
nitya's notes + building a search engine @exaailabs
San Francisco, CA
Joined December 2020
- Tweets2.8K
- Following783
- Followers7.3K
- Likes10.5K
had a long conversation with a flow cytometrist in Boston on how she thinks AI will most impact science
my working conclusions are essentially a bullish case for Period Labs + Medra, the emergence of AI discovery maps to the ability and pace of experimental validation
4
150
Great work by Nikhil + team!
Exa Snapshot allows you to search the web as if you were in the past and it was used here for their llm forecasting research
FutureSim Q3 is here!
Opus 5.5 dethrones GPT, displaying strong priors and incredible test-time adaptation.
Opus 5.5 is a beast, calling 50+ subagents and taking over 400 research notes during the 90 day simulation while also being 50% more token-efficient than 6.1 Sol. (1/5)
20
1,304
how do you all focus while working with both short and long running agents effectively?
there are periods you are just waiting for the agents to all work but you can't really focus on anything else since there is always a thread which is ready with a response
the framing of "ambient work" feels right
teapawt.net/poolnoodle93/status/21…
my theory for this is that we are now entering the zone of “ambient work” due to the increasing capabilities of long horizon agents. you’ll be working all the time it will just be broken up by real world commitments and work is always happening in the background.
3
420
funniest thing about navier strokes drama that no one mentioned is that
Strokes and Navier also came up with their ideas independently and thought it was distinctive enough to be released separately, mirroring the language in the OAI blog weirdly similarly in spirit
Nitya retweeted
Introducing Pomerado (YC F26): the faster, cheaper, and more reliable alternative to browser use.
Open source. Sign up for free to supercharge your agents 🧵
this is incredible
Our newest book is finally here!
Making the Modern Laboratory is a spectacular 504-page hardcover that tells the story of everything in the research laboratory and how it came to be, from pipettes to heat-resistant glass to microscopes and more.
It features dozens of articles (most of which will never appear in print), 100+ full-color images, and original writing from 36 authors. It's printed on Arena Natural Rough paper from Fedrigoni. These books are THICC and would make a great gift for both the science-curious and the working scientist. :)
We have spent the last year working on this project. It has been an insane labor of love. The best way to support our independent science journalism is to buy a copy. <3
Pre-order for $49, with free shipping: press.asimov.com/articles/mo…
16
1,590
signing up for a race is a competitive sport now someone give me your Copenhagen half bib pls
3
6
1,682
Nitya retweeted
TIL the onboarding for macOS 9 had a whole section explaining what the internet is and how it works
a lot of erratic behavior by smart people on twitter can be explained by studying scientific heretics
often when I am scrolling this app I think about Brian Josephson, who at 22 proved that electric currents can flow between two superconductors if separated by a thin layer of insulating material. He was early to something that was correct and rewarded the highest prize for it - Nobel for physics.
He then spent the rest of his career studying things like telepathy, refusing to change what he worked on. Being early and right, makes it hard to believe people when they tell you you're wrong later on.
6
1,301
Product team at Exa is hiring! come work with me
the work is just really interesting - building time versioned search of the web, productionizing research across our retrieval team, driving search quality, API design and there is still so much more to do
Exa just passed $100m in annualized revenue.
We've grown almost exactly 10x per year for 3 years. We now handle 2 billion requests a month from 200k+ teams. And it’s still so early:
We predict agents will search 1000x more than humans by 2030.
Agents don't search like humans. They write long, complex queries and want all the world's data, not just the web. We've spent 5 years building a search engine designed for them.
We now serve personal agents, AI labs, coding agents, GTM agents, really any agent. Exa gives these agents access to 100s of billions of web documents and tons of proprietary data.
Our customers are increasingly noticing that information is the bottleneck to agent performance. While the world seems close to perfect intelligence with models like Fable and Astra, we are very far from perfect information.
To our small but legendary team, thank you for getting us here, from engineering to research to GTM to ops.
To our customers, it is an honor to serve you. We have some exciting releases coming soon and can't wait to share them 🫡
evals are also distinct across tasks in ways you might not expect
retrieval beyond parametric memory requires specific focus on recency, tasks that are consistently changing - it is an important part of how I think about product quality. Wrote about this + open sourced our code evals a while back, linked below
Evals have been coming up more and more in my conversations with podcast guests and PM friends.
Nearly half of the 25 awesome PM job openings I shared last week ask for experience writing evals. And leading companies keep sharing what investing in evals bought them:
— @tryramp took its automatic receipt collection from 35% to 83% accuracy.
— @Shopify shipped an AI workflow builder that's 2.2x faster and 68% cheaper than the frontier-model setup it replaced.
— @harvey__ai rebuilt its AI contract reviewer, nearly doubling its internal quality score.
— @cursor_ai tuned its Auto Balance routing, with much higher user satisfaction at 41% lower cost.
So I asked the 🐐s of evals, @HamelHusain and @sh_reya, to write an advanced sequel to their very popular "Building eval systems that improve your AI product." Drawing on their work with 50+ AI companies, they share the key step most teams skip, what you should (and shouldn't) automate, and a free plugin that lets a coding agent do most of the heavy lifting.
Read it here: lennysnewsletter.com/p/advan…
2
21
1,363
manual dogfooding -> golden set of evals that comprehensively capture a task
exa.ai/blog/webcode
2
298

































