search APIs have different cost and latency structures, but these can be offset where they help an agent use fewer turns and save on costly language model inference
Key results:
➤ Parallel, Exa, and Firecrawl have the strongest overall performance, with Artificial Analysis Search Index scores of 75, 74, and 73 respectively at launch
➤ All search providers tested substantially improve knowledge-based benchmark performance: the model only baseline scores 33 on the Search Index, while search-included provider results score between 65 and 75
➤ Focused search results reduce spend on model inference: Parallel Search (advanced) search costs more per task than Parallel Search (basic) but less per task in total ($0.084 vs $0.11). Higher quality results cut the model's token use by over 40% in this case, more than offsetting increased search costs while reaching higher benchmark scores
➤ Fast search calls do not guarantee fast tasks: Parallel Search (turbo) has the fastest average search calls among Parallel's tiers (0.51s per query vs 1.03s for Parallel Search (basic)) but the basic tier scores higher on quality (73 vs 67) and the two land close on total time per task
Parag Agrawal reposted
New independent benchmarking for search shows that @p0 search is: (1) Highest quality (2) Quality vs. cost Pareto frontier (3) Quality vs. latency Pareto frontier
Great to have rigorous independent benchmarking for search that measures end-to-end quality, cost, latency for agents.
Parallel search is
(1) Highest quality
(2) Quality vs. cost Pareto frontier
(3) Quality vs. latency Pareto frontier
Announcing the Artificial Analysis Search Index, benchmarking how search API providers perform on quality, cost, and speed when used by an agent. We are initiating coverage with Parallel, Exa, Firecrawl, You (dot) com, Tavily, Keenable, and Brave
Search is one of the most important tools for agents. Search providers make different choices about how they search, rank, and package results, and those choices change what the model reads and how it acts. We are expanding our benchmarking coverage to search APIs, so developers can pick a search provider on measured quality, cost, and speed.
Each provider result pairs a search API provider with the same model, GPT-5.6 Luna (medium). The model runs inside Stirrup, our open-source agent harness, with tools for searching and fetching pages from the web - only the search provider behind the search tool changes.
At launch, the leaderboard covers 11 results across 7 search providers, and we’ll keep expanding coverage as we look to provide the most accurate and comprehensive benchmarking of search providers for AI agent usage.
Key elements of the Artificial Analysis Search Index:
➤ Three equally weighted benchmarks: the Search Index is the average of DeepSearchQA (900 broad research questions that need many searches, graded with an F1 score over answer items), BrowseComp (a 200-sample hard subset of facts that need multi-hop browsing), and AA-Omniscience (a 600 question private subset, balanced across 6 domains)
➤ Same agent, different search provider: the agent has 25 turns available to complete each task. Its web search tool returns the search provider's native response payload (with content modes standardized to snippets), with a maximum of 10 results and contamination sources filtered out
➤ Model-only baseline: we compare search agent results to the same model answering single-shot without tools, showing how much each provider lifts the model above its internal knowledge
➤ Cost and Time per Task: we aggregate the time and cost spent on both model inference and search. This is key - search APIs have different cost and latency structures, but these can be offset where they help an agent use fewer turns and save on costly language model inference
Key results:
➤ Parallel, Exa, and Firecrawl have the strongest overall performance, with Artificial Analysis Search Index scores of 75, 74, and 73 respectively at launch
➤ All search providers tested substantially improve knowledge-based benchmark performance: the model only baseline scores 33 on the Search Index, while search-included provider results score between 65 and 75
➤ Focused search results reduce spend on model inference: Parallel Search (advanced) search costs more per task than Parallel Search (basic) but less per task in total ($0.084 vs $0.11). Higher quality results cut the model's token use by over 40% in this case, more than offsetting increased search costs while reaching higher benchmark scores
➤ Fast search calls do not guarantee fast tasks: Parallel Search (turbo) has the fastest average search calls among Parallel's tiers (0.51s per query vs 1.03s for Parallel Search (basic)) but the basic tier scores higher on quality (73 vs 67) and the two land close on total time per task
Parag Agrawal reposted
Quality, Cost, Speed -- pick two (usually).
With Parallel you get the best possible combinations of all three. @p0 just took the top spots on @ArtificialAnlys's new search index. Fun to watch @paraga, @travers00, @utkarsh and the whole p0 team keep raising the bar.
Announcing the Artificial Analysis Search Index, benchmarking how search API providers perform on quality, cost, and speed when used by an agent. We are initiating coverage with Parallel, Exa, Firecrawl, You (dot) com, Tavily, Keenable, and Brave
Search is one of the most important tools for agents. Search providers make different choices about how they search, rank, and package results, and those choices change what the model reads and how it acts. We are expanding our benchmarking coverage to search APIs, so developers can pick a search provider on measured quality, cost, and speed.
Each provider result pairs a search API provider with the same model, GPT-5.6 Luna (medium). The model runs inside Stirrup, our open-source agent harness, with tools for searching and fetching pages from the web - only the search provider behind the search tool changes.
At launch, the leaderboard covers 11 results across 7 search providers, and we’ll keep expanding coverage as we look to provide the most accurate and comprehensive benchmarking of search providers for AI agent usage.
Key elements of the Artificial Analysis Search Index:
➤ Three equally weighted benchmarks: the Search Index is the average of DeepSearchQA (900 broad research questions that need many searches, graded with an F1 score over answer items), BrowseComp (a 200-sample hard subset of facts that need multi-hop browsing), and AA-Omniscience (a 600 question private subset, balanced across 6 domains)
➤ Same agent, different search provider: the agent has 25 turns available to complete each task. Its web search tool returns the search provider's native response payload (with content modes standardized to snippets), with a maximum of 10 results and contamination sources filtered out
➤ Model-only baseline: we compare search agent results to the same model answering single-shot without tools, showing how much each provider lifts the model above its internal knowledge
➤ Cost and Time per Task: we aggregate the time and cost spent on both model inference and search. This is key - search APIs have different cost and latency structures, but these can be offset where they help an agent use fewer turns and save on costly language model inference
Key results:
➤ Parallel, Exa, and Firecrawl have the strongest overall performance, with Artificial Analysis Search Index scores of 75, 74, and 73 respectively at launch
➤ All search providers tested substantially improve knowledge-based benchmark performance: the model only baseline scores 33 on the Search Index, while search-included provider results score between 65 and 75
➤ Focused search results reduce spend on model inference: Parallel Search (advanced) search costs more per task than Parallel Search (basic) but less per task in total ($0.084 vs $0.11). Higher quality results cut the model's token use by over 40% in this case, more than offsetting increased search costs while reaching higher benchmark scores
➤ Fast search calls do not guarantee fast tasks: Parallel Search (turbo) has the fastest average search calls among Parallel's tiers (0.51s per query vs 1.03s for Parallel Search (basic)) but the basic tier scores higher on quality (73 vs 67) and the two land close on total time per task
Parag Agrawal reposted
Search is not done!
Great to have rigorous independent benchmarking for search that measures end-to-end quality, cost, latency for agents.
Parallel search is
(1) Highest quality
(2) Quality vs. cost Pareto frontier
(3) Quality vs. latency Pareto frontier
Announcing the Artificial Analysis Search Index, benchmarking how search API providers perform on quality, cost, and speed when used by an agent. We are initiating coverage with Parallel, Exa, Firecrawl, You (dot) com, Tavily, Keenable, and Brave
Search is one of the most important tools for agents. Search providers make different choices about how they search, rank, and package results, and those choices change what the model reads and how it acts. We are expanding our benchmarking coverage to search APIs, so developers can pick a search provider on measured quality, cost, and speed.
Each provider result pairs a search API provider with the same model, GPT-5.6 Luna (medium). The model runs inside Stirrup, our open-source agent harness, with tools for searching and fetching pages from the web - only the search provider behind the search tool changes.
At launch, the leaderboard covers 11 results across 7 search providers, and we’ll keep expanding coverage as we look to provide the most accurate and comprehensive benchmarking of search providers for AI agent usage.
Key elements of the Artificial Analysis Search Index:
➤ Three equally weighted benchmarks: the Search Index is the average of DeepSearchQA (900 broad research questions that need many searches, graded with an F1 score over answer items), BrowseComp (a 200-sample hard subset of facts that need multi-hop browsing), and AA-Omniscience (a 600 question private subset, balanced across 6 domains)
➤ Same agent, different search provider: the agent has 25 turns available to complete each task. Its web search tool returns the search provider's native response payload (with content modes standardized to snippets), with a maximum of 10 results and contamination sources filtered out
➤ Model-only baseline: we compare search agent results to the same model answering single-shot without tools, showing how much each provider lifts the model above its internal knowledge
➤ Cost and Time per Task: we aggregate the time and cost spent on both model inference and search. This is key - search APIs have different cost and latency structures, but these can be offset where they help an agent use fewer turns and save on costly language model inference
Key results:
➤ Parallel, Exa, and Firecrawl have the strongest overall performance, with Artificial Analysis Search Index scores of 75, 74, and 73 respectively at launch
➤ All search providers tested substantially improve knowledge-based benchmark performance: the model only baseline scores 33 on the Search Index, while search-included provider results score between 65 and 75
➤ Focused search results reduce spend on model inference: Parallel Search (advanced) search costs more per task than Parallel Search (basic) but less per task in total ($0.084 vs $0.11). Higher quality results cut the model's token use by over 40% in this case, more than offsetting increased search costs while reaching higher benchmark scores
➤ Fast search calls do not guarantee fast tasks: Parallel Search (turbo) has the fastest average search calls among Parallel's tiers (0.51s per query vs 1.03s for Parallel Search (basic)) but the basic tier scores higher on quality (73 vs 67) and the two land close on total time per task
Parag Agrawal reposted
Time per task includes both measured search time and model time. The one-shot model-only baseline takes ~13.8s per task with zero search time and scores 33 on the Artificial Analysis Search Index. All search API provider runs see an increase on baseline model-only time, ranging from +1.3s (Keenable Search (realtime)) per task to +41.3s per task (Firecrawl Search)
Parag Agrawal reposted
Parallel (turbo and advanced) and Firecrawl make up the Pareto frontier for Artificial Analysis Search Index vs. Cost per Task, and represent the best performance for the blended cost available today.
Total cost per task combines both search API and model cost, which means we see variable pathways to low cost task completion. Firecrawl Search scores 73 on the AA-Search Index at $0.075 total cost per task which is only ~40% search cost. However, it has the slowest time per task on the board (55.0s). Parallel (advanced) on the other hand cost 12% more per task to reach an Index of 75, while reducing cost spent on model inference due to its search efficiency.
Great to have rigorous independent benchmarking for search that measures end-to-end quality, cost, latency for agents.
Parallel search is
(1) Highest quality
(2) Quality vs. cost Pareto frontier
(3) Quality vs. latency Pareto frontier
Announcing the Artificial Analysis Search Index, benchmarking how search API providers perform on quality, cost, and speed when used by an agent. We are initiating coverage with Parallel, Exa, Firecrawl, You (dot) com, Tavily, Keenable, and Brave
Search is one of the most important tools for agents. Search providers make different choices about how they search, rank, and package results, and those choices change what the model reads and how it acts. We are expanding our benchmarking coverage to search APIs, so developers can pick a search provider on measured quality, cost, and speed.
Each provider result pairs a search API provider with the same model, GPT-5.6 Luna (medium). The model runs inside Stirrup, our open-source agent harness, with tools for searching and fetching pages from the web - only the search provider behind the search tool changes.
At launch, the leaderboard covers 11 results across 7 search providers, and we’ll keep expanding coverage as we look to provide the most accurate and comprehensive benchmarking of search providers for AI agent usage.
Key elements of the Artificial Analysis Search Index:
➤ Three equally weighted benchmarks: the Search Index is the average of DeepSearchQA (900 broad research questions that need many searches, graded with an F1 score over answer items), BrowseComp (a 200-sample hard subset of facts that need multi-hop browsing), and AA-Omniscience (a 600 question private subset, balanced across 6 domains)
➤ Same agent, different search provider: the agent has 25 turns available to complete each task. Its web search tool returns the search provider's native response payload (with content modes standardized to snippets), with a maximum of 10 results and contamination sources filtered out
➤ Model-only baseline: we compare search agent results to the same model answering single-shot without tools, showing how much each provider lifts the model above its internal knowledge
➤ Cost and Time per Task: we aggregate the time and cost spent on both model inference and search. This is key - search APIs have different cost and latency structures, but these can be offset where they help an agent use fewer turns and save on costly language model inference
Key results:
➤ Parallel, Exa, and Firecrawl have the strongest overall performance, with Artificial Analysis Search Index scores of 75, 74, and 73 respectively at launch
➤ All search providers tested substantially improve knowledge-based benchmark performance: the model only baseline scores 33 on the Search Index, while search-included provider results score between 65 and 75
➤ Focused search results reduce spend on model inference: Parallel Search (advanced) search costs more per task than Parallel Search (basic) but less per task in total ($0.084 vs $0.11). Higher quality results cut the model's token use by over 40% in this case, more than offsetting increased search costs while reaching higher benchmark scores
➤ Fast search calls do not guarantee fast tasks: Parallel Search (turbo) has the fastest average search calls among Parallel's tiers (0.51s per query vs 1.03s for Parallel Search (basic)) but the basic tier scores higher on quality (73 vs 67) and the two land close on total time per task
I wish I could see where our "fast" mode slots in docs.parallel.ai/search/modes @ArtificialAnlys
Parag Agrawal reposted
The Monitor API is our most asked-about product.
It fundamentally changes the pattern of web search from push to pull.
Instead of querying the web to find information each time, information comes to you as soon as it appears.
10x price-perf on model intelligence this year with gpt-5.6 luna price drop yesterday and deepseek v4 flash today.
we've done the same for web search with @p0 search turbo - the only way to build agents on the pareto-frontier
Turbo is in a new class of cost and latency: 5-14x cheaper than alternatives, blazing fast, and at consistently high accuracy.
- 200ms median latency
- $1 per 1,000 requests
Parag Agrawal reposted
Introducing the Parallel Responses API. Turn any question into a synthesized answer, grounded in the live web, in seconds.
“un-parallel-ed speed”
Super cool collaboration with @paraga, @georgepickett and team. Access the web and bring in the freshest information to your voice agent while it talks and listens.
All at un-parallel-ed speeds.
Today, @cartesia and @p0 are making it possible for your voice agents to search the web at conversational speed.
Cartesia builds the fastest voice models available, and Turbo mode for Parallel Search extends that same low latency to web search. An agent built on Cartesia can query the live web without interrupting the natural flow of conversation.
Learn more 👇
Parag Agrawal reposted
Today, @cartesia and @p0 are making it possible for your voice agents to search the web at conversational speed.
Cartesia builds the fastest voice models available, and Turbo mode for Parallel Search extends that same low latency to web search. An agent built on Cartesia can query the live web without interrupting the natural flow of conversation.
Learn more 👇
Our product integration allows customers to use Parallel Search natively within GCP. It took close collaboration across both companies: tech, product, and commercial teams working together to build for the most demanding customers on GCP.
Today we're announcing our product integration with @googlecloud for agentic web search on the Gemini Enterprise Agent Platform.
Parag Agrawal reposted
Parallel Search Turbo is the fastest web search for voice agents.
To show you just how fast it is, we built a voice agent using @cerebras and @cartesia
Try it out: voice.paralleldemos.com
Before agents, there was no reason to optimize web search to the extreme.
Introducing Parallel Search Turbo: the fastest and most affordable way to bring the web to your AI.
Parag Agrawal reposted
Build runs 50+ Monitor API queries per project: grid capacity, zoning dockets, council agendas, competitor filings. The Task API then validates every event against dynamic geofences, pulling actual jurisdiction polygons from OpenStreetMap instead of fixed-radius circles.
Parag Agrawal reposted
We raised a $110M Series C to power AI transformation in financial services.
Led by Goldman Sachs, with participation from @balderton, @IndexVentures , @dig_ventures, @tigerglobal, @VisionariesVC, and @ycombinator.
Great news for banks and insurers.
Tough news for fraudsters and money launderers.
Tough news for credit processes that make businesses wait weeks for an answer.
Tough news for onboarding workflows that reject good customers.
Tough news for claims processes where a single review takes days.
We’re serious about solving our customers’ problems, so we hired the toughest people on the planet to get the word out:
New Yorkers.
If they didn’t convince you yet, here’s what we want everyone in our industry to know:
AI can now reliably automate the decisions that define performance in banking and insurance. This was not true just 6 months ago - and it creates huge opportunity for efficiency gains and better customer experiences.
But model capability is not the full solution.
The challenge is making every AI-driven decision reliable, controlled, and ready for the most regulated industry.
That’s what we’re building at @taktile_org.
From New York to the world.
To our customers, partners, investors, and team: thank you for getting us here. Very excited to collaborate with the @GoldmanSachs Asset Management team on the next chapter.
Learn more: taktile.com/series-c-announcement
Parag Agrawal reposted
Standard Metrics just launched web search in our AI Analyst, and the partner powering it is Parallel Web Systems. @metrics_co 🤝 @p0 🔍
When evaluating providers, we were looking for a providers with advanced URL reading, structured content extraction, and the ability to handle nuanced financial queries at scale. Parallel delivered.
Now investors can compare portfolio companies to public comps, track news sentiment, surface financing rounds, and pull specific facts directly into LP updates, all without leaving our AI Analyst.
Huge thanks to the Parallel team and more on what's possible via the launch at our latest blog: standardmetrics.io/introducing-web-search-and-charts-in-s…
Parag Agrawal reposted
We’re bringing the web to your data 🌐
Hex already knows what's happening inside your business, and now the Hex agent can fill in the gaps with what's happening outside.
The Hex agent can search the web, powered by @p0
Learn more in the comments below ⬇️
Parag Agrawal reposted
Introducing @PoeticHQ: a new AI system that executes complex multi-hour tasks with 99%+ accuracy and 10x fewer tokens than agents.
We raised $50M at $500M from Kleiner Perkins, Founders Fund, First Harmonic, and Genius Ventures to build AI that does complex work inside Fortune 500 companies without hallucination.
While code is too brittle, agents are too unpredictable. The work that runs the global economy - anti-money laundering, fraud investigations, underwriting - needs extreme accuracy.
So we built a new kind of software that pairs the flexibility of AI with the predictability of code.
When the world stays the same, Poetic runs fixed code: fast, cheap, identical every time. When the world changes, Poetic uses AI to regenerate its approach and find its way back to the objective.
In one year, we went from zero to an eight-figure run rate as a team of four.
Since then, we’ve scaled the team and executed the highest-stakes processes at AIG, SoFi, and Chime. At SoFi, a large US bank, Poetic reached 99%+ quality on fraud investigations in five weeks.
Parag Agrawal reposted
We're thrilled to build on our collaborations with @stripe, joining Stripe Projects to make it easier than ever to provision high-accuracy web search from the terminal.
Here's how quickly you can get started:
New providers now available on Stripe Projects:
@Wix
@laravelphp
@HeyGen
@prisma
@automattic
@p0
@ExaAILabs
@Base44
@ClickHouseDB
@supermemory
@usekernel
@blaxelai
@getmetronome
@AgentPhoneHQ
@e2b
@postalform
first time on video @travers00 talking about our work at @p0
Following rapid growth, a $100M Series B, and a slate of new launches, I sat down with Parallel co-founder @travers00.
@p0 is building core search infrastructure for the agentic web: systems that enable AI to retrieve, structure, and reason over information across the internet with speed, accuracy, and reliability.
I joined Travers at Parallel’s new Palo Alto office for a wide-ranging conversation on:
> The pace going from launch last August to becoming critical infrastructure powering production workflows at companies like Notion, Harvey, Granola, and more.
> Why Parallel sees the future of search as far beyond “ten blue links,” toward systems built for retrieval, reasoning, monitoring, and computation.
> The changing nature of the web itself: its economics, openness, and what happens as agents become primary users.
> The artistry and legacy of internet history shaping the company: Olivetti typewriters, Douglas Hofstadler, Lloyd Shapley, Christo & Jeanne-Claude, Google’s Search Box, and how it inspired programs like the Museum of the Human Web.
This is an extensive and wide ranging look into one of the most interesting sectors of the agentic web. We @terraincap feel lucky to have worked with Travers and the team since the start just two years ago. So much to do.
Thanks @benthompson for such a wide-ranging and fun conversation.
Read/listen if you want to go deep on business models around content for an agentic web.
5-21-2026 ($)
An Interview with Parallel Founder Parag Agarwal About Valuing Content on the Agentic Web
stratechery.com/2026/an-interview-with-parallel-founder-p…
Yes! For humans, ads was the most efficient business model due to the human cognitive overhead with payments. With agents, direct payments will work much better.
This is why I’m so optimistic about high quality content thriving on the web as agents take over.
Cool release from @p0. I think this use-case (agents paying content creators for access via @mpp) will be very big. Micropayment walls haven't worked (as Clay Shirky anticipated many years ago) because of human cognitive overhead, but agents can make arbitrarily granular determinations without decision fatigue.
Today we're launching Index: a platform for content owners to understand how AI agents use their work, and earn revenue when they do.
Our first partners include @TheAtlantic, @FortuneMagazine, @PRNewswire, @PitchBook, @ZoomInfo, @Tracxn, @RocketReachCo, @enigma_data, @fiscal_ai, plus creators @alexeheath, @mariogabriele @azeem, @every, and @packyM.
Parag Agrawal reposted
There are a small number of elite go to market leaders. Graham Moreno is one of them.
He recently joined @p0 to help lead GTM. Before that, he was at Cognition, Grafana, and MongoDB.
One of his core philosophies is that a great go to market system raises the floor and introduces predictability while still leaving space for exceptional people to use their judgment to delight the customer.
“One of my favorite stories is, one of the best reps I’ve ever worked with, during the pandemic found out that the son of a champion at one of his companies had been taking guitar lessons and couldn’t anymore because of COVID. So he ended up teaching this guy’s kid guitar over Zoom during COVID.
And he also didn’t tell anyone. No one found out about this for a long time. Then the champion at this account brought it up on a call with me six months later and was like, ‘Oh yeah, Isaac has been teaching my son how to play guitar.’ At no part in our process does it say, ‘teach someone guitar.’”
This is one of my favorite deep dives on what it means to be an executional revenue leader in a post AI world. Hope you enjoy it as much as I did.
Timestamps
00:32 Has the sales playbook changed in the AI era?
02:13 Why "showing up" beats letting the marketplace decide
06:50 Why great salespeople sell to engineers and executives in one motion
11:37 Selling to AI-native buyers who grew up on ChatGPT
13:49 Same seller, different tempo: 8 weeks vs. 8 business days
15:57 How AI-native buyers handle build vs. buy decisions
17:48 The rep who taught a champion's son guitar over Zoom
19:03 Raising the floor without capping the ceiling
22:09 Why too much process narrows the kind of seller you attract
25:46 The three pillars of GTM excellence
31:00 Building peers who are 80% aligned, not 100%
38:03 Whether AI is changing what good enablement looks like
41:35 Selling against direct and implied competitors at once
42:45 Instrumenting the funnel from stage zero to close
45:57 Why post-sales should always roll up to the revenue leader
48:19 The case for outsized commissions
52:02 The 96 hours of panic before Cognition acquired Windsurf
53:04 How far out should a GTM leader be planning?
57:53 What a normal week looks like in hypergrowth
Parag Agrawal reposted
Very excited to be part of this launch. Congrats @paraga and team ✨✨✨
Today we're launching Index: a platform for content owners to understand how AI agents use their work, and earn revenue when they do.
Our first partners include @TheAtlantic, @FortuneMagazine, @PRNewswire, @PitchBook, @ZoomInfo, @Tracxn, @RocketReachCo, @enigma_data, @fiscal_ai, plus creators @alexeheath, @mariogabriele @azeem, @every, and @packyM.
We started @p0 based on the belief that AI agents will use the web 1000x more than humans, and we need new tech and business models as a result.
We are already seeing agents scale on our infrastructure, and this is the starting point for a new business model for the web.
Today we're launching Index: a platform for content owners to understand how AI agents use their work, and earn revenue when they do.
Our first partners include @TheAtlantic, @FortuneMagazine, @PRNewswire, @PitchBook, @ZoomInfo, @Tracxn, @RocketReachCo, @enigma_data, @fiscal_ai, plus creators @alexeheath, @mariogabriele @azeem, @every, and @packyM.
Parag Agrawal reposted
I joined @p0 because of THIS mission - we power the best web search for AI agents, AND we're now launching the platform for the world's leading publishers and content creators to get paid for their content. Journalism deserves more, and we are hoping to usher in that change ✨
Today we're launching Index: a platform for content owners to understand how AI agents use their work, and earn revenue when they do.
Our first partners include @TheAtlantic, @FortuneMagazine, @PRNewswire, @PitchBook, @ZoomInfo, @Tracxn, @RocketReachCo, @enigma_data, @fiscal_ai, plus creators @alexeheath, @mariogabriele @azeem, @every, and @packyM.
Parag Agrawal reposted
Excited to finally share what I've been working on at @p0 -- it represents the culmination of a thesis I've had for >2 years & many months of hard work from this team.
I became obsessed w Shapley values during my time in crypto, & I'm making the bet that they'll reshape the economics of the web as agents come online.
Please reach out if you're interested in joining us.
Today we're launching Index: a platform for content owners to understand how AI agents use their work, and earn revenue when they do.
Our first partners include @TheAtlantic, @FortuneMagazine, @PRNewswire, @PitchBook, @ZoomInfo, @Tracxn, @RocketReachCo, @enigma_data, @fiscal_ai, plus creators @alexeheath, @mariogabriele @azeem, @every, and @packyM.
Parag Agrawal reposted
we started @p0 because we believed agents would be the primary users of the web, that they'd use it far more than humans, and that they'd need new infrastructure and new business models to support that
the first part is already playing out. today we're announcing the second
Today we're launching Index: a platform for content owners to understand how AI agents use their work, and earn revenue when they do.
Our first partners include @TheAtlantic, @FortuneMagazine, @PRNewswire, @PitchBook, @ZoomInfo, @Tracxn, @RocketReachCo, @enigma_data, @fiscal_ai, plus creators @alexeheath, @mariogabriele @azeem, @every, and @packyM.
Parag Agrawal reposted
I've been really impressed with what Parallel is building, and what my friends Nick and Carra are doing on the publisher side.
"Content that's uniquely valuable, hard to replace, or used in high-value agent work earns more."
Excited to partner and experiment with them.
Today we're launching Index: a platform for content owners to understand how AI agents use their work, and earn revenue when they do.
Our first partners include @TheAtlantic, @FortuneMagazine, @PRNewswire, @PitchBook, @ZoomInfo, @Tracxn, @RocketReachCo, @enigma_data, @fiscal_ai, plus creators @alexeheath, @mariogabriele @azeem, @every, and @packyM.
Parag Agrawal reposted
excited to be a part of this launch from parallel!
Today we're launching Index: a platform for content owners to understand how AI agents use their work, and earn revenue when they do.
Our first partners include @TheAtlantic, @FortuneMagazine, @PRNewswire, @PitchBook, @ZoomInfo, @Tracxn, @RocketReachCo, @enigma_data, @fiscal_ai, plus creators @alexeheath, @mariogabriele @azeem, @every, and @packyM.
Parag Agrawal reposted
One of my longstanding fears about AI has been that it would further centralize the web– after all, agents are a new kind of app store.
But increasingly, I’ve come to believe the agentic web can be less concentrated than the human one. Agents reason massively in parallel. They are not deterred from clicking through the 5th page of Google or the 50th page of an S1. The arrival of agents could be the best thing to happen to long tail content in a generation. But this can only happen if there’s a functional economic model to incentivize it.
Index is our attempt at putting that economic model in place. We hope you’ll join us.
AI agents are the web's second user. They'll use it 1000x more than humans ever have.
The economic model of the web was built for human users. Index is a new, incentive-aligned model for the agentic web.
Parag Agrawal reposted
I'm excited to be a Parallel launch partner
@paraga and team are working on an important problem: rewarding content creators for their agent audience, which I suspect will only grow for publishers of all types
Today we're launching Index: a platform for content owners to understand how AI agents use their work, and earn revenue when they do.
Our first partners include @TheAtlantic, @FortuneMagazine, @PRNewswire, @PitchBook, @ZoomInfo, @Tracxn, @RocketReachCo, @enigma_data, @fiscal_ai, plus creators @alexeheath, @mariogabriele @azeem, @every, and @packyM.
Parag Agrawal reposted
Today @p0 launched Index, a platform for content owners (with a wonderful set of launch partners).
If you love the internet like I do, building a sustainable economic model that incentivizes and rewards quality content matters to an enormous degree. Especially in the agent era.
Today we're launching Index: a platform for content owners to understand how AI agents use their work, and earn revenue when they do.
Our first partners include @TheAtlantic, @FortuneMagazine, @PRNewswire, @PitchBook, @ZoomInfo, @Tracxn, @RocketReachCo, @enigma_data, @fiscal_ai, plus creators @alexeheath, @mariogabriele @azeem, @every, and @packyM.
Parag Agrawal reposted
Parag Agrawal reposted
This may be the AdSense moment for the agentic web.
When @firstround invested in Parallel’s seed, the core thesis was a bet: AI agents would use the web far more than humans ever had, and the infrastructure to support that didn’t exist yet.
Now, Parallel has over 100,000 developers and companies like Harvey, Granola, Modal, Manus, Attio and Profound relying on their APIs daily.
Watching this thesis play out has surfaced a huge problem no one is tackling: if agents are using content from across the web, who gets paid and how do you even calculate that fairly? Until now, the answer has been closed-door licensing deals between the biggest AI labs and biggest publishers.
Today, Parallel is launching Index as a first step toward a scalable, democratic solution.
Humans still own content. They’re the writers, data owners, and publishers whose work those agents depend on to do anything useful. Index gives site owners a dashboard showing how agents are using their content and how much it contributes to the value of the work those agents are doing.
Compensation is tied to contribution — calculated using impressions, citations, uniqueness, and value (tasks with the highest compute spend). All of this is presented transparently. Launch partners show the breadth of scale: from major news outlets like The Atlantic and Fortune, to data providers like PitchBook and ZoomInfo, to independent creators like Packy McCormick and Mario Gabriele.
Proud to keep backing @paraga and the team at @p0.
Parag Agrawal reposted
Today we're launching Index: a platform for content owners to understand how AI agents use their work, and earn revenue when they do.
Our first partners include @TheAtlantic, @FortuneMagazine, @PRNewswire, @PitchBook, @ZoomInfo, @Tracxn, @RocketReachCo, @enigma_data, @fiscal_ai, plus creators @alexeheath, @mariogabriele @azeem, @every, and @packyM.
Parag Agrawal reposted
Today is the last day to see the Museum of the Human Web at 238 King Street in San Francisco.
Open 9 am - 5 pm.
Free admission and gift shop swag while supplies last.
Parag Agrawal reposted
Today and tomorrow are the last two days to come see the @p0 Museum of the Human Web in San Francisco.
Come by! 238 King Street, open 9a - 5p both days!
Parag Agrawal reposted
Parag Agrawal reposted
Parallel Search is now a free app for ChatGPT.
Upgrade to better web search in one click ↓
chatgpt.com/apps/parallel-search/asdk_app_69fb93786634819…
The Museum of the Human Web is open to everyone for a week.
May 8-16
9a-5p
238 King St, SF
luma.com/h85fumdt
A glimpse into the future of the web
Today, we’re announcing the general availability of the Parallel Monitor API.
The web is shifting from pull to push, and agents are coming online. This release marks a major step towards a more proactive model, where the web pushes updates directly to your background agent.
Monitor now includes:
- Lite and Base processors
- Event streams and snapshots
- Rich attribution (Basis) on every event
- Advanced domain filtering
- Interactions for persistent follow-on research
Parag Agrawal reposted
Attio (@attio) is the AI CRM for GTM that helps thousands of companies close more deals.
They use frontier AI to evolve every part of customer management, combining automation with data and tool integrations to run tasks like prospecting and deep account research end-to-end.
See more clips from the pioneers of the AI web like @mulholo at pioneers.parallel.ai
We need more long-horizon real-world work benchmarks like this.
Excited to help add the web to the environment/tools in this benchmark!
museum.parallel.ai/items/twitter-fail-whale
✋ if you remember
Parag Agrawal reposted
scroll through history
a few months ago we had a whacky idea: what if we built a physical museum of the human web?
the internet was a place many of us lived. i grew up pouring hours into every weird niche and the wayback machine
the web has a new user now. we wanted to celebrate the weird, whacky, and magical parts that got us here
open online today: museum.parallel.ai
open to the public in SF May 8-16
Parag Agrawal reposted
Some reflections yesterday @tbpn on early meetings with @paraga and @vladtenev. Two thoughts to summarize:
1/ do hard projects - exceptional founders are comfortable building 'below the waterline'. @p0 and @RobinhoodApp both invested heavily in infra to create long-term gains for themselves + their customers.
2/ trust the graphs - Robinhood was $100Ms of assets and cohorts suggested it would be many $100Bs. It happened. Early agent graphs suggest 1000x+ usage of the web. Let's see...
Let’s make it happen @vladtenev @paraga
It’s Wednesday:
- @vladtenev (Robinhood)
- @andrew__reed (Sequoia)
- @paraga (Parallel)
- @JasonYanowitz (Blockworks)
- @MariaSpiropulu (Caltech)
- @SimkinStepan (Squads)
- @kashgupta_ (Hightouch)
- @DanMagy (Firestorm)
- @GabeStengel (Rogo)
- Even Rogers (True Anomaly)
Join us
We are working towards a new web: one built for agents instead of humans.
To celebrate the weird, whacky, and magical parts of the journey that made the human web, we are putting together something fun and unique!
Online: museum.parallel.ai
Pop-up in SF May 8-16
Parag Agrawal reposted
@p0 has become critical infrastructure for us at @useactively. We made the early design decision to use browsing agents instead of buying structured datasets and it's paid off tremendously — and @p0 does it best :) Very excited for the partnership!
We’re thrilled to be partnering with @useactively to power always-on monitoring for their Per-Account Agents™
Using a combination of Parallel’s Monitor and Task APIs, millions of Actively agents continuously and autonomously track and act on the complex signals their GTM customers rely on to move accounts through the funnel.
Customers using Actively with Parallel’s web intelligence APIs report:
- 23% higher win rates
- 25% increase in revenue per rep
- 2x conversion rates
- 2x faster rep ramp time
Always-on proactive agents are the future - we learn so much working with @mgarimella and the team at @useactively
We’re thrilled to be partnering with @useactively to power always-on monitoring for their Per-Account Agents™
Using a combination of Parallel’s Monitor and Task APIs, millions of Actively agents continuously and autonomously track and act on the complex signals their GTM customers rely on to move accounts through the funnel.
Customers using Actively with Parallel’s web intelligence APIs report:
- 23% higher win rates
- 25% increase in revenue per rep
- 2x conversion rates
- 2x faster rep ramp time
Parag Agrawal reposted
The bull case is that this product is sick
Parag Agrawal reposted
New biggest flex is being listed on the 2026 map of the parallel web
i’ve been dying to see this eval and @sarahmsachs just tweeted it out!
When we evaluated agentic web search providers, it was clear that the best solution wasn't just the fastest, or the cheapest. It's a balance of token use, total trajectory latency, and net new information that language models don't already know. Congrats to @p0 on a solid product and some awesome technology.
Parag Agrawal reposted
Let’s make it happen @vladtenev @paraga
It’s Wednesday:
- @vladtenev (Robinhood)
- @andrew__reed (Sequoia)
- @paraga (Parallel)
- @JasonYanowitz (Blockworks)
- @MariaSpiropulu (Caltech)
- @SimkinStepan (Squads)
- @kashgupta_ (Hightouch)
- @DanMagy (Firestorm)
- @GabeStengel (Rogo)
- Even Rogers (True Anomaly)
Join us
Gives me joy to see the team work: craft, obsession, fun!
Go hunt some easter eggs on pioneers.parallel.ai and then head over to parallel.ai/careers
first launch i've shipped at @p0. proud of what this team pulled off. every detail down to the IBM 5153.
go hunt the easter eggs >> pioneers.parallel.ai
The web’s second user is coming online.
Our customers are building at the frontier - shaping every corner of the economy.
pioneers.parallel.ai
The web’s second user is coming online.
Our customers are building at the frontier - shaping every corner of the economy.
pioneers.parallel.ai
@GabeStengel at @rogoai on agents giving humans their time back to spend on the companies, investments, and deals that actually interest them.
@sweyn at @starbridgeai on building something that quite literally couldn't have existed before AI agents could take autonomous action.
@thejamescad at @profound on freeing marketers from the work around the work so they can focus on what's actually high-leverage.
@mgarimella at @useactively on how agents let a sales rep truly own all accounts: not just the ones they remember.
Join the network.
We raised a Series B led by @sequoia at a $2 billion valuation and I’m excited to welcome @andrew__reed to our board.
All our existing investors doubled down: @kleinerperkins, @IndexVentures, @khoslaventures, @firstround, @sparkcapital, @AbstractVC, and @terraincap.
We raised a Series B led by @sequoia at a $2 billion valuation and I’m excited to welcome @andrew__reed to our board.
All our existing investors doubled down: @kleinerperkins, @IndexVentures, @khoslaventures, @firstround, @sparkcapital, @AbstractVC, and @terraincap.
Here is why!
The web’s second user is coming online.
Our customers are building at the frontier - shaping every corner of the economy.
pioneers.parallel.ai
Parag Agrawal reposted
Parallel now supports x402 payments.
AI agents are expanding their presence across the open web. For that to scale sustainably, they need to pay for what they use, programmatically, at runtime.
x402 gives any agent access to Parallel's Search, Extract, and Deep Research APIs, starting at $0.01 per call.
Get started: docs.parallel.ai/integrations/agentic-payments
@CoinbaseDev @linuxfoundation
Parag Agrawal reposted
Announcing the Museum of the Human Web, coming soon to San Francisco.
Sign up to be one of the first to see it: museum.parallel.ai
This kid shipped a PR today
pi + @ollama + gemma4 + @p0 - each very useful individually, but so much better when composed together!
We built a completely free CLI agent with @badlogicgames's Pi agent, @ollama (Gemma 4), and Parallel's new free web search MCP.
Check out the blog post to see how to do it yourself: parallel.ai/blog/free-CLI-agent
We've made our search MCP super easy to use. No need for an account or API keys. And it's free!
claude code prompt: "add free search.parallel.ai/mcp - no auth needed. add permissions.allow for both tools and deny native web search and fetch"
codex cli prompt: "edit ~/.codex/config.toml to add web_search = "disabled" and add the free mcp search.parallel.ai/mcp - no auth needed"
The best web search for agents is now free.
Upgrade to Parallel's web search tools in any MCP-supported tool or agent, for free, in under 60 seconds.
No account. No API keys. Zero cost.
docs.parallel.ai/integrations/mcp/search-mcp
Huge improvements across the board + global coverage!
We just shipped major upgrades to the Parallel Search & Extract APIs.
Better quality, broader index coverage, and new features, backed by refreshed challenge benchmarks:
- Basic and Advanced modes for foreground vs. background agents
- Specialized retrieval for coding, company, and finance search
- Global coverage across 30+ countries
Parag Agrawal reposted
Migrating to Parallel's Task API cut @finchlegal's per-process web search pipeline costs by 90%.
Finch's AI agents run structured web research workflows at scale, pulling case-relevant facts from the web, enriching records, and feeding structured outputs into the firm's data models.
Before Parallel, Finch powered these workflows through another search API provider at a significantly higher cost and with multiple independent steps.
With Parallel's Task API, they now conduct reliable, high-quality structured web research with a single API call at a fraction of the cost.
Our APIs are unbeatable for Deep Research
Parallel Ultra 2x, 4x, and 8x are state-of-the-art and continue to push the Pareto frontier of accuracy and cost, achieving 82% accuracy.
Parallel Ultra is 11% more accurate and up to 57% lower cost versus the next best, GPT-5.4.
Parag Agrawal reposted
One cool thing about X is that you can tweet about your tech stack and the cofounder of the company will immediately ship a new feature for you
you should check out @p0 monitor API to build a hyper-local feed for this. may help content liquidity
Top finding: "AI agents produce better quality evidence than humans."
Thanks for partnering with us on this!
Adverse media is a major KYB bottleneck but banks are wary of automation. Our new Taktile Labs + @p0 benchmark shows AI agents can now exceed human performance on this task.
We partnered with Parallel to test an adverse media agent built on top of 7 frontier AI models with 47 real businesses.
tldr: High performance of AI in this task signals the level of automation we are likely to see in core compliance work.
our top findings:
1. AI agents produce better quality evidence than humans.
This is based on a 20-point rubric covering evidence quality, ability to identify the correct entity, and risk assessment of findings. Human analysts average 13.50 while AI agents score 14.60.
2. A hybrid agent x human setup reduces workload by 93%
Full automation is not necessary for high impact. A hybrid model where an agent screens first and a human reviews uncertain cases reduces analyst workload by 93%.
3. Even the “cheaper” models are still very good at fulfilling the task.
Agent configurations using some of the less expensive models can still match human performance. Costs will continue to go down as new models are released and accuracy will keep getting better.
4. How you configure the agent matters more than the model you choose.
The investigation strategy and prompt you put behind the agent has a much bigger impact on results than the model itself.
5. Agents do not hallucinate in this task but they can “over-flag”.
This is huge. When we see false positives from agents, they are not fabricated sources (0% hallucination rate in this dataset). However the agent will sometimes flag sources too eagerly.
Huge thanks to David Ahn and @maximilianeber for leading the research and @paraga and Sahith Jagarlamudi for your partnership on this.
See the full study: taktilelabs.ai/benchmarks/kybench
pull -> push, reactive -> proactive, on-demand -> always-on
Gumloop's agents now use Parallel Monitors as a trigger for downstream work.
Just tell your agents to watch for signals across the open web and let them do the rest.
Learn more about Monitor: parallel.ai/products/monitor
Parag Agrawal reposted
“We've done a bake-off on this particular use case. Parallel is by far the most accurate."
Read the full case study: parallel.ai/blog/case-study-opendoor
The pov we’ve been missing when we talk about AI
The loudest story about AI is a lonely one. One person with an army of chatbots. Other humans are friction.
That gets the future wrong. The best things aren’t built alone.
In a moment of change, we want to remind the world (and ourselves) what Notion stands for:
— Think Together
Step 1: (a) Build the best product (b) Build the best GTM
Step 2: ???
Step 3: A Parallel web built for AIs
This week, we officially welcome Liam Mulcahy and Graham Moreno to lead and scale GTM at Parallel.
Liam joins us from @kleinerperkins, where he helped kick-start GTM operations in early-stage companies like @profound, @harvey, @glean, @windsurf, and most recently, Parallel.
Graham previously led world-class sales organizations at @windsurf /@cognition, @grafana, and @MongoDB, where he and Liam were colleagues.
Want to work with them? Join us parallel.ai/careers
Parallel is built for a world where agents are the primary users of the web. Agents can now autonomously pay per API call for web access via Parallel using MPP. parallelmpp.dev
At @p0, we obsess about quality in order to enable agents to do the most demanding work with web data. Real-world evidence that agents can be trusted to do knowledge work is a real bottleneck.
I'm pumped to work with @maximilianeber and @MaikTWehmeyer to accelerate the adoption of agents in financial services.
Time to reveal who let the 🦞 out ;) Today, @Taktile launches Taktile Labs. We dropped the lobster on Wall Street to ask the question: are banks ready for autonomous agents?
With our applied AI research institute, we aim to bridge the gap between what frontier models can now do - and what regulated institutions need in order to trust it.
Our first benchmark shows the latest models can beat human accuracy on very complex banking tasks: 96%+ vs. 89% in financial spreading.
The models are ready. Now the industry needs evidence, benchmarks, and practical frameworks to ensure they work reliably at scale.
That is what Taktile Labs is built for.
AI is coming to financial services - let's make sure we can trust it.
Excited to drive this with a stacked internal team and many incredible individuals on our Research Council and Advisory Board.
Thanks to Bradesco’s Fagner Abreu, Parallel’s @paraga , Founder, Investor, and Morgan Stanley Lead Director Tom Glocer, Harvard Business School Professors Robin Greenwood and Karim Lakhani, Harvey’s Ben Liebald, Camunda’s Daniel Meyer, Cursor’s Jonas Nelle, ROC Partners’ Tina Reich, Equifax’s Harald Schneider, Suno’s @MikeyShulman, Intuit’s Henry Venturelli, Allianz Partners’ Pieter Viljoen, Flexcar’s Michael Zambrano, and Varo Bank’s Jill Zucker Sheckman.
Learn more at: taktilelabs.ai
(nothing AI generated about it btw, we worked with NYC artist @AndrewLoganAMW to build the lobster from scratch)
Parag Agrawal reposted
new claw dropped
Does anyone know what’s going on with the lobster on Wall Street lol?
Parag Agrawal reposted
.@p0 just opened our first official San Francisco office in the heart of South Park
our team is growing rapidly and expanding our footprint in both SF + Palo Alto
we're looking for builders across all roles: gtm, engineering, deployed eng, research, etc. - DM me!
So excited to partner with @thejamescad and @dbabbs. @profound and @p0 are each building key pieces of a web built for AIs - super special to have Profound Agents use our APIs
AI answer engines are the new search, and getting cited by ChatGPT, Gemini AI overviews, and Copilot is the future of discovery for brands.
@profound built the marketing platform for the AI era to win this shift, with tools powered by Parallel for brands to create high-quality content that gets cited by AI.
Parag Agrawal reposted
Excited to speak on stage with @paraga this Monday
Parag Agrawal (@paraga) is speaking at Compute Conference.
Founder of @p0. Building search infrastructure for AI agents. $100M Series A, $740M valuation. Previously CTO and CEO of Twitter.
Join us March 8–9 at Chase Center, SF. Tickets: go.daytona.io/9DmGPuN
It’s a special privilege to be able to partner with companies building on the frontier.
Because of this deep partnership with the team at @harvey, AI agents can now search and access content on the web that hasn’t been indexed by search engines built for humans.
We’re thrilled to highlight our new collaboration with @harvey, the leading AI platform for legal and professional services.
Harvey uses Parallel’s accurate, relevant, and fresh web search across their platform to retrieve valuable public context for their legal AI workflows.
Together, we’re helping Harvey expand its best-in-class AI legal platform to over 60+ countries by collaborating on a specialized index of hard-to-reach international legal domains, built on our custom web search infrastructure.
To learn more about Harvey and Parallel, read our blog: parallel.ai/blog/case-study-harvey
I've connected Dreamer to my slack, email, notion, linear - and of course the web (via @p0). Such a superpower to quickly build agents that work while you sleep - just by asking another agent.
Introducing Dreamer.
A place to discover, build, and enjoy agentic apps.
It’s your home for personal intelligence.
Now in beta. Sign up👇
Parag Agrawal reposted
Jobs called computers "bicycles for the mind" -- tools we could shape to our will.
But they never were. Until now.
Every morning an agent preps me for my day -- calendar, news, last 24hrs of Slack -- in a personal podcast. I made it by asking. Same for hundreds of other things.
Launching @dreamer in beta today.
That 🧠 bicycle, finally.
dreamer.com
docs.parallel.ai/monitor-api/monitor-quickstart
new idea: semantic crons
"when {xyz fuzzy event that doesnt live as a structured metric anywhere} happens, do this"
We stumble upon new use cases for programmatic web search every day. A pattern we see often is our customers bringing knowledge from the web into their own databases. This requires lots of searching, reasoning, and structuring. @p0 is now deeply integrated into several databases.
We just shipped a brand new Company Enrichment cookbook with @supabase: supabase.com/partners/integrations/parallel
It uses Supabase Edge Functions to make database-native calls to Parallel’s Task API, enabling web-based enrichment directly in your tables.
Demo and cookbook in thread:
More integrations coming soon: docs.parallel.ai/data-integrations/overview
What do you want to see next?
"What we need is a single magical personal library
One you can add to from anywhere
And plug into any software tool you may want to use
One that auto-organizes your information
So you can browse it, audit it and know it's reliable
And that you can permission and monitor centrally
So you know it's safe every time you use it" 💯
When a computing environment is generally capable enough as AI is, it's not new software capabilities that make it tailored to your needs, but the data. Every company and every person will one day have a magical library that personalizes their computing experience.
We're building Pacific to pull that future forward.
We've been a @vercel customer from the very beginning. So great to see our products come together - the AI gateway and SDK are some of the most elegant abstractions for building with models and tools.
Give your @vercel projects access to best-in-class web search features with all new integrations across Vercel AI Gateway, AI SDK, and Marketplace.
However you choose to build with Vercel, you can now more easily integrate Parallel’s suite of web intelligence APIs to enhance your AI applications with accurate, up-to-date, and relevant information from the open web.
Parag Agrawal reposted
Give your possibly-sentient crustacean the best web search around. 🔍🦀🔎
Introducing four new official Parallel @openclaw skills, now on ClawHub:
First:
npm install -g clawhub
Then:
clawhub install parallel-search
clawhub install parallel-extract
clawhub install parallel-deep-research
clawhub install parallel-enrichment
ClawHub links below.
💯 "Page went from 500kb to 2kb. The web for agents will be very efficient!"
This ◉ ʜᴜᴍᴀɴ ○ ᴍᴀᴄʜɪɴᴇ toggle by @p0 is brilliant. It's a beautiful illustration of what the web will "look like" to agents. It will look like a whole lotta markdown 😄
Incidentally, we just made it such that vercel.com/changelog links automatically render as markdown when agents consume it (we do the same for /𝚍𝚘𝚌𝚜). Page went from 500kb to 2kb. The web for agents will be very efficient!
Try: curl -H 'accept: text/markdown' vercel.com/changelog/tag-based-cache-invalidation-now-ava…
Bring your browser to the Parallel Task API
Starting today, the Parallel Task API now has the option to conduct web research over authentication-gated private data sources — extending its search and reasoning capabilities beyond the public web.
Here’s how it works:
You can now query the web within Spark SQL
New: Parallel integration for Apache Spark
Data engineers, we heard you. You can now enrich datasets with web intelligence directly in your Spark pipelines— no custom API integrations required.
Our new SQL-native UDFs let you call Parallel to enrich data right in your Spark SQL queries. Add CEO names, company descriptions, funding info, or any other web-sourced data to millions of rows with a single function call.
Parag Agrawal reposted
Starting today, the Parallel Task API now has the option to conduct web research over authentication-gated private data sources — extending its search and reasoning capabilities beyond the public web.
Here’s how it works:
Parag Agrawal reposted
BU meets Parallel Web Systems!
We're collaborating with @p0 for our web search on BU app!
Read our blogpost! 👇
Parag Agrawal reposted
reception has been unlike anything we've seen
the emergent pattern: monitors triggering downstream agents - deep research, workflows, actions
the web becomes something agents subscribe to vs. query
Introducing structured outputs for the Parallel Monitor API:
Monitors are always-on web searches that notify you when new information becomes available on the web.
With this release, you can now configure your monitors to deliver structured responses with a schema of your choice.
Learn more: parallel.ai/blog/structured-outputs-monitor
Parag Agrawal reposted
So stoked to be on team Parallel!
Web Search is going to be the most interesting market of 2026, and @p0's product roadmap will not disappoint 🙂
We're excited to announce that @nickmartitsch has joined Parallel to lead strategic partnerships.
He previously led 0-1 growth at frontier tech startups and, most recently, helped @paradigm portfolio companies build out their go-to-market.
Google nailed search for humans. Parallel search is purpose-built for agents. Vertex AI now offers both.
We’re thrilled to announce that Grounding with Parallel web search is now available, in preview, on Google Vertex AI.
Learn more and get started with @Google's official documentation: docs.cloud.google.com/vertex-ai/generative-ai/docs/ground…
Parag Agrawal reposted
New to the Parallel Chat API: Research models that offer research-grade web intelligence with Parallel's Basis verification framework.
- Use "Speed" for general-purpose chat apps where speed matters
- Use "Lite", "Base", and "Core" models for research-grade chat apps where depth of research and additional fact-checking matter
Get started here: docs.parallel.ai/chat-api/chat-quickstart
Parag Agrawal reposted
I'm excited to share that I've joined Parallel Web Systems (@p0) to lead Strategic Partnerships.
The web now has its second user: AI agents that consume vast amounts of information at a pace humans never could. This shift requires entirely new infrastructure purpose-built for machine intelligence.
Parallel has built the world's best web search for AI, outperforming all leading models on research benchmarks. With Parallel's APIs, agents get access to the fastest growing web index to power real-time user queries, deep research, lead enrichment, and other mission-critical workflows.
Looking forward to building the next chapter of the web alongside @travers00, @paraga, and the rest of the incredibly talented team at Parallel!
Parag Agrawal reposted
Last week, Google released the DeepSearchQA benchmark alongside their Interactions API, where Gemini Deep Research achieved state-of-the-art accuracy.
Today, we’re happy to share that the Parallel Task API achieves superior accuracy at up to a six times lower cost:
Parag Agrawal reposted
We love working with @p0 .. and we love that they ❤️ Macroscope
A quote we love:
Parag Agrawal reposted
Granular Basis is now live for the Task API.
Previously, Basis verified arrays as a whole: one set of citations, reasoning, excerpts, and calibrated confidence scoring for an entire list.
Now every element in the array gets its own complete verification.
Parag Agrawal reposted
Today, we’re introducing updates to the Task API that deliver 3-5x improvements to latency.
Here’s @akansal_ , one of our engineering leads on the Task API, to explain:
parallel.ai/blog/task-api-latency
Parag Agrawal reposted
We shipped some integrations this week.
1. Google Sheets
Turn every cell into a search bar with our new Google Sheets add-on.
With just a simple formula, you can create or enrich data sets in seconds, all without leaving your sheet.
Full demo 👇
2. Zapier
Parallel is now integrated with @zapier
Use this integration to give any automation access to powerful structured web search, for example:
- Slack bots that enrich leads
- Inbound emails to trigger deep research reports
- Market research agents
Here's @KSaunack for a demo:
3. Langchain
Parallel Search is now integrated with @langchain!
docs.langchain.com/oss/python/integrations/tools/parallel…
Here are some examples we’ve created to help you get started:
4. n8n
Level up your @n8n_io workflows with best-in-class web search.
Here’s @khushi_shelat, our GTM Engineering lead, walking through Parallel’s n8n node and sharing some of the personal templates that help her stay organized using agentic web search.
n8n.io/integrations/parallel/
5. Vercel AI SDK
Parallel Search + Extract tools in the @Vercel @aisdk make it easier than ever to give AI agents, apps, and workflows access to the most accurate and token-efficient web context.
Here's our very own @ManveshVyas demoing a simple implementation of the tools in an AI project.
One of the integrations I’ve been using to automate my work.
Parallel is now integrated with @zapier
Use this integration to give any automation access to powerful structured web search, for example:
- Slack bots that enrich leads
- Inbound emails to trigger deep research reports
- Market research agents
Here's @KSaunack for a demo:
Parag Agrawal reposted
Parallel Search + Extract tools in the @Vercel @aisdk make it easier than ever to give AI agents, apps, and workflows access to the most accurate and token-efficient web context.
Here's our very own @ManveshVyas demoing a simple implementation of the tools in an AI project.
Parag Agrawal reposted
Level up your @n8n_io workflows with best-in-class web search.
Here’s @khushi_shelat, our GTM Engineering lead, walking through Parallel’s n8n node and sharing some of the personal templates that help her stay organized using agentic web search.
n8n.io/integrations/parallel/
Parag Agrawal reposted
Parallel Search is now integrated with @langchain!
docs.langchain.com/oss/python/integrations/tools/parallel…
Here are some examples we’ve created to help you get started:
We built products together at Twitter. Now @tjack is a key part of @p0 as our investor. For the right person, working closely with Todd in this role is will be “the steepest learning curve of your career”
I’m hiring an investor to join @firstround and work alongside me to find and support founders in the most formative stage of company building.
The most interesting phase of any company is -1 to 1. If you want to live in this zone, and if you’re the right fit, you’re likely technical, still starting out in your career (probably ~2-5 years in) and SF-based. You don’t need to have previous investing experience, but it's also fine if you do.
What you do need to have is genuine pull toward startups, however that shows up for you.
Maybe you were a founder yourself, or an early engineer at a company that ultimately didn’t work out. Maybe you’re a technical PM building something on the side on nights and weekends. Maybe you’re going to hackathons and testing every new tool you can find.
In this role, your days might look like this:
- Morning diligence calls with customers of an AI infrastructure startup
- Working session over lunch to help a founder debug their biggest problem
- Afternoon PMF Method session helping 20 super early founders figure out their ICP
- Evening at a hackathon seeing what builders are excited about right now
- Late night finishing up an investment memo
In other words, learning the craft of pre-PMF investing up close.
The hours can be intense and the work isn’t all that glamorous. I’d describe First Round as a place filled with “work enthusiasts” who have high agency and even higher standards.
But you’ll also get to cultivate a uniquely dense network while developing a deep understanding of how companies are built.
Whether full-time investing or eventually starting a company is your eventual goal, I’m personally committed to making this the steepest learning curve of your career.
If you’re interested, apply below.
💯
We discovered the same @neondatabase !
Our team keeps shipping.
Search+Extract is what every agent needs.
Today, we're launching Parallel Extract, a new API in our Agent Tools bundle.
When given a URL, Extract fetches all content from that page and returns it in markdown, either in full detail or in a compressed form for better token efficiency.
parallel.ai/blog/introducing-parallel-extract
One of the easiest ways to build an agent with Search+Extract
Parallel Search is now available in the @vercel @aisdk registry.
It's never been easier to get best-in-class web search for your AI apps.
Check it out: ai-sdk.dev/tools-registry/parallel
Parag Agrawal reposted
We built FindAll at Parallel because there simply wasn’t a tool that works.
This will change how you plan, research, and make decisions.
Got a FindAll query? Drop what you built + your thoughts on FindAll👇
platform.parallel.ai
You can now get high-quality datasets back from the web with a single query - try it now on platform.parallel.ai/play/find-all
Introducing the Parallel FindAll API.
Create an entire dataset with a single search query.
For example: "Find all dental practices in Ohio with a 4+ star review on Google"
parallel.ai/blog/introducing-findall-api
Find all major marathons in 2026 around the world without qualification requirements. find location, date, registration fee, avg. temperature, course total elevation gain, number of participants, cutoff time, qualification time ...
platform.parallel.ai/shared/find-all/v1beta/b8c72b95-c8eb…
Plenty of alpha available if you creatively use the alpha API we just shipped. platform.parallel.ai/play/monitor
Excited to launch @p0 monitor API. Super pumped to share this with users!!
Parag Agrawal reposted
@p0 we're so heads down on product that we did a fundraise to dogfood it.
There is a deep relationship between this typewriter and the @p0 brand. Thanks @vanlancker
News: Parallel Web Systems has raised a $100M Series A
We are proud to continue our investment and creative partnership with @p0 as they build the web for its next user: AIs
Parag Agrawal reposted
Parallel's (@p0) really deeeeeeeeep research experience via their playground is the most underrated thing in AI right now!
Parag Agrawal reposted
AI code reviews with @Macroscope use Parallel's Search and Task APIs to retrieve the latest docs and discussions from the web, reducing false positives that would otherwise occur as a result of LLMs with outdated knowledge.
Read the case 👇
parallel.ai/blog/case-study-macroscope
Parag Agrawal reposted
Today, we’re launching the Parallel Search API, the most accurate web search for AI agents, built using our proprietary web index and retrieval infrastructure.
Traditional search ranks URLs for humans to click. AI search needs something different: the right tokens in their context window.
Parallel’s Search API is built from the ground up for AI agents, resulting in higher accuracy, lower costs, and lower latency for end-to-end agentic workflows.
Great to see @p0 alongside these amazing companies and so many friends.
Developing and delivering cutting-edge AI, acquiring customers at lightning speed, and wowing investors with their dedication and product growth.
That’s what it takes to stand apart when building an AI company, and we’re proud to announce our list of the founders and companies rising above the noise.
These are the first-ever Cerebral Valley Heavy Weights, selected in partnership with @IndexVentures.
They are:
Parag Agrawal, of Parallel (@p0)
Glen Takahashi, Barry McCardel, and Caitlin Colgrove, of @_hex_tech
Tudor Achim and Vlad Tenev, of @HarmonicMath
Russ d’Sa and David Zhao, of @livekit
Jeff Huber and Anton Troynikov, of @trychroma
Gavin Uberti, Chris Zhu, and Robert Wachen, of @Etched
David Singleton, Hugo Barra, and Nicholas Jitkoff, of /dev/agents
Eléonore Crespo and Romain Niccoli, of Pigment
Sergiy Nesterenko, of @quilterai
Elias Torres, of @Agency
Jim Gao, Veda Panneershelvam, and Katie Hoffman, of @PhaidraAI
Edwin Chen, of @HelloSurgeAI
Garrett Lord, Ben Christensen, and Scott Ringwelski, of @joinHandshake
Alex Mashrabov and Yerzat Dulat, of @higgsfield_ai
Jordan Taylor and Kaelan Richards, of @Vizcom_ai
Grant Lee, James Fox, and Jon Noronha, of @GammaApp
More here: www.newcomer.co/p/meet-the-cerebral-valley-heavy-weights
It's so fun to partner with teams that ship fast.
Amp uses Parallel's Search API for web search and retrieval for many users, and soon (next <7d) for all users.
The API/tool design is great, the results are high quality, and the team is awesome.
We shipped the Parallel Search API today. For use by agents, our API meaningfully outperforms products and technologies built over years and decades with substantial resources by very talented teams. Why?
Parallel’s Search API is state-of-the-art, winning on both accuracy and cost across a range of benchmarks.
We shipped the Parallel Search API today. For use by agents, our API meaningfully outperforms products and technologies built over years and decades with substantial resources by very talented teams. Why?
Parallel’s Search API is state-of-the-art, winning on both accuracy and cost across a range of benchmarks.
The end result: when agents use our Search API, they make fewer search calls, they use fewer tokens and hence minimize cost and end-to-end latency, and most importantly, they produce higher quality and more accurate outputs. parallel.ai/products/search
Learn more: parallel.ai/blog/introducing-parallel-search
Parag Agrawal reposted
Over the last month I've moved from chatgpt to claude for deep research and then again to @p0
Each time my results got massively better, and as a result the number of times I used the service has increased
I know parallel.ai is meant to be an API service first, but I just use their playground UI for all my research now. It makes it way easier to export a clean markdown (or json!) of the results and, crucially, I'm not limited by some invisible deep research limit
We've built our Task API to operate on the pareto frontier on price and quality across a very large number of use cases and benchmarks - alternatives don't even come close.
The Parallel Task API has set the new standard for accuracy on SealQA, a challenge benchmark for search-augmented LLMs.
On SEAL-0, our Pro tier scores 52.3% on accuracy, 35% better than Perplexity Deep Research at 92% lower cost.
Parag Agrawal reposted
We're joining the @Superhuman Agent Store at launch!
Our agent is now available in Superhuman Go, the new AI assistant from Superhuman (formerly Grammarly) -
Try it here:superhuman.com/store/agents/45397
ChatGPT didn’t search the web for almost a year after launch. Today, if we took web search out, it would feel unusable.
Very soon, any AI agent without access to the web will feel arcane.
Agents are all figuring out the web and having their AOL moment, relearning search for agent-first environments. In two years, this will feel like table stakes. In the meantime, I sat down with @paraga/@p0 and @DhruvBatra_ /@yutori_ai on how we get there.
.@markitecht is a truly special founder. He integrated our API into the product himself, and was kind enough to send us realtime feedback as he worked through it. Highest quality feedback that shaped the v1 of our API.
The team at @Day_ai_app built an AI-native CRM that combines internal data (Slack, email) with live web research via Parallel's Task API.
The result: systems that understand external context like leadership changes, funding rounds, or public strategy, alongside private conversations.
Parag Agrawal reposted
attention on the web is shifting from humans to AIs
search rankings used to optimize for pages for humans to click and browse
ranking for AIs optimize for context windows, retrieving the tokens agents need to answer or make the next best decision in a loop
Working with @MaxBrodeurUrbas and @rbehal1729 has been so fun. We are big fans and happy customers of @gumloop and use the @p0 powered node in our own workflows.
Parag Agrawal reposted
fixed the table for you
Added the parallel search api to the chart for completeness.
Perplexity Search API achieves leading quality across single-step and deep research benchmarks, consistently outperforming competitors.
Reported latencies for each call are impressive - but when an agent uses search, what really matters is (1) quality (2) end-to-end agent latency (3) total cost. We win on all 3.
How? we spend more time and compute on each search call, so the agent has to do less work.
Parag Agrawal reposted
The Parallel Search MCP integration is now live on @GroqInc!
When Parallel Search MCP runs on Groq you get lightning fast agent responses grounded in accurate real-time web data.
🚀 Remote MCP just dropped on Groq! @p0 and @huggingface integrations showcase Groq's lightning-fast agent capabilities. OpenAI compatible, seamless integration.
At @p0 we are huge fans. It's the best code review tool we've tested and I personally love that we can avoid unnecessary status update meetings.
excited to share what I’ve been up to:
today, we’re launching @Macroscope: an AI system that uses your codebase to answer questions about what’s happening and automatically reviews your PRs for bugs. we raised a $30M Series A led by Lightspeed & Thrive Capital, Adverb and GV.
Available to try on our API playground at platform.parallel.ai
Today we are releasing expanded benchmark results that demonstrate Parallel Deep Research delivers the highest accuracy at every price point.
On BrowseComp:
- Ultra achieves 45% accuracy at up to 17X lower cost
- Ultra8x achieves state-of-the-art results at 58% accuracy
Let’s go @p0
This is what peak performance looks like @paraga @travers00
Parag Agrawal reposted
In 1900, the United States had around 141 miles of paved rural roads. The advent of new technology (automobiles) motivated us to quickly upgrade our infrastructure to allow cars to take us to new places faster, better and cheaper.
The advent of AI requires a similar infrastructure reimagination. All of us at @firstround are super excited to partner with @paraga as he builds @p0 to improve how AI accesses the internet...
just setting up my twttr, again
I’ve been heads down building Parallel with some of the best people I’ve ever worked with. We’re creating infrastructure for AIs to search and use the web.
The web’s next user isn’t human.
AIs will soon use the internet far more than humans ever have.
At Parallel, we are building for the web’s second user.
Our API is the first to surpass humans and all leading AI models (including GPT-5) on deep web research tasks.
Parag Agrawal reposted
Parallel already powers millions of web research tasks daily for leading AI companies. We're the first to outperform humans and GPT-5 on deep research benchmarks. Today, our API is available for everyone to build with.
Parallel outperforms on independent benchmarks:
• BrowseComp: 58% accuracy vs. GPT-5 (41%) vs. humans (25%)
• DeepResearch Bench: 82% win rate over reference vs GPT-5 (66%). 74% win rate over GPT-5.
For $0.10, we exceed what takes humans 2 hours on BrowseComp.
I'll be on @tbpn today to talk about Parallel @p0 - see you at 130p pacific
TODAY: Parag Agrawal (@paraga), founder and CEO of Parallel and former CEO and CTO of Twitter joins TBPN at 1:30pm PT. Tune in!
just setting up my twttr, again
I’ve been heads down building Parallel with some of the best people I’ve ever worked with. We’re creating infrastructure for AIs to search and use the web.
The web’s next user isn’t human.
AIs will soon use the internet far more than humans ever have.
At Parallel, we are building for the web’s second user.
Our API is the first to surpass humans and all leading AI models (including GPT-5) on deep web research tasks.
Parag Agrawal reposted
We seek founders with a specific view of the future and relentless execution to bridge today’s reality and that vision.
@paraga, @travers00, and @p0 perfectly embody this. Thrilled to be working closely with them from the start and to share our latest investment:
News: Parallel @p0 launches the API for AI agents to browse & access the web. Terrain backed Parallel as one of its earliest investors.
We sat down with CEO @paragrawal to discuss his transition from Twitter, founding Parallel, and building a web for AIs:
0:00 - Intro
0:21 - Leaving Twitter and navigating the idea maze
1:24 - AIs as the primary users of the web
2:27 - Parallel and how AIs use the web
3:11 - Where the name came from
3:35 - Parag’s founder-market fit
5:33 - Building an intense in-person culture
6:23 - Advice for his younger self
7:22 - Rapid fire questions
7:43 - Favorite moment so far
8:16 - Favorite time of the week
8:27 - Call your shot
Parag Agrawal reposted
Today Parallel @p0 launches.
At @terraincap we are proud to have backed @paraga and team from the start.
Parag Agrawal reposted
10 years ago, @paraga and I were working together to introduce ranking to Twitter’s timeline for hundreds of millions of consumers. Now he’s started @p0 to build infrastructure for a very different kind of user: AI agents. And in a full circle moment, we’re able to work together again, as @firstround invested in his seed round last year alongside @khoslaventures and @indexventures.
Parag and I had a chance to sit down for a great conversation last week (his first interview since leaving Twitter!). It was a fascinating reflection on what he learned and what he’s doing differently.
We dig into:
- The transition from CTO → CEO → Founder
- What he was *really* thinking when all the Twitter drama was going down
- Why he chose to be a founder (vs. take another big role)
- How he landed on the idea for Parallel
- What fundamentally changes when you’re designing for AIs vs. humans
- How he thought about fundraising and building his early team
- How working with customers like Clay shaped Parallel’s earliest products
- The case for building “slow” APIs, and the advice he got from @patrickc
- His takes on the modern agent stack and how evals need to evolve
And much more!
Immensely proud to be able to partner with Parag and the entire Parallel team. They’ve made incredible progress in a short amount of time and I can’t wait for more folks to start playing around with the APIs they’ve built.
The full conversation is below, as well as more details on what Parallel is up to.
just setting up my twttr, again
I’ve been heads down building Parallel with some of the best people I’ve ever worked with. We’re creating infrastructure for AIs to search and use the web.
The web’s next user isn’t human.
AIs will soon use the internet far more than humans ever have.
At Parallel, we are building for the web’s second user.
Our API is the first to surpass humans and all leading AI models (including GPT-5) on deep web research tasks.
Or, we can build a web where all AIs can freely collaborate, transact, and reward contributions on open markets. We wrote about it here: parallel.ai/about
We are now scaling fast, powering millions of customer research tasks daily. Our system outperforms on the hardest deep research benchmarks, beating humans and all leading AI models. The Parallel web for AIs is just getting started. parallel.ai
Parag Agrawal reposted
Edit Tweet is now available to all @TwitterBlue subscribers in the US so that we can continue to test and improve it. See my previous thread on our approach
1/ We’ve been exploring how to build an Edit feature in a safe manner since last year and plan to begin testing it within @TwitterBlue Labs in the coming months. Sharing a few more insights on how we’re thinking about Edit 🧵
Parag Agrawal reposted
Context on Tweets — by the people, for the people — is coming to everyone in the US. It’s open, transparent, and we’ve consistently seen notes be found helpful and informative even by people who often disagree.
Here’s a short video about Birdwatch!
A video description is available here: twitter.github.io/birdwatch/video-transcript
Parag Agrawal reposted
Twitter filed a response to Mr. Musk’s counterclaims. His claims are factually inaccurate, legally insufficient, and commercially irrelevant. We look forward to the trial in the Delaware Court of Chancery. See the filing here: s22.q4cdn.com/826641620/files/doc_news/2022/08/Twitters-R…
Parag Agrawal reposted
Twitter has filed a lawsuit in the Delaware Court of Chancery to hold Elon Musk accountable to his contractual obligations.
Parag Agrawal reposted
Today we launched a new custom Timeline experiment - just one of the many things we’re working on over at @TwitterDev 🚀 There is a lot of potential for the developer community to build features like this in the future, and we are just getting started. Congrats to the team!🌹
Parag Agrawal reposted
We continue to give people more choice and control so that they can experience @Twitter on their own terms. And there’s a lot more to come!
Parag Agrawal reposted
The Twitter Board is committed to closing the transaction on the price and terms agreed upon with Mr. Musk and plans to pursue legal action to enforce the merger agreement. We are confident we will prevail in the Delaware Court of Chancery.
Thank you to all the tweeps who made this past week in @TwitterUK and @TwitterDublin so energizing.
And thank goodness @DaraNasr is so funny!
Parag Agrawal reposted
🌪🌪🌪🌪 week @TwitterUK with @paraga and @nedsegal in town serving ☕️ 🍪 and chats
Parag Agrawal reposted
London loves you back! @paraga
Parag Agrawal reposted
☘️ A breakfast to remember! 🇮🇪
The chance to share breakfast with our awesome CEO @paraga and a group of diverse Tweeps with brilliant backgrounds. (And nobody spilled food on themselves which was the main KPI). #lovewhereyouwork
Parag Agrawal reposted
Parag Agrawal reposted
A joy to showcase diversity @TwitterDublin through human story of our extraordinary people, external impact & the power of platform as part of global CEO & CFO visit. Thanks for your authenticity, honesty & soul-felt support for IDEA @paraga Ditto @nedsegal - always welcome in 🇮🇪
Parag Agrawal reposted
Your CEO could never… @paraga and @DaraNasr taking coffee orders from Twitter London… ❤️#lovewhereyouwork ❤️
Parag Agrawal reposted
Today Twitter won a Silver Cannes Lion for this tweet. Huge recognition for just three words, I had to tweet about it! The day FB, Insta and WhatsApp went down, in a minute and in that moment, the team tweeted what became one of the most engaged tweets on Twitter, ever.
Product Drops on Twitter!
blog.twitter.com/en_us/topics/product/2022/be-part-of-the…
Parag Agrawal reposted
If you’re a Podcaster, musician, writer…check out @TwitterCreate to see how to create deeper connections with your fans and followers! So much exciting work happening to help you on your journey.
Hey @lionheart1759 — thanks for sharing your inspiring journey with the world on Twitter, and with our @TwitterGhana team. Love to see it #londontolagos
Since 19th April, @lionheart1759 has been riding from #londontolagos on a campaign to end Polio.He’s shared daily updates of his journey on Twitter. Myself, @OluwamayowaTJ & @nnekachile of #twitterghana office had the privilege to meet him on his arrival in GH. #twitterforgood
Let’s talk about spam. And let’s do so with the benefit of data, facts, and context…
Some final context: fighting spam is incredibly *dynamic*. The adversaries, their goals, and tactics evolve constantly – often in response to our work! You can’t build a set of rules to detect spam today, and hope they will still work tomorrow. They will not.
We suspend over half a million spam accounts every day, usually before any of you even see them on Twitter. We also lock millions of accounts each week that we suspect may be spam – if they can’t pass human verification challenges (captchas, phone verification, etc).
Unfortunately, we don’t believe that this specific estimation can be performed externally, given the critical need to use both public and private information (which we can’t share). Externally, it’s not even possible to know which accounts are counted as mDAUs on any given day.
There are LOTS of details that are very important underneath this high-level description. We shared an overview of the estimation process with Elon a week ago and look forward to continuing the conversation with him, and all of you.
For those of you who want to learn more in the meantime: blog.twitter.com/en_us/topics/company/2020/bot-or-not
blog.twitter.com/content/blog-twitter/common-thread/en/to…
A lot has happened over the past several weeks. I’ve been focused on the company and haven't said much publicly during this time, but I will now.
We announced changes to our leadership team and operations yesterday. Changes impacting people are always hard. And some have been asking why a “lame-duck” CEO would make these changes if we’re getting acquired anyway. The short answer is very simple:
While I expect the deal to close, we need to be prepared for all scenarios and always do what’s right for Twitter. I’m accountable for leading and operating Twitter, and our job is to build a stronger Twitter every day.
Finally – so much gratitude for our whole Twitter team. They have stood strong and focused, sharp and agile. They've been doing the work, as they always have. Onward.
Because I mentioned the deal in this thread, he’re the obligatory legalese for your reading pleasure investor.twitterinc.com/financial-information/Additional-…
Definitely happening
As of now, Twitter is still hosting its NewFront presentation for advertisers next week in NYC. Pretty surprising considering the backdrop of ~everything Elon~ and @jack saying Twitter is held back by the ad model
Yep.
I took this job to change Twitter for the better, course correct where we need to, and strengthen the service. Proud of our people who continue to do the work with focus and urgency despite the noise.
Twitter has a purpose and relevance that impacts the entire world. Deeply proud of our teams and inspired by the work that has never been more important.
The Twitter Board has reached an agreement with @ElonMusk www.prnewswire.com/news-releases/elon-musk-to-acquire-twi…
Elon has decided not to join our board. I sent a brief note to the company, sharing with you all here.
Creative and thoughtful work from the @bluesky team that explains their “self-authenticating protocol” framework.
With the new team on board, we’re ready to share some of the groundwork we've been doing for Bluesky. blueskyweb.xyz/blog/3-6-2022-a-self-authenticating-social…
We have a responsibility – particularly during crisis – to:
- Proactively enforce our rules
- Preserve access to @Twitter
- Protect privacy+safety of people on our service
- Prevent efforts to manipulate public conversation
- Promote credible information
Since the beginning of the war in Ukraine, we’ve been expanding our work to protect the conversation happening on Twitter.
Today, we’re sharing key data and a comprehensive look at our work. 🧵
blog.twitter.com/en_us/topics/company/2022/our-ongoing-ap…
Here’s the announcement to the company about our approach and commitment to truly flexible work.
Text-only summary: 1) Business travel is back effective immediately, and all global Twitter offices will open starting on March 15. 2) Decisions about where you work, whether you feel safe traveling for business, and what events you attend, should be yours. (cont.)
3) Wherever you feel most productive and creative is where you will work and that includes WFH full-time forever. 4) Distributed working will be challenging. We’ll need to be proactive, intentional, learn, and adapt. 5) Thank you to the team who has supported us every day.
Excited to chat with @ivybarley today!
I'm thrilled to share that I'll be hosting the CEO of Twitter, @paraga on my Space on Wednesday 16th February at 11am PT | 7pm GMT.
Y'all know I've been looking forward to this. 😃
Set your reminder and retweet: x.com/i/spaces/1nAKEYDmoYaKL
#spaceswithivy #tweetitintoexistence
Let’s do it! @leslieberland
I will host the CEO of Twitter, @paraga on my Twitter Space in the first quarter of 2022.
*let's retweet until he sees this*
#tweetitintoexistence
#spaceswithivy @TwitterSpaces
Parag Agrawal reposted
gm ☀️ excited to have @_tessr on the team! There are so many opportunities for Twitter to leverage & add on to web3 – especially in the creator space around monetization, ownership, identity, community, and more.
Twitter building out a crypto team. Probably nothing.
I’m thrilled to share that I’ve joined Twitter, to lead a new team focused on crypto, blockchains, and other decentralized technologies—including and going beyond cryptocurrencies.✨
Super excited to welcome @_tessr to lead our Crypto engineering team. More about this team 👇
I’m thrilled to share that I’ve joined Twitter, to lead a new team focused on crypto, blockchains, and other decentralized technologies—including and going beyond cryptocurrencies.✨
Parag Agrawal reposted
Proud of all the teams at Twitter who have been working hard to increase our product development velocity. We’ve been busy! Here’s a recap of a bunch of things we’ve shipped or started experimenting with just in the past 2 weeks:
Parag Agrawal reposted
Today we’re introducing @SuperFollows, a way for the people driving the conversations on Twitter to earn money from their most engaged followers on Twitter 🧵 blog.twitter.com/en_us/topics/product/2021/introducing-su…
Parag Agrawal reposted
All systems go! Help shape the future of Twitter Spaces with the new Spaces endpoints. ✨
blog.twitter.com/developer/en_us/topics/tools/2021/shape-…
I’m thrilled to share that @arcalinea will be the @bluesky lead.
The ideal candidate is someone who is hands-on technically, curious across disciplines, has experience working in the open on the blockchain, an opinionated vision for this effort, and is thrilled by the idea of shaping the future of social media.
The team and community that Jay will bring together can shape the future of social media through decentralization using open standards. Join her👇
If you’re interested in joining @bluesky, we have open roles on our website blueskyweb.org. Or join the community on @Discord - we’re excited to open up this discussion space to everyone. discord.gg/YmrHBhCuFa
🙏 to everyone that participated. Amazing to see so much creative energy help build a better understanding around algorithms and their impact on people.
Drumroll please! Our first-ever algorithmic bias bounty at #defcon wrapped this weekend, and the winners were announced at yesterday’s #defcon @aivillage_dc workshop. ICYMI, you can watch a recording of the presentation here youtu.be/KLwdzLiJywI
Parag Agrawal reposted
Finally let my son “splurge” on something my immigrant parents could never bring themselves to allow me to do with money: deposited $0.51 into a machine to get back only $0.01, which has itself lost its value in the name of becoming a ‘souvenir’ 😅
Parag Agrawal reposted
Responsible ML isn't a team, it's a company-wide imperative. With META as the driver, we're excited to share our new RML initiative. 1/5
Today we’re introducing the Responsible ML initiative, our effort to understand the use and impact of machine learning at Twitter, and take action when needed.
We’ve been working on this for a while. Here’s how we see the path forward:
blog.twitter.com/en_us/topics/company/2021/introducing-re…
Parag Agrawal reposted
Thank you Dr @Bob_Wachter for your extraordinary leadership over the past year — can’t even count the number of times @paraga and I referred to your tweets when assessing a situation, making choices for our family, or simply deciding how to feel at any particular moment in time
23/ …social media can be brutal, I’ve found Tweeting over the past year to be massively gratifying. And, while I hope my tweets have helped people make sense of an incredibly confusing and ever-changing landscape, I’ve also learned a ton in putting them together.
Parag Agrawal reposted
.@williams_jutta is guiding force on META, looking forward to this panel!
Excited to partner with @ruchowdh to accelerate our work on Responsible ML and push into exciting new directions around providing people more choice and control over algorithms.
Big news! - I’m thrilled to be joining the @TwitterEng team today as Director of ML Ethics, Transparency & Accountability. With the META team, we’ll work to improve ML transparency, inclusivity and accountability. 1/
Parag Agrawal reposted
1/ I’m excited to share that we are partnering with the @BigHatBio team @MarkDePristo @peytongreenside, and leading their Series A financing! BigHat is building a platform for the future of biologic drug design, with tremendous potential for patient impact a16z.com/2021/02/11/bighat-biosciences/
We’ve seen a lot of benefit from our partnership with @GCPcloud and are now doubling down for offline data and ML workloads.
I'm thrilled to be greatly expanding our partnership with @Twitter on their Data and Machine Learning applications! @paraga @nicktornow @derekalyon
We can’t 'like' this enough. We’re partnering with @Twitter to move its offline analytics, data processing, and #ml workloads to Google Cloud. Learn more, via @TechCrunch ↓ techcrunch.com/2021/02/04/twitter-expands-google-cloud-pa…
More 👇
We’re excited to be expanding our strategic partnership with @googlecloud. This will enable more teams at Twitter to gain richer insights, faster, about how people use our service so we can improve the product & recommend more relevant content. www.prnewswire.com/news-releases/twitter-expands-strategi…
For those interested in learning about various efforts around developing protocols for social media.
Working virtually through the pandemic, this group self-organized, invited additional experts, and created a review of the ecosystem around protocols for social media. You can read it here: matrix.org/_matrix/media/r0/download/twitter.modular.im/9…
👇
I do not celebrate or feel pride in our having to ban @realDonaldTrump from Twitter, or how we got here. After a clear warning we’d take this action, we made a decision with the best information we had based on threats to physical safety both on and off Twitter. Was this correct?
Parag Agrawal reposted
1/ Everyone's talking about video visits. But are they really the greatest gift of digital health? I think they’re a red herring. Just a new medium for the same old episodic care. REAL change will come as we transition to systems of "Always-On Triage"
a16z.com/2021/01/08/always-on-triage/
The linked blog details the analysis that led to our decision.
Yes! Thrilled to strengthen our partnership with @awscloud and continue working towards an even better, lower-latency experience for people using Twitter all around the world!
Really excited about working w/ @paraga & the whole @Twitter team to help power the real-time, global traffic of this vibrant platform w/ @AWScloud services & #infrastructure. #aws #reinvent www.zdnet.com/article/twitter-taps-aws-cloud-to-deliver-t…
Parag Agrawal reposted
In honor of the first wave of RNA vaccine dosing this week, we brushed up on the #centraldogma
RNA is awesome 👏🏼
Thx @narayan_pri for these amazing (and timely!) books, and thanks @cara_florance for creating this series! @littleansh might just be a budding baby biochemist
Parag Agrawal reposted
Communication that is inspiring
Communication that is convincing
Communication that is simplifying
...all the communication
Also communication that is so frequent and predictable that the content ironically starts to feel ‘known’ and expected (learned this from @zachweinberg)
When you think of a Product Manager who's great at execution, what skills come to mind?
“one person, working alone in their bedroom, was able to build a capability which would allow them to seriously compromise iPhone users they'd come into close contact with.” The specific issue has been fixed, but crazy to imagine the implications.
Excited to finally publish my lockdown project from earlier this year: an iOS zero-click radio proximity exploit odyssey.
googleprojectzero.blogspot.com/2020/12/an-ios-zero-click-…
Fleets enable serendipity in a way that reminds me of the way I first experienced Twitter many years ago - in just the past few hours, they have been an ice breaker to facilitate DM conversations with people I might not have connected with otherwise!
📣 Starting today, Fleets will be rolling out globally for everybody on Twitter! 🧵
Parag Agrawal reposted
Thank you to @paraga and @littleansh for today's #storytime! They read Binny's Diwali, a beautiful story about the Indian festival of lights by @ThrityUmrigar kids.scholastic.com/kids/book/binnys-diwali-by-thrity-umr… #happydiwali #diwali
Really great paper that describes a commonly seen problem with ML models in production.
NEW from a big collaboration at Google: Underspecification Presents Challenges for Credibility in Modern Machine Learning
Explores a common failure mode when applying ML to real-world problems. 🧵 1/14
arxiv.org/abs/2011.03395
Have always loved @Radiolab — works even better as a Twitter thread.
The US election is tomorrow. If you, like me, are tired of horse-race-style reporting, and need to zoom out, I wanna tell you a story. It’s about an ancient force influencing the election. And, as a bonus, it’ll give you an Easter egg to watch for as the returns come in. THREAD
A cohesive set of changes that we expect will lead to a more deliberate and thoughtful conversation on Twitter. And we will learn a lot by observing ways in which the service changes as a result.
🧵 Thread 👇🏽
As we head into #election2020, this morning we announced some important updates to our policy and to the product experience which you can read about in detail here: blog.twitter.com/official/en_us/topics/company/2020/2020-…
Welcome to the team @rinkisethi!
A lot of the work our teams do to protect Twitter happens behind the scenes. blog.twitter.com/en_us/topics/company/2020/our-continued-…
Awesome work, fascinating results, great summary thread 💯
We have a new paper on learned optimizers! We used thousands of tasks (and a lot of compute 😬) to train general purpose learned optimizers that perform well on never-before-seen tasks, and can even train new versions of themselves.
arxiv.org/abs/2009.11243
1/8
“The acceleration of our pipeline was critical in our ability to quickly perform exploratory data analysis (EDA) and led to discovering a range of effective features used in the final solution”
Winning the 2020 @recsyschallenge challenge wasn't easy. @kaggle grandmasters, fast #gpu, crafty feature engineering in @rapidsai, and @XGBoostProject made it possible. Now it's all in @nvidia #nvtabular! nvda.ws/360d4iD
“Their friendship was built on a bedrock foundation of shared values, and chiefly, I think, the value of a sober, neutral court ... These were guardians of a sacred institution, and they believed that institution must stand beyond politics.” Also, cc @jack
the friendship between ginsburg and scalia wasn't a story of civil disagreement. on highest principles they agreed, and friendship without such agreement is impossible. i wrote about shared values, tech utopianism, and (evergreen) the twitter apocalypse. solana.substack.com/p/more-important-than-votes
This is a very important question. To address it, we did analysis on our model when we shipped it, but needs continuous improvement.
Love this public, open, and rigorous test — and eager to learn from this.
Parag Agrawal reposted
Hearing from many teams (incl my own) that being forced to remotely onboard new employees has actually made the process *better*.
Lots of things that [extroverted] ppl used to randomly learn in hallway convos are now codified, written, repeatable...and more widely accessible 👏🏼
Another step that enables people to have more ownership over conversations they start.
Sometimes people are more comfortable talking about what’s happening when they can choose who can reply. Today, we’re launching conversations settings so unwanted replies don’t get in the way of meaningful conversations.
Amazing execution on an incredibly challenging re-architecture of our ads platform. I helped build the previous system, and couldn't be happier that our decisions from 6-8 years ago no longer hinder progress. Huge thanks to the team that worked on this.📈blog.twitter.com/engineering/en_us/topics/infrastructure/…
Parag Agrawal reposted
If you aren’t following this list during the watch party, you are missing out. #hamilfilm @HamiltonMusical x.com/i/lists/1278834808559296513
Parag Agrawal reposted
Parag Agrawal reposted
Huge congrats to @jessicatysu and team! And thanks @twitter for encouraging this work! @krishna_kamath @aneeshs @5harad
🏅🏅🏅
For #icwsm2020's best paper, congratulations to Jessica Su, Krishna Kamath, Aneesh Sharma, Johan Ugander, and Sharad Goel for "An Experimental Study of Structural Diversity in Social Networks"
aaai.org/ojs/index.php/ICWSM/article/view/7332
Parag Agrawal reposted
Both Twitter and Square are making #juneteenth (June 19th) a company holiday in the US, forevermore. A day for celebration, education, and connection.
www.juneteenth.com/history.htm
Super excited to learn from this test. And how it works together with easily discovering all retweets with comments.
To experience what happened 50 years ago, follow along
T-3 hours, 30 minutes and counting. The Apollo 13 countdown has now resumed. There are no more holds planned before launch.
#apollo13
#apollo50
Define leadership
I’m moving $1B of my Square equity (~28% of my wealth) to #startsmall LLC to fund global COVID-19 relief. After we disarm this pandemic, the focus will shift to girl’s health and education, and UBI. It will operate transparently, all flows tracked here: docs.google.com/spreadsheets/d/1-eGxq2mMoEGwgSpNVL5j2sa6T…
Parag Agrawal reposted
Not all quarantine memes are created equal.
Thrilled to welcome @InteranaBobby and the rest of @InteranaCorp team to Twitter!
We are excited to bring their expertise to help us better understand how people use Twitter so we can deliver better products and features, faster.
Parag Agrawal reposted
"Wuhan Virus", "Chinese Virus", COVID-19. Doesn't matter the name - my hometown will forever be known for that and only that. I drew this comic to shine light on what people don't know: the beautiful culture, rich history, and strong people of Wuhan. 加油!
www.lauragao.com/wuhan
Parag Agrawal reposted
New massive recommender system dataset from Twitter! We hope this will stimulate more research on recommender systems which didn't really have datasets of this size.
We are releasing the biggest ever (160 million samples) recommender system public dataset today with @recsyschallenge 2020. Please go check out the website: recsys-twitter.com
Really excited to see how people use this product so that we can improve it and bring it other markets.
Every day, people come to Twitter to see what’s happening. One of the unique things about Twitter is that “what’s happening” is fueled by people sharing their thoughts openly, through Tweets. But sharing your thoughts publicly can be intimidating! 🧵
Parag Agrawal reposted
I wrote a blog post that details Saturday's finding of "cryptic transmission" of #sarscov2 in Washington State. Genomic evidence suggests that #covid19 has been circulating here since Jan 15, but undetected due to lack of testing. bedford.io/blog/ncov-cryptic-transmission/
A Developer Labs endpoint that helps us learn in many new dimensions. I am most excited to see how this API empowers people by enabling choice among ML algorithms.
In November, we gave people the ability to hide replies to their Tweets. Starting today, we’re opening this feature up to developers!
✨Introducing a new Labs endpoint so developers can help people hide replies they find irrelevant, off-topic, or toxic. twittercommunity.com/t/introducing-the-hide-replies-endpo…
Parag Agrawal reposted
You’re better off in the long term, at least to me, thinking about the long-term prospects of your ideas vs the short-term prospects of your team.
🙏 for helping set this up for the team
Changing lives one bag at a time. Shout out to the amazing team that prepped 5,000 bags in 2.5 hours. @RMHChou #oneteam #twitterforgood
Thank you for your leadership!
Writing cards, assembling care packages for @RMHChou @TwitterForGood #oneteam!
Parag Agrawal reposted
4,400 tweeps serving for 4 hours each today at 51 charities and organizations throughout Houston. 1,200 at the @HoustonFoodBank alone! #oneteam
🧵
What do you find most difficult about working with engineers? I’m at Twitter #oneteam in Houston, Texas, asking product managers, designers and other non-engineering folk this question. Here’s what they have to say👇
Parag Agrawal reposted
Is it even a #oneteam if @boo doesn’t breakdance??? Live from the EPDR Summit! #oneteam2020 #epdr
Prioritization is hard! #oneteam2020 #epdrbreakouts
It’s only getting harder.
Parag Agrawal reposted
Sooooo if you follow any folks who work at Twitter, get ready for 3 days of tweets from our all-company offsite (sorry or you’re welcome?!)
The tech behind the scenes at #oneteam2020
Traveling for work is always a reminder for me - it can be hard. And we don’t acknowledge our families and support structures enough - thanks to all who are traveling to Houston, and thanks to all who are staying behind to hold down the fort at home! #oneteam2020
🧵 about SIM swaps. What about @googlefi (with and without advanced protection)?
Unfortunately, all five carriers used authentication methods that are considered insecure in the computer security community. Taken together, these findings help explain why SIM swaps have been such a persistent problem. More details in our paper: www.issms2fasecure.com/assets/sim_swaps-01-10-2020.pdf
Recommend reading this great summary.
What I did over my winter break!
It gives me great pleasure to share this summary of some of our work
in 2019, on behalf of all my colleagues at @GoogleAI & @GoogleHealth.
ai.googleblog.com/2020/01/google-research-looking-back-at…
Parag Agrawal reposted
Thrilled to be joining this amazing team! Look forward to working with founders who will change the future of healthcare and make it work better for patients 👊🏼
I'm thrilled to announce that Vineeta Agarwala will be joining @a16z as our newest General Partner! @vintweeta has had many different roles in the healthcare space: as a physician, as an operator at tech-driven healthcare companies, and as an investor. a16z.com/2020/01/07/vineeta-agarwala/
Parag Agrawal reposted
First hike of the new decade: Slacker Hill! An undeserved view after a super short climb
I’m incredibly excited for Twitter to kick off @bluesky, a new independent effort to develop a decentralized standard for social media. Please see @jack’s thread for more context. I have the privilege of finding a lead for this team.
The ideal candidate is someone who is hands-on technically, curious across disciplines, has experience working in the open on the blockchain, an opinionated vision for this effort, and is thrilled by the idea of shaping the future of social media.
3 - The traditionally slow and deliberate consensus-building approach to evolving standards might fail to keep up with a rapidly changing ecosystem and set of consumer needs.
Despite these hurdles, we believe that there is the potential for massive positive impact for Twitter and for society. If you agree, and are as excited to confront these challenges as we are, please reach out by DMing @bluesky. I look forward to meeting you and learning from you.
👇
Today, we’re sharing the full details of Twitter's new political ads policy. I encourage you to read through it for the full detail, but I wanted to share some of the thinking that went into its creation. This new policy goes into effect on 11/22.
business.twitter.com/en/help/ads-policies/prohibited-cont…
👋 Look forward to coming back to 🇳🇬 🇬🇭
Parag Agrawal reposted
Help us shape Twitter's policy on synthetic and manipulated media. Feedback due by 11/27. Details in the blog! blog.twitter.com/en_us/topics/company/2019/synthetic_mani…
Thank you for inviting us to 🇳🇬 and to your home @NOIweala #sheinspiresme
🤯 What a conversation! Bitcoin + 🇳🇬 = 🔥
This was a lot of fun — very heated at times but bullish on the future of bitcoin and crypto on the continent @timigod @iaboyeji @chrismaurice @MunachiOgueke @adetolaov @paraga @jack
Re: this @leslieberland thread, I can confirm all the faces of @kayvz are a whole mood.
Even when he isn't trying. 😂
Grateful to @ethereumJoseph for joining us at @Twitter for a lively conversation. We are passionate about working with people and communities everywhere to make our future internet enable more trust, fairness, and access.
Thanks to @paraga & @Twitter for inviting me to their beautiful campus for a great dialogue about decentralization and the evolution of the World Wide Web.
We’re all building the internet together and it takes the leaders of Web2 exploring at the edges to move us all forward.
Parag Agrawal reposted
Well, shoot. My kid’s first interpretable word is definitely “Alexa”
Machine 1, Mom 0
Parag Agrawal reposted
Today in Twitter Developer Labs we’re releasing a new way to filter Tweets in real-time. This is one of the most popular features of our API. It allows developers to listen to hashtags, @mentions, and study what’s happening.
twittercommunity.com/t/new-filtered-stream-endpoints-avai…
Parag Agrawal reposted
What a way to end the week: We just received a generous donation of 40 laptops from @Twitter! This will definitely go a long way for our students who are working hard to break into the tech industry. 👩🏻💻👨🏽💻👨🏾💻👩🏾💻
Thank you for your support, Twitter! 🙏🏽🙏🏾🙏🏿
Parag Agrawal reposted
I wrote a Twitter blog post!! Excited to share some of the ways that the new Twitter web platform makes it a delight to iterate on.
The new x.com responds to more than just your screen size. Read more about how we brought our PWA to desktop in the next episode of our blog series.
blog.twitter.com/engineering/en_us/topics/infrastructure/…
👇
Today we're launching the new twitter.com! So many inspiring people worked on this 2-year project, and we learned so much. I want to share our learnings with the world, so I asked my teammates: What's the most valuable thing you learned while building the new site? 👇
Parag Agrawal reposted
1/6 We’re very excited to announce that, next week, we will be launching a test in Canada 🇨🇦that allows people to hide replies to their Tweets.
You asked for more control over your conversations, so starting next week we’re testing a new feature in Canada that will let you hide replies to your Tweets.
For transparency, viewers everywhere can see hidden replies by going to a new icon or the dropdown menu.
Great write up by @prasadwagle describing the challenges involved in enabling batch and interactive analysis over many petabytes of data. Our Data Platform team is hiring! blog.twitter.com/engineering/en_us/topics/infrastructure/…
Credit for making this happen goes to three incredible leaders: @KCPacini, @lientm and @d_lux_brand. Grateful that I get to work with you.
Starting today, we’re accepting applications for our first cohort of engineering apprentices! We know that great talent comes from all kinds of backgrounds, and we need to do a better job of creating and fostering a sustainable pipeline of diverse talent.
blog.twitter.com/en_us/topics/company/2019/engapprentices…
Parag Agrawal reposted
The below cases are all real patients I have seen in the span of just *two weeks*. The full patient stories are heartbreaking.
Parag Agrawal reposted
Reading some @NASAhistory, and was curious who this was...
Have been subscribed to The New Paper for a week, and find myself looking for it in my email most mornings. Really quick to consume, and does feel factual -- would love to see them on Twitter. www.thenewpaper.co/r?r=joo2b3d23b
Constraining each rule to 280 characters does create a lot of clarity. The fact that the team was able to do this reminds me of the research we did that led us to pick 280 as the increased character limit. blog.twitter.com/engineering/en_us/topics/insights/2017/O…
We refreshed Twitter's rules so they are easier to understand and to make clearer what's not acceptable on Twitter and why. From 2500 words to under 600! Take a look and let us know what you think...
We are very excited to welcome @mmbronstein and the @fabula_ai team to Twitter!
📣 We're excited to announce @Fabula_AI is joining Twitter! They'll bring their machine learning talent to help us grow our research team and graph deep learning technology to improve the health of the conversation on Twitter. Welcome to the flock! blog.twitter.com/en_us/topics/company/2019/Twitter-acquir…
Deep learning has been very successful in modeling images, video, and speech, all of which contain an underlying Euclidean structure. However, a wide range of problems at Twitter benefit immensely from the ability to model and learn from non-Euclidean graph structured data.
Geometric deep learning aims to bring key ingredients of deep learning methods such as convolutional neural networks to non-Euclidean data, and the Fabula team has pioneered important work in this area.
The Fabula team’s expertise in geometric deep learning is applicable to many problems at Twitter. The team will initially focus on improving the health of the public conversation.
I am personally thrilled to have the opportunity to work closely with @mmbronstein, @frederickmonti, and the @fabula_ai team. Welcome to the flock, Fabula!
Parag Agrawal reposted
Based on the observation that the GPT-2 medium-size model has memorized (and can spit back word-for-word) very long extracts from the web, such as the Gorilla Warfare meme, I had an idea for a very simple ML-less text generation algorithm. I spent the past 20 min implementing it.
Parag Agrawal reposted
We made a subtle change that has made images on ultra-retina devices load 33% faster and save upwards of 35% on data with no perceptible change in quality. Check out our blog!
blog.twitter.com/engineering/en_us/topics/infrastructure/…
Some great work from the #twitterjvm team showing effectiveness of JVM tuning using Bayesian optimization and Graal
Corrected link x.com/christhalinger/status/1128874143921737728
We are looking for your feedback as we build the next generation of the Twitter API. Want to make sure it will work for those who use it the most.
📣We’re launching Twitter Developer Labs in the coming weeks so we can build the future of our API together with those who use it most. Want to learn more about how you can join us? 👉blog.twitter.com/developer/en_us/topics/tools/2019/buildi…
"Our time series metric ingestion service grew to more than 5 billion metrics per minute, stores 1.5 petabytes of logical time series data, and handles 25K query requests per minute."
Check out how our Observability team has improved latency, cost, and more 👇blog.twitter.com/engineering/en_us/topics/infrastructure/…
Parag Agrawal reposted
Looking for your help to learn the best treatment options incl clinical trials for FLT3 negative #aml in primary induction failure.
Appreciate input, anecdotes, suggestions, references. All info helps, thx much @KNOW_AML @LLSusa @CureLeukaemia @LeukemiaMDA @leadleukaemia @LeukUK
This thread just made my day
Parag Agrawal reposted
Thanks Singapore! We’re outta here! #tweeptour
Parag Agrawal reposted
This stop on the #tweeptour wouldn’t be complete without #twitterforgood. We helped package hundreds of meals for 5000 families in need. The @WillingHeartsSG lead said it best — “You all at Twitter are in the business of growing, but we hope one day we can shut down” 💙@TwitterSG
OH: I'd eat durian over chocolate any day.
Parag Agrawal reposted
This is the biggest, hardest, and most important project I have ever been involved with - it’s central to the future of the company. Come join @jianhangg and team if you like REALLY hard technical challenges and want to work on a once-in-a-lifetime project
👇🏼
Parag Agrawal reposted
My PhD advisor Jennifer Widom taught me how to learn an area of scientific research to be able to make original contributions. And gave me the skills to write and present technical content. But more importantly, she set me on a path of continuous learning. #sheinspiresme
👇🏼
In honor of #internationalwomansday tomorrow, I'd like to start a thread about women who have inspired us, taught us, and mentored us in our careers. Please comment with someone who has inspired you! (cont)
In honor of #internationalwomansday tomorrow, I'd like to start a thread about women who have inspired us, taught us, and mentored us in our careers. Please comment with someone who has inspired you! (cont)
My PhD advisor Jennifer Widom taught me how to learn an area of scientific research to be able to make original contributions. And gave me the skills to write and present technical content. But more importantly, she set me on a path of continuous learning. #sheinspiresme
👏🏼👏🏼👏🏼 @yz @arvind__t @timssweeney and the entire team that worked on this
Learn more about the #poweredbytf ML model that ranks the tweets in your timeline! The Cortex and Home Timeline teams at @Twitter describe re-tooling their platform for TensorFlow in this guest post. #tfdevsummit
Read it here ↓ medium.com/tensorflow/ranking-tweets-with-tensorflow-932d…
Find and replace doesn't work for this use case. Need to build some technology.
Parag Agrawal reposted
Want to use our prototype version of Twitter and help give us early feedback? See below!
Parag Agrawal reposted
Harvey is legit world class! On many levels!
Parag Agrawal reposted
I’d like you to know that the final message we played for Opportunity was a song.
We sang her to sleep with Billie Holiday’s “I’ll Be Seeing You.” It ends with:
I’ll find you
In the morning sun
And when the night is new.
I’ll be looking at the moon,
But I’ll be seeing you.
Look forward to the insights and understanding that this collaboration will bring to Twitter and the broader ML community. I’m confident that our systems and products will improve as a result.
We’re excited to reveal a new partnership between Twitter and UC Berkeley:
a new lab, lead by @mrtz and @beenwrekt, dedicated to understanding and improving how ML systems work inside social systems.
blog.twitter.com/en_us/topics/company/2019/ucberkeley-twi…
Parag Agrawal reposted
Is this thing on?
I have neck control, a double chin, and this cool @twitter onesie now, so figure it’s time to start tweeting! I’ll be sharing my candid thoughts via @littleansh #helloworld
Ansh was born on Thanksgiving. His name means “a part of” - a part of something bigger, a part of our family, a part of our community. I’m so grateful for all the support @vintweeta and I have gotten from our Twitter family over the last many weeks. #oneteam
Parag Agrawal reposted
Terrible. All of our SF Twitter and Square folks are encouraged to stay indoors tomorrow. Work from home (or come in if you feel air quality is better in office).
The air quality in San Francisco is now 'very unhealthy' x.com/i/events/1063220965562171392
Parag Agrawal reposted
Hey IIT Delhi, here's your opportunity to ask @jack your questions! Tweet with #powerof18
Parag Agrawal reposted
You voted, and the winner is in! Tweet in all the languages below to see the deep or diya emoji appear!
Have a wonderful Diwali!
Parag Agrawal reposted
#happydiwali tweeps!
Parag Agrawal reposted
November is here, and that means a massive shift is coming. And by "massive" I am of course referring to the redefinition of the kilogram unit of mass that the world has been building up to for more than 100 years. Let me explain:
Parag Agrawal reposted
We want to make it easier to toggle between seeing the latest tweets the top tweets. So we’re experimenting with making this a top-level switch rather than buried in the settings. Feedback welcome.. what do you think?
So great to meet the incredibly collaborative team at @TwitterDublin
Delighted to welcome our CTO @paraga to @TwitterDublin - now in conversation with @tpogculture
My heart rate watching the hearing while in bed over the last hour
Really great meeting our @TwitterUK team over the last couple of days. Can see why this is the birthplace of #lovewhereyouwork
Parag Agrawal reposted
What else would you call this?
Parag Agrawal reposted
Thank you @chefjoseandres ❤️ #oneteam
Excited to attend the NYC @TwitterEng talks and look forward to answering your questions.
Want to know what it's like to work in engineering at @Twitter? Join us here on 9/12 at 9:30am ET from @TwitterNYC to hear from members of the team! Submit your questions using #whatshappeningnyc.
👏🏼
Congrats to GV's @byersblake and @vintweeta, recognized by Business Insider among the top Silicon Valley biotech investors under 40 — www.businessinsider.com/silicon-valley-life-science-biote…
Parag Agrawal reposted
Something I've been working on for a while. Happy to share it more widely. Gargantuan effort. This is a big milestone, but we are just getting started. So proud of the team. Send feedback and let us know what you think!
Parag Agrawal reposted
Had a great time judging the @TechCrunch Startup Battlefield session this evening alongside @msquinn @SarahRCannon @bobkocher @peterpham!
Easy for us to ask questions, much harder to pitch. Major kudos + thx to all the founders who are sharing ideas and demos at #tcdisrupt!💡
Here are all the Startup Battlefield companies competing for $100K and the renowned Disrupt Cup #tcdisrupt tcrn.ch/2wIhK9F
💯
I joined Twitter on September 3, 2013. Today marks my fifth anniversary. I wanted to take time today to share some lessons I have learned, usually the hard way, that have help me along the way
This one is a winner!
Parag Agrawal reposted
If you're a member of a large, complex, or highly tribal software engineering organization, then this thread is for you.
Let's talk a little about living technical documentation for software systems.
1/n
Parag Agrawal reposted
Twitter Lite is now available in the Play Store in 21 more countries! 🌏🌍🌎
“It’s clear that teams using Strato enjoy increased leverage: it saves them considerable effort, so enables them to deliver features more quickly, and experiment at lower risk”
paper on a system I've been working on for 3+ years was accepted to Scala Symposium! drive.google.com/open?id=1aYupExDuAbUheDX4aycrxZDrc6stTcig with @ramenlabs @marius @erikfroese @greenberg @mjeffryes @THISWILLWORK @objmag1c @rmgn @BenNavetta @slyphon @msol @joyousx
OH: we need those pants on the Goldbird team
Parag Agrawal reposted
Oh my god
Parag Agrawal reposted
“For most problems you encounter someone else has encountered a similar problem and has probably built a solution that you can re-use, modify or be inspired by. So try to be lazy and make use of those solutions.” - @paraga, CTO @twitter
Full house for @igb and @ysr1729 as they discuss auto-tuning JVMs using ML at #oneteam
Great audience questions during @LukeEvansSimon’s talk about ads prediction.
I have so much FOMO for to all the talks I missed. Thanks to @blackmad, @lessachu, @pandemona, @enunge, @kathleencodes, @mkruz, @billdarrow and all the presenters for putting this together. #oneteam
Parag Agrawal reposted
You go back in time and give yourself one piece of advice about engineering. What is it?🤔 I’m putting this question to @Twitter engineers from around the world who are in San Francisco for #oneteam this week. Here’s what they have to say👇
👇
Parag Agrawal reposted
Twitter is migrating 300 petabytes of data to @GCPcloud. Twitter's CTO @paraga shares the migration experience & his team's work.
More on how to get the most out of your databases for your business → cloudplatform.googleblog.com/2018/07/on-gcp-your-database…
Very insightful discussion.
Enjoyed participating in the chat session with @eric_lander and @ericschmidt at #googlenext2018 to discuss healthcare and machine learning yesterday! The video is now up.
www.youtube.com/watch?v=N9mdn3nI3HM&start=28m30s
Parag Agrawal reposted
A short thread addressing some issues folks are encountering as a result of our conversational health work, specifically the perception of “shadowbanning” based on content or ideology. It suffices to say we have a lot more work to do to earn people’s trust on how we work.
To be clear, our behavioral ranking doesn’t make judgements based on political views or the substance of tweets. We recently publicly testified to Congress on this topic judiciary.house.gov/wp-content/uploads/2018/07/Pickles-Te…
Parag Agrawal reposted
Twitter CTO on the #googlenext18 stage talking about how network performance allowed them to decouple storage and compute. Not requiring data-local computation with large data sets gives a lot of flexibility!
Welcome to the team @nicktornow! Thrilled to have the opportunity work together.
Welcoming @NickTornow on day1⃣! He will join our engineering leadership team as our new Platform Lead.
Parag Agrawal reposted
1/ Pleased to welcome @NOIweala and @authorzoellick to the @twitter Board! www.prnewswire.com/news-releases/twitter-appoints-ngozi-o…
Parag Agrawal reposted
We've heard some questions recently about the status of Verification on Twitter, so wanted to address directly. Updating our verification program isn’t a top priority for us right now (election integrity is). Here’s some history & context, and how we plan to put it on our roadmap
Parag Agrawal reposted
Fascinating:
www.nytimes.com/2018/07/11/sports/world-cup/harry-kane-en…
Research in endurance sports like cycling and running has shown that athletes can gain a performance boost during intense bouts of exercise by rinsing their mouths with a carbohydrate solution, then spitting it out without swallowing.
I believe most of Twitter’s impact is in the future. Really excited to build technology with all our incredible engineering, product, design, and research teams to bring about positive change. #lovewhereyouwork
Look forward to working with the @HelloSmyte team on improving the health of the public conversation on Twitter.
Really proud of the team, and excited about the impact their work will have on changing Twitter!
So proud of the team #twittercortex !! Great blog post by @cibelemh and @strife076 about Twitter moving its production deep learning platform onto TensorFlow, along with the challenges and adaptations associated to it: blog.twitter.com/engineering/en_us/topics/insights/2018/t…
PS: we're hiring!
Really good starting point for fostering a conversation. Lots of work needed across the community around bias, transparency, accountability, and privacy.
Today we’re sharing our AI principles and practices. How AI is developed and used will have a significant impact on society for many years to come. We feel a deep responsibility to get this right. blog.google/topics/ai/ai-principles/
AI does need values, and being explicit about them publicly can be powerful. cc/ @naz_erkan, @fhuszar, @sandeep_pndy, @nkoumchatzky
“Artificial Intelligence Needs Values. Here Are Ours!” medium.com/huggingface/artificial-intelligence-needs-valu…
Parag Agrawal reposted
Update on @Twitter’s request for Health metrics proposals. We received great submissions and appreciate all the effort from those involved…1/
Thank you everyone who submitted @Twitter Health proposals! 230+ proposals from institutions globally across topics: echo chambers, bots/misinformation, healthy discourse, info analyses & more. We'll be reviewing over the coming weeks & will follow up with semi-finalists then!
Glad Uber and Lyft agree that this problem is best solved together. Also great to see Twitter play a part in fostering this collaboration.
A meaningful step towards improving the health of public conversation. Really proud of our teams that collaborated to make this happen.
I should not have said we didn’t have to share. I have felt strongly that we should. My mistake.
We are sharing this information to help people make an informed decision about their account security. We didn’t have to, but believe it’s the right thing to do.
We are sharing this information to help people make an informed decision about their account security. We didn’t have to, but believe it’s the right thing to do.
I’m sorry that this happened, but am proud to work at a company that puts people who use our service first.
Parag Agrawal reposted
Nice shoutout to @christhalinger and the #twittervmteam in @forbes today for our work with the Graal JIT compiler. www.forbes.com/sites/oracle/2018/04/18/graalvm-1-0-gives-…
Always energizing to spend time @TwitterBoston
Parag Agrawal reposted
Intellectually, it feels collectively like having Nobel laureates in all disciplines hanging out in the tea room just down the hall. I can pop in whenever I like, hear them noodling, occasionally contribute myself.
Parag Agrawal reposted
6) Note that the trace is also the sum of the eigenvalues. also, the sum of the eigenvalues of A^3 is the sum of eigenvals of A to power 3. SO the third moment of the eigenvalues of A is the number of triangles in a graph.
(this is my most favorite fact in all of everything)
Solid party favors. Happy birthday @Twitter #lovetwitter
Parag Agrawal reposted
As someone in healthcare, I am always grateful for the opportunity to work with mission-driven people and companies. I’ve come to appreciate that Twitter has a *deeply* mission driven team. And what a mission they are on!
Parag Agrawal reposted
Some of the cool people of Infrastructure Engineering are presenting at ☕️ time.
Parag Agrawal reposted
Always great to hear when people notice these subtle design details we’ve added to the Twitter app. @dbedingfield @_dml
The haptics in Twitter around the character limit are SO GOOD. You can keep typing (so annoying when forms/apps block you 😡) yet vibrations let you know as you approach and hit the limit. Great use of colors, numbers, and animation too.
Kudos to whoever designed that!! 👍🏻🙏🏻
If you care about improving the health of public conversation, it doesn’t matter if you work on frontend or hardware engineering or anything in between. You can contribute. careers.twitter.com/content/careers-twitter/en/jobs-searc… #jointheflock
I have wanted this feature for so long!
Parag Agrawal reposted
Yesterday we took action against a number of accounts that we believe have violated our policies by creating new accounts after being suspended. This is a part of the work we're doing every day to improve Twitter.
Parag Agrawal reposted
Like @jack, we think we can do much more. Come help us build a kickass AI platform at #twittercortex
Such a privilege to be inspired by @jack and to learn from this amazing team everyday.
It’s a big day for Twitter. While his humility usually precludes us from sending tweets like this, I'm doing it anyway:
On behalf of Tweeps everywhere, thank you @jack for your leadership. We're fortunate and deeply grateful for your incredible impact on the company, and us!
Parag Agrawal reposted
20
Parag Agrawal reposted
Example of crazy systems failure that occurs when hospital #ehr systems are fragmented. A real patient story. (1/n)
Cutting-edge technology applied at scale to solve real user problems. #lovewhereyouwork
We have developed a new ML model to improve smart auto-cropping of images on Twitter. Learn more: blog.twitter.com/engineering/en_us/topics/infrastructure/…
Parag Agrawal reposted
👋 from Twitter HQ! We made an animated map so you can show your students how it spread around the world:
twitter-maps.carto.com/u/twitterdata/builder/60ed65be-8fc…
Hard to not be proud of this list of accomplishments!
On behalf of the Google Brain team, I wrote a summary of some of the things we did in 2017. This is part 1 of 2. I'm very proud of the whole team and what we accomplished in 2017! research.googleblog.com/2018/01/the-google-brain-team-loo…
Parag Agrawal reposted
Twitter is very nice some days
Parag Agrawal reposted
Introducing the CVPR 2018 Learned Image Compression Challenge, a collaboration between Google, ETH and Twitter! 🚀 www.compression.cc
In order to advance the state-of-the-art in image compression with #machinelearning, we are co-sponsoring the Workshop and Challenge on Learned Image Compression (CLIC) at #cvpr2018 in June. We hope you'll enter, and may the tiniest win! research.google/blog/introducing-the-cvpr-2018-learned-im…
And off he goes. Thanks for the food delivery @StarshipRobots 🤖@DoorDash
Good morning NYC
Parag Agrawal reposted
Parag Agrawal reposted
Launching tweetstorms/threads today! Thank you to everyone who helped create and refine this format on Twitter. 🙏🏼 blog.twitter.com/official/en_us/topics/product/2017/nicet…
/cc @alkispolyzotis
Machine Learning just ate Algorithms in one large bite, thx to @tim_kraska, @alexbeutel, @edchi, @JeffDean & Polyzotis at @Google—faster, smaller trees, hashes, bloom filters www.arxiv-vanity.com/papers/1712.01208v1/
1/ I’m thrilled to announce this lineup of amazing women in tech who will be hitting the Twitter stage during CES on Jan 10th! We’ll also livestream on Twitter via @twitterwomen. #hereweare #changetheratio
Parag Agrawal reposted
Really fascinating use of multi-dimensional #ehr data to help guide early utilization of palliative care. Thank you for sharing @AndrewYNg @DrSteph_H! It was eye-opening to be the recipient of a few of these predictions for our patients during the pilot @StanfordMed
Here's our full paper on improving palliative care with deep learning. Neural net with 18 hidden layers inputs EHR, estimates chance of mortality in next 3-12 months. Also generates report to explain predictions to clinicians. arxiv.org/abs/1711.06402
A thoughtful and nuanced contribution to the debate. Worth a read for anyone following this discussion.
Alchemy, Rigour and Engineering
my opinion on @alirahimi19's talk, @ylecun's response, and the ensuing discussion
www.inference.vc/my-thoughts-on-alchemy/
Parag Agrawal reposted
In September, we tested Twitter Lite in the Play Store in the Philippines. We received a lot of valuable feedback from people, and made some refinements. Starting today, we're making it available in 24 more countries! 🌏
Parag Agrawal reposted
Check out our poster on "Meta-learning for cold start recommendations" at NIPS on Mon, 12/4. Poster #72 nips.cc/Conferences/2017/Schedule?showEvent=9457 @hugo_larochelle @arvind__t
Parag Agrawal reposted
Today we’re launching Twitter premium APIs to kick off the next era of innovation for Twitter developers. 🙌
These APIs provide a new transparent, predictable way to scale solutions on Twitter. blog.twitter.com/developer/en_us/topics/tools/2017/introd…
Thank you to Japan 🇯🇵 for showing us the way blog.twitter.com/engineering/en_us/topics/insights/2017/O…
Parag Agrawal reposted
Engineering analysis:
Testing the limit! Here’s how we measured the impact of longer Tweets via experimentation: blog.twitter.com/engineering/en_us/topics/insights/2017/E…
Parag Agrawal reposted
The man of the hour: thank you @nabokov7 for your leadership and for showing us every day that passion and personality matter #thankyouiku
Parag Agrawal reposted
Thanks @ProjectOpenHand for having us today! #twitterforgood
Thanks to all the Tweeps who volunteered with us on #fridayforgood at @ProjectOpenHand! #twitterforgood
Seems very useful for model explorations and research. Does Eager Execution mostly negate the edge PyTorch has had for such use cases?
Introducing Eager Execution - an imperative, define-by-run interface for #tensorflow!
Learn more here → developers.googleblog.com/eager-execution-an-imperative-d…
How the universe celebrates #halloween
1/ #happyhalloween2017 everyone! I'll be celebrating by posting scary, ghastly, ghostly, and downright crazy things we find in the #universe
Parag Agrawal reposted
The Mechanics of History by Yoann Bourgeois. Performed at the Pantheon (via @fguzfguz)
Parag Agrawal reposted
Here’s our safety release calendar: blog.twitter.com/official/en_us/topics/company/2017/safet…
On this day in Central Park, @vintweeta said yes. #anniversary
Parag Agrawal reposted
Name another company that matters that has ever done this with important parts of its product.
Parag Agrawal reposted
We have a Trust & Safety Council which reviews and gives feedback on our policy and product changes. We emailed them our proposed changes in order to get critical feedback. Email found in this article: www.wired.com/story/here-are-twitters-latest-rules-for-fi…
And👇
Summary of #gravitationalwaves #ligo news:
* LIGO/Virgo detected neutron stars colliding
* Also seen from event: light at many wavelengths
👇
Parag Agrawal reposted
👇🏼 thread. We need to be a lot more transparent in our actions in order to build trust.
Parag Agrawal reposted
ICYMI: Tensorflow & PyTorch Design Tradeoffs (ML meets general purpose programming)
medium.com/@kovasb/ml-framework-design-reliability-compos…
Parag Agrawal reposted
This is interesting. Seems to imply something about determinism too.
We are not living in a simulation, FYI. cosmosmagazine.com/physics/physicists-find-we-re-not-livi…
2017
Parag Agrawal reposted
This is a small change, but a big move for us. 140 was an arbitrary choice based on the 160 character SMS limit. Proud of how thoughtful the team has been in solving a real problem people have when trying to tweet. And at the same time maintaining our brevity, speed, and essence!
T-5 hours now. As good a time as any to catch up on the adventures of @CassiniSaturn
T-6 days until @CassiniSaturn burns up in the upper atmosphere of #saturn. It's has flown for 13 years, and will take pics to #grandfinale
Parag Agrawal reposted
Larger pupils may have powered the evolutionary leap from oceans to land:
www.quantamagazine.org/why-did-life-move-to-land-for-the-…
Best of both worlds! Really important to enable rapid experimentation without trading off the ability to operate in production easily.
Facebook and Microsoft introduce ONNX, a new, open ecosystem for interchangeable AI frameworks... www.facebook.com/login/?next=https%3A%2F%2Fresearch.faceb…
Parag Agrawal reposted
A tribute to his grandmother turning 100 years old, Chinese artist builds 'sky ladder' firework. x.com/SadHappyAmazing/status/648510422908207104/video/1 via @ValaAfshar
Parag Agrawal reposted
This #dreamliner has been drawing a giant #dreamliner for 13 hours now!
Live link:
www.flightradar24.com/BOE004/e54ad6b
Parag Agrawal reposted
Twitter is getting a lot better at blocking accounts of people/bots who threaten violence... so report anything abusive folks
Parag Agrawal reposted
Parag Agrawal reposted
It used to be that reporters wrote stories & then Tweeted them out. Now we Tweet & then convert it into a story: www.nytimes.com/interactive/2017/07/11/world/middleeast/w…
Parag Agrawal reposted
Confirmed from an NYT source: "We were preparing to publish" story -- Don Jr. camp "asked for more time" to comment "and then pre-empted us"
Parag Agrawal reposted
Technical details on our Deep Learning+ECG (detecting irregular heartbeats/arrhythmia) work: stanfordmlgroup.github.io/projects/ecg/
Parag Agrawal reposted
Is it just me or is Twitter's algorithmic timeline so much better these days?
Parag Agrawal reposted
Twitter algorithm on point af
Took the long way home - back in SF!
EWR-BOM-BKK-CNX-HKT-SIN-BOM-SEZ-JNB-LLW-JNB-CPT-JNB-IST-NCE, MXP-MUC-SFO
www.gcmap.com/mapui?P=EWR-BOM-BKK-CNX-HKT-SIN-BOM-SEZ-JNB…
Best part about traveling with @DJIGlobal #mavicpro: making friends with geeky TSA agents worldwide
What a joy to be back @Inria catching up with my undergrad research mentor Frederic Cazals. His intellectual curiosity is always inspiring!
Dinner topics included topology of biomolecules, graphs, flu vaccines, and of course skiing :) Still as awed by Frederic as I was 13 yrs ago
Parag Agrawal reposted
Hello! I made you a mix!
This is
What Twitter Feels Like: open.spotify.com/user/tobillo24601/playlist/0Ayyn9Nl4y8Rg…
Parag Agrawal reposted
As a candidate, Trump criticized the Clinton Foundation for taking money from Middle East despots. But these donations are different.
Saudis And The UAE Will Donate $100 Million To A Fund Inspired By Ivanka Trump www.npr.org/sections/thetwo-way/2017/05/21/529417148/saud…
Best part is that it's headed to Taco Bell
Parag Agrawal reposted
Had a great talk with @jack dorsey -- got to @twitter changes, harassment and @realDonaldTrump. backchannel.com/jack-dorsey-on-donald-trump-9864d2e542f4
Amazed by how accurate @googlemaps estimation of driving times are in Mumbai even w/ crazy traffic. We're talking w/in 3min on a 2hr ride!
It's so cool that Finland bestows this responsibility on its scientists in such a direct way #marchforscience #defendthetruth
In Finland, PhDs are awarded sword as a symbol to defend what's right and true. My sword stays at home today, but I #marchforscience
Parag Agrawal reposted
2 unit tests. 0 integration tests.
Parag Agrawal reposted
What a convo. I'll type it up quick.
Live tweet of a phone call with the president! Fastest way to get the news.
President Trump just called me. Still on phone.
"We just pulled it," he tells me.
⚡️ “The 'Fearless Girl' takes on Wall Street”
x.com/i/moments/839195816283762688
Congrats @StanfordEng! Jennifer was an incredible PhD advisor. She continues to inspire me, alongside many others: medium.com/@profwidom/from-mooc-to-moic-launching-an-inst…
Jennifer Widom, professor of computer science and electrical engineering, has been named dean of @StanfordEng. stanford.io/2l5RfrM
Parag Agrawal reposted
Unexpected highlights of ski wknd in Aspen @Snowmass: après ski gluwein w @kcoleman and Ice Age mastodon bones (found here in 2010)!
Parag Agrawal reposted
.@tfadell @sacca @ACLU Thank you for your initiative, encouraged @paraga and I to double down. Let's do this.
Sport at its very best.
Proud
Parag Agrawal reposted
Subway train poetry. Captures what those who currently feel "welcome" in America must remember. We're ALL immigrants #neverforget #resist
Parag Agrawal reposted
Solo girl crushes it at a team programming contest. So proud of my little sister! pclassic.org/2016f #girlsruntheworld
cc @badass_w
Parag Agrawal reposted
Go Barca! #elclasico @FCBarcelona
Churros for breakfast, anyone?
Making dinner reservations and came across this google translate gem. Guess I'll dine with the Mrs @vintweeta as a Lord tonight
Parag Agrawal reposted
Go team!
Parag Agrawal reposted
Best 10mi motivational sign ever! #montereyhalf #lovetrumpshate #girlsruntheworld
This was insane! #whiteout @PennStateFball
Bokeh gets old fast when it's all you can see. #pacificatemyglasses
Giant Ceibo Tree www.instagram.com/accounts/login/?next=https%3A%2F%2Fwww.…
Grateful that I got this one right 5 yrs ago! #twitterversary
Looks like I might enjoy this job. #ballmerpeak @ Twitter www.instagram.com/accounts/login/?next=https%3A%2F%2Fwww.…
Parag Agrawal reposted
I know i've put on weight but i am not 400lbs. 😤
#debates2016
Parag Agrawal reposted
"In Presidential Debates, there is really only one winner: Twitter"
-- Wired www.wired.com/2016/09/first-debate-best-tweets/
Parag Agrawal reposted
My story on the debate: www.thestar.com/news/world/uselection/2016/09/26/trump-an…
Parag Agrawal reposted
Priors confirmed, now back to work everyone
Parag Agrawal reposted
This is the top Retweeted Tweet during the first US presidential debate:
The concept of global warming was created by and for the Chinese in order to make U.S. manufacturing non-competitive.
Parag Agrawal reposted
This @npr annotated transcript is fantastic. www.npr.org/2016/09/26/495115346/fact-check-first-preside…
Parag Agrawal reposted
Take 2 minutes and help someone register to vote. #debatenight www.vote.org
Parag Agrawal reposted
U.S. Citizens, Q4 OKR: vote.
That's it. Your only obligation.
Parag Agrawal reposted
The only winner tonight is the Voyager probe, which is speeding away from the Earth at 17 kilometers/second #debatenight
Parag Agrawal reposted
This is a debate between a politician and your uncle's Facebook page.
Parag Agrawal reposted
.@BorowitzReport: CNN Launches Manhunt After Lester Holt Vanishes From Debate nyer.cm/ybxUKFo
Parag Agrawal reposted
Watch @bpolitics' special coverage of the first debate. We're streaming it LIVE on @Twitter #debates debates.twitter.com
Parag Agrawal reposted
2.2M people watched #tnf (2.1M last week). Average viewers up 34% to 327k. I think solid given blowout $TWTR
www.recode.net/2016/9/23/13033836/how-many-watched-nfl-tw…
Parag Agrawal reposted
This is pretty much the best parse error ever.
Parag Agrawal reposted
The @HoustonTexans.
The @Patriots.
It's LIVE on @Twitter!
💻📱: tnf.twitter.com
📺: @NFLonCBS #tnf #houvsne
Parag Agrawal reposted
Parag Agrawal reposted
Parag Agrawal reposted
Unbelievable! @CMU_Football shocks #22 @CowboyFB 30-27 with a ridiculous play to end the game! snpy.tv/2cNI8ss
Parag Agrawal reposted
The life of a back home baller -@paraga @Stanford #gostanford #farmphoto
Great to be able to talk to the NYC ML Meetup group about all the fun work we are doing here at @TwitterNYC
Parag Agrawal reposted
This is my new favorite comic.
Parag Agrawal reposted
Hey guys, I downloaded #periscope this morning! Recommended.
Parag Agrawal reposted
Remember how oppressed minorities use Twitter to get the word out? #nobillnobreak
The cameras may be off, but Democrats are using Twitter to broadcast their gun control sit-in on the House floor www.wired.com/2016/06/house-dems-use-twitter-make-gun-con…
Parag Agrawal reposted
Now C-SPAN is showing a periscope of sit-in from the House floor - never seen anything like this.
Parag Agrawal reposted
35,000 have signed up for Machine Learning Yearning book. Join us by Jun 24th! mlyearning.org
Parag Agrawal reposted
Parag Agrawal reposted
Some tips on finding good PMs :-)
firstround.com/review/find-vet-and-close-the-best-product…
Parag Agrawal reposted
Easily the most stunning places I've ever visited! Geology is crushing it @BryceCanyonNPS @ZionNPS #findyourpark
Parag Agrawal reposted
One way to bridge the 2-body problem: @tweethue. While sitting in SF, @paraga is messing w/ light colors around me in NYC - thx @ZhiTsun!
National Park Week starts 4/16! Gotta plan the next #parineeta2016 trip with @vintweeta this week #findyourpark
FIFA World Cup challenged us to build adaptive quality factor -- and now it foundational to our ad serving system.
Twitter’s ad server — our revenue engine — performs ad matching, scoring & serving at an immense scale. Learn more: blog.twitter.com/2016/resilient-ad-serving-at-twitter-scale
Parag Agrawal reposted
"We're seeing stars" Totality is here! www.nasa.gov/ntv #eclipse2016 #solareclipse #eclipse
⚡️ “Basketball's game of the year wasn't in the NBA or college”
x.com/i/moments/705832108745318400
One more arch for good measure by @vintweeta
Lunch at the other @birdfeeder
@birdfeeder snowbird ski resort
Parag Agrawal reposted
Linux kernel bug delivers corrupt TCP/IP data to Mesos, Kubernetes, and Docker containers /cc @epcjones medium.com/vijay-pandurangan/linux-kernel-bug-delivers-co…
Timeline Personalization
It's time for champagne
Parag Agrawal reposted
Tweets you missed from people you follow, recapped in reverse chron! Refresh back to live. Flip it on to try it now!
#parineeta2016 in 30 secs (thanks @googlephotos)
I have never had a full time job without @aroetter in my reporting chain. Thanks for all you have done to grow me along with Twitter.
Five years ago, talking to @kevinweil was what convinced me to come work at Twitter. Thanks for that and all your leadership since then.
Selfie on a horse #parineeta2016
Haldi! #parineeta2016
That time we were immobilized by #mehndi #parineeta2016
Morning run with @vintweeta to kick off #parineeta2016
Heading to Jaipur for the start of #parineeta2016 (cc @vintweeta)
I have seen orders of magnitude growth on many dimensions at @TwitterSF over the last 4 years. Hope for the same at @TwitterNYC.
Today is my last day at the SF office. I am really excited about moving to NYC.
Parag Agrawal reposted
When you fix a bug in production.
Parag Agrawal reposted
How Elon Musk and other tech giants plan to stop computers from taking over medium.com/backchannel/how-elon-musk-and-y-combinator-pla… Thanks to @evanatmedium
Parag Agrawal reposted
We're launching @open_ai, a non-profit AI research company: openai.com/blog/introducing-openai/.
Got disconnected after being on hold for 20 minutes to get customer service. DM to @AirCanada solved issue in one minute.
Clouds below, blue skies above
Vacation auto-responder ON
Parag Agrawal reposted
When you get engaged to a @twitter man, you must celebrate on his medium :)
She said yes! @vintweeta
She said yes! @vintweeta
#seethemomentwhen you see the human side of news x.com/i/moments/651424481815212032
Congratulations to @justmadhu @vijayp @blackmad and the entire team #twitternyc x.com/justmadhu/status/651382440167145472
Tesla Model X Launch
Will this poll get 200 votes?
Parag Agrawal reposted
The #twitterkdd team #lovewhereyouwork #kdd2015
Join us for the Twitter party at Sydney's secret rooftop bar @TheRookSydney after the poster session at #kdd2015
Parag Agrawal reposted
#kdd2015 Keynote:Online Controlled Experiments: Lessons from Running A/B/n Tests for 12 Years. www.exp-platform.com/Pages/KDD2015KeynoteExPKohavi.aspx
Parag Agrawal reposted
Come meet our @TwitterEng team working in the #ml & data science space: twitterkdd.eventbrite.com #kdd2015 #twitterkdd
Parag Agrawal reposted
@paraga, @iam_karam, @yue, @SonnyWhy and @ZJYIN tweeting from the Twitter booth at #kdd2015 #twitterkdd
Come to @TheRookSydney to meet the Twitter team working in the #ml & data science space. RSVP: www.eventbrite.com/e/twitterkdd-social-hour-tickets-18009… #kdd2015
Need to work on my geography. Couldn't recognize a destination for an Air France flight out of LAX
Parag Agrawal reposted
The ethics of experiments vs intervening without a control group mobile.nytimes.com/2015/06/21/opinion/sunday/please-corpo…
Parag Agrawal reposted
On the left is the best image we had of Pluto before. Now we have this. Just astonishing. #plutoflyby
Best set ever. #federervsdjokovic
Roger
Who wins?
Parag Agrawal reposted
The NBA emoji war on Twitter yesterday was amazing. See the best of it in one place here: x.com/twittermedia/timelines/618914838467600384
Parag Agrawal reposted
Best photo of the day: the running of the media interns to relate news of Supreme Court rulings. (note shoes)
Parag Agrawal reposted
Data science meets xiao long bao -- and I'm not talking about counting the days to DTF: www.smartshanghai.com/articles/dining/the-man-who-spent-a… @mattocko @nehanarkhede
Maintain my caffeine sensitivity so that I can use it when I need to stay out on a Friday night.
Parag Agrawal reposted
When my mom discovers I'm Snoop Dogg's nephew.
What a conversation between @levie and @SnoopDogg
Parag Agrawal reposted
Let's get this started, @dickc. #costolo2016
Parag Agrawal reposted
Thanks to a certain CEO for adding "cadence", "vis-a-vis", "in service to" and "got it" to my lexicon.
Parag Agrawal reposted
Twitter conversations are now easier to read!
blog.twitter.com/2015/making-conversations-easier-to-foll…
Congrats @akarshan_ @yipe @DSGNRSteve @samuelmaskell @justinwang!
The mighty @KarlTheFog taking over again.
Parag Agrawal reposted
For the second straight game we have OT! This #nbafinals!
Parag Agrawal reposted
You have prejudices. I have priors.
An earthquake on Mt. Kinabalu www.bbc.com/news/world-asia-33017758
Parag Agrawal reposted
For the 1st time ever in #sf history, May was colder than Jan., Feb., March, & April... #cawx
Parag Agrawal reposted
Fun seeing @fredwilson comment on our "While you were away" feature. avc.com/2015/06/while-you-were-away/
Props to the team @ptr @paraga @allidryer!
I am nervous starting this episode of @GameOfThrones — suspect that something bad will happen.
Phew!
Parag Agrawal reposted
In @Melbourne at #sigmod15 & want to learn more about @TwitterEng? Swing by our booth! #twittersigmod
Parag Agrawal reposted
@kellyshalk, @megsraye and @goldeng8pridge G'day from #sigmod15 in Melbourne! #sigmod15 #twittersigmod
nice branding @larsonite
Glad to read about what @rkakkad83 is up to with @englishdost. This is great! yourstory.com/2015/04/english-dost-app/
Showed up at work without my laptop. #achievementunlocked
From Hoare's Turing Award Lecture www.cs.fsu.edu/~engelen/courses/COP4610/hoare.pdf
Parag Agrawal reposted
1. Go to @periscopetv
2. Follow
3. Click gear, select "turn on device notifications"
4. Never miss the best Periscopes
lets launch control2
Parag Agrawal reposted
1/ Just published over @LinkedIn - Who’s the best seed money in India? linkd.in/1N7kD5L
"… hit the center. The center of an Asprin tablet is exactly the same size as the center of a beach ball." www.youtube.com/watch?v=Q8Yp9SjCU5E
Goodbye Bombay! www.instagram.com/accounts/login/?next=https%3A%2F%2Fwww.…
Parag Agrawal reposted
How Twitter's ad team is using neuroscience to find out what your tweeting brain looks like medium.com/backchannel/this-is-your-brain-on-twitter-cac0…
Just witnessed the effectiveness of a threat to tweet about bad customer service at a restaurant in Bombay #twitterindia
Parag Agrawal reposted
My time at @Nest was incredible - an amazing team. But now I’m excited to join the flock at @Twitter as VP of Technology and Analytics.
Parag Agrawal reposted
Starting today, we syndicate our business model in the same way we syndicate our content, w/ same great native ads: blog.twitter.com/2015/promote-content-on-and-off-twitter-…
Parag Agrawal reposted
Pizza and planning @TwitterIndia. #friday #oneteam #watchoutworld
Related Searches
Parag Agrawal reposted
See a unique look at the green #lovejoy Comet ti.me/1GpuWBy via @TIME Photo: Dave Lane
Parag Agrawal reposted
Breaking: 4 shot dead in San Francisco's Hayes Valley... All victims were male adults. No suspects in custody. cbsloc.al/1A25F7G
Good morning SF #nofilter www.instagram.com/accounts/login/?next=https%3A%2F%2Fwww.…
Parag Agrawal reposted
Parag Agrawal reposted
The Sony Pictures hack: Everything we know so far feedproxy.google.com/~r/TheNextWeb/~3/_DpW4fYbScI/
Goodbye
Lots of concurrent work to describe images using text. Anyone know which paper is the most readable? news.ycombinator.com/item?id=8621917
Another amazing OK Go video: shot using a drone half a mile up and japanese schoolgirls moving in perfect synchrony www.billboard.com/articles/news/6296760/ok-go-i-wont-let-…
Parag Agrawal reposted
So this is happening outside of Whole Foods at Market and Dolores.
Parag Agrawal reposted
I'm happy to announce we've reached an agreement with Twitter to give them twitpic.com - more here: blog.twitpic.com/2014/10/twitpics-future/
Loved reading this. storify.com/WildlifeFilms/lion-drama-at-great-plains-seli…
Parag Agrawal reposted
Time and money spent on meaningful choices is commonly associated with a greater sense of well-being in the long run: stanford.io/1wsbwmz
Does anyone have a Google inbox invite for me?
Just composed a poem! #cannonballapp #nature
Now installing: Cannonball, the magnetic poetry game appsto.re/us/7WTA3.i
Parag Agrawal reposted
Twitter Fabric feels similar to what AMZN did w/ AWS. Open up platform capabilities they already run at massive scale. Cool #twitterflight
Parag Agrawal reposted
"On @android it's even easier. How many times have you heard that sentence?" /cc @jeffseibert @Flight
Parag Agrawal reposted
Announcing Fabric: The easiest way to make the best apps.
Parag Agrawal reposted
Almost Go time. #twitterflight
Reading about the broad questions around news.sciencemag.org/biology/2014/10/updated-u-s-halts-fun…. Fascinating!
Parag Agrawal reposted
"Tell me about Twitter, Dad." "Well son, we would piece together the news of the day by reading strangers' decontextified jokes about it."
If you just saw the Blue Angels, watch the view from inside the cockpit youtu.be/u4D0yx4DvBk
Parag Agrawal reposted
been in India for last 2 wks. seriously impressed by @TwitterIndia promoted tweets. great targeting and creative. going to miss these ads!!
Have had a great 3 years. Thanks @utkarsh for getting me into the flock.
Looks like I might enjoy this job. #ballmerpeak @ Twitter www.instagram.com/accounts/login/?next=https%3A%2F%2Fwww.…
2nd of 3 sunsets for me on the 29th! www.instagram.com/accounts/login/?next=https%3A%2F%2Fwww.…
2 days in a jungle www.instagram.com/accounts/login/?next=https%3A%2F%2Fwww.…
Parag Agrawal reposted
Today's XKCD. Yes.
Parag Agrawal reposted
India successfully orbits Mars -and their first tweets with NASA's Curiosity Rover are perfect fw.to/mpFNgQW
Parag Agrawal reposted
Howdy @MarsCuriosity ? Keep in touch. I'll be around.
Tropical rainforest and caves.
Snorkeling and curry laksa to close out Sabah. On to Sarawak tomorrow. www.instagram.com/accounts/login/?next=https%3A%2F%2Fwww.…
monkeys, rain, sunset, terrible food, fireflies, beer #borneo www.instagram.com/accounts/login/?next=https%3A%2F%2Fwww.…
Sunset grilling www.instagram.com/accounts/login/?next=https%3A%2F%2Fwww.…
Parag Agrawal reposted
Where does your industry fall on our 2x2 matrix of uncertainty? s.hbr.org/1tDyMME
Walk at sunrise #thankstojetlag
Kopi in Singapore is a dangerous drink. On 3rd cup so I can enjoy that last sip full of condensed milk.
Parag Agrawal reposted
The inexorable march of Ivy League grade inflation.
That thing where the flight gets delayed by 10 minutes every 10 minutes.
Almost hoping this continues so that I will be able to watch Stanford beat USC before I board. #gocardinal
Parag Agrawal reposted
I am here to lose my verified badge;) @paraga @vyas instagram.com/p/smRUp3A0aA/
Parag Agrawal reposted
Parag Agrawal reposted
this is an eye opening comparison between NY and SF income, however, I think they should compare SF Bay Area to NYC www.slate.com/blogs/moneybox/2014/08/29/income_distributi…
Parag Agrawal reposted
ICYMI: How the #napaquake affected sleep in the Bay Area. See more data on the blog: www.jawbone.com/blog/napa-earthquake-effect-on-sleep/
What would you get?
Exit row FTW!
Amazing story. Stood out how short-term direct impact on reality can fulfill, inspire and reboot a career. x.com/echen/status/501500578615160832
Parag Agrawal reposted
Last Week Tonight with John Oliver: Ferguson, MO and Police Militarization youtu.be/KUdHIatS36A
Parag Agrawal reposted
‘Hatching Twitter’ season three, episode three: as the deploy queue grows, a "beard-off" between two senior engineers gets out of hand
Idealistic, but can it really work in practice and scale? open.bufferapp.com/introducing-open-salaries-at-buffer-in…
Machine predicts heart attacks 4 hours before doctors www.newscientist.com/article/mg22329814.400-machine-predi…
Check out the live lunar flyby of a vintage spacecraft at 1030am pacific. spacecraftforall.com
Parag Agrawal reposted
Expertise www.youtube.com/watch?v=dbHvMajApvw
Looks stunning
Here's the first #film trailer. www.aviationtheinvisiblehighway.com
Imagining PowerPoint animations to explain my own thoughts to myself.
I was the shortest among 6 guys in an elevator today.
Parag Agrawal reposted
It's a big day for mobile app advertising and Twitter. We've agreed to acquire TapCommerce! blog.twitter.com/2014/welcoming-tapcommerce-to-the-flock
Super excited to be launching mobile app promotion blog.twitter.com/2014/introducing-mobile-app-promotion
Looks like I might enjoy this job. #ballmerpeak @ Twitter www.instagram.com/accounts/login/?next=https%3A%2F%2Fwww.…