Jeffrey Wang
@jeffzwang
Followers
17K
Following
5K
Media
103
Statuses
2K
cofounder @exaailabs
San Francisco
Joined January 2020
We raised $85M to continue building the best search engine for AIs. Why a search engine for AIs? LLMs are trained on the internet but they can’t memorize it - they need search. But traditional search engines were built for humans. LLMs are now ingrained in thousands of popular
We raised $85M in Series B funding at a $700M valuation, led by Benchmark. Exa is a research lab building the search engine for AI.
61
23
482
There are a LOT more papers that need to be desk rejected at ICLR. Somehow, this hallucination wasn't caught. So, I went on a crazy rabbit hole and found 50 more (many are just as funny). We're approaching crisis levels Paper titles included below 👇 https://t.co/PeTjJAbkrH
This paper has been desk rejected. LLM-generated papers that hallucinate references and do not report LLM usage will be desk rejected per ICLR policy ( https://t.co/HcsxTDAVwl) Reviewers of other versions of this submission have been notified.
8
14
149
OpenAI and Google should provide limited usage of GPT 5.1 Pro and Deep Think on their $20/month plans, not because I'm a cheapskate (although I am), but because I'm pretty sure this would cause mass adoption of their premium subscriptions. Almost no one realized just how GOATed
81
14
528
It boggles my mind that Google is paywalling Deep Think at $250/month (Ultra plan) Market cap: $4T Gemini market share: ~15% cool bro
Gemini 3 Deep Think is here. Deep Think is our most advanced reasoning mode that explores multiple hypotheses simultaneously to give you an even more sophisticated output.
28
4
270
In this world, nothing can be certain except: - death - taxes - ✨ EXA'S ANNUAL NEURIPS PAPER SEARCH✨
We embedded all 5000+ NeurIPS papers! https://t.co/QoKAw0oORx Cool queries: - "new retrieval techniques" - "the paper that elon would love most" - "intersection of coding agents and biology, poster session 5" It uses our in-house model trained for precise semantic retrieval 😌
2
0
36
https://t.co/OfSPHCMSKT Discovered this super interesting article from @iamgrigorev thanks to @ExaAILabs's new search engine. Very nice
ggrigorev.me
Byte-level BPE from first principles: what matters for speed and quality, how to implement it cleanly, and why a SuperBPE variant can lift sample efficiency.
2
2
19
🔥
2
1
61
Exa is now my default search engine! For the longest time, Exa wasn’t general enough or fast enough for daily driving. But we’ve dramatically improved our index, algorithm, and latency. What happens with simple searches? We detect that and return you useful information as fast
Smarter than your default search engine, faster than your default chat app Try the new https://t.co/cQ6UlWHnKY
23
16
379
Hint: the event is called SOTA on a BOATA
Exa and Benchmark are hosting a special event at NeurIPS Thursday evening. DM me your spicy takes on semantic retrieval if you want to come aboard ⛴️
1
0
33
I used to run an erotica writing eval across new frontier models, but now they all refuse except for Grok 😭
10
0
74
I wish I could work at each AI lab for like a week - go in, fix like 5 janky things, then leave
33
2
278
wonder how the current AI timeline would be different if SBF was still in the game
4
0
33
💯
we've added both a web search and code search tool to opencode powered by exa - it just works nothing to setup now opencode can lookup examples of any library you're using
2
3
37
Google dominated the search algorithm for so long that we've all been trained to think search can't get much better. Sure, the interface to search has changed (LLMs), but the underlying search algorithm has not. There are many reasons for this, mostly financial incentive
Introducing Exa 2.1 We scaled our pre-training and test-time compute by an order of magnitude, unlocking frontier search API performance for both fast and agentic search. Deep dive below:
12
10
240