Ex

Exa (YC S21) – embeddings search agent with >20x recall than Google

Hacker News

Exa (YC S21) – embeddings search agent with >20x recall than Google

Hey HN! I'm Will, founder of Exa (YC21 - https://exa.ai ). Today we're opening up Websets, a search engine that finds massive lists of correct results given complex queries. For example, you can search for: - “software engineers in the Bay Area, with experience in startups and big tech, who know Rust and have published technical content”: https://websets.exa.ai/cm7ax5bl8003276q689v0dde5 - “US based healthcare companies, with over 100 employees and a technical founder": https://websets.exa.ai/cm6lc0dlk004ilecmzej76qx2 - "research paper about ways to avoid the O(n^2) attention problem in transformers, where one of the first author's first name starts with "A","B", "S", or"T", and it was written between 2018 and 2022”: https://websets.exa.ai/cm7dpml8c001ylnymum4sp11h We built Websets because the web is humanity’s grand collection of all knowledge, and yet it’s totally unorganized. So much valuable content is too hard to find. Traditional search engines, like Google, were built to handle simple keyword queries over the web, not arbitrarily complex SQL. While agentic tools like Deep Research help a bit, they rely on traditional search under the hood and so are similarly bottlenecked. Websets works well because under the hood it uses our in-house embedding-based search engine, trained specifically to handle complex natural language queries. Crucially, Websets uses LLMs to agentically verify each result to ensure correctness. Websets is therefore a test-time compute search engine – it might take minutes or even hours to run. We believe this is a worthwhile sacrifice for high value searches. It’s hard to eval these things, but we did our best and measured that Websets found 20x more results than Google on a set of complex queries and 10x more than Deep Research. These numbers could be arbitrarily higher with more compute per query. Blog post here: https://exa.ai/blog/websets-evals While Websets isn’t perfect search yet, it’s a significant first step, and we’re excited to share the first version of the product with you all. We have a free limited tier, and we set up a special HN code for Pro plans. Use PERFECTSEARCH for a two-week free trial of Pro. Can try it here: websets.exa.ai Initial launch video here: https://x.com/ExaAILabs/status/1864013080944062567 Would love to hear thoughts, feedback, and suggestions! I know HN thinks about search sometimes :)

Share card

Actual performance

28points
4comments
Made the leaderboard

Launch Intel predictions

Analyze your own launch →
Product HuntOn track for Day 1 leaderboard · Strong signals: agent, agentic, google · Missing: mac, agents, macos
89%89% predicted probability of success on Product Hunt, based on ML models trained on real launch data.
best fitHighest predicted score across all platforms for this description.
Indie HackersFits the IH revenue-focused audience · Missing: supports, reddit linkedin, podcasting
85%85% predicted probability of success on Indie Hackers, based on ML models trained on real launch data.
Hacker NewsStrong engagement from HN community · Strong signals: excited, lua, ide · Missing: https docs, just released, exist
54%54% predicted probability of success on Hacker News, based on ML models trained on real launch data.
nativeThis product was originally launched on this platform.
AppSumoMay struggle as an AppSumo deal · Missing: plus, platform, intuitive
43%43% predicted probability of success on AppSumo, based on ML models trained on real launch data.
TrustMRRLess likely to generate early MRR · Strong signals: video, google, way · Missing: mobile apps, ios, personal
28%28% predicted probability of success on TrustMRR, based on ML models trained on real launch data.
Acquire.comPre-revenue stage for this audience · Missing: arr, mrr, revenue
18%18% predicted probability of success on Acquire.com, based on ML models trained on real launch data.
BetaListMay not resonate with beta-testers · Missing: web3, chat, crypto
0%0% predicted probability of success on BetaList, based on ML models trained on real launch data.

Correct prediction on native model

Similar products

Al
Algolia (YC W14) Presents DocSearch74%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

Algolia (YC W14) Presents DocSearch

Hacker News9
Wh
Why we killed imwith and launched GIFted (Guggy YC S17)73%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

Why we killed imwith and launched GIFted (Guggy YC S17)

Hacker News2
Cr
Crowdfunding by BackerKit (YC W13)63%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

Crowdfunding by BackerKit (YC W13)

Hacker News2
Wh
Why YC when there's YZ?59%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

Why YC when there's YZ?

Hacker News5
So
Somethingservice (YC W23)57%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

Somethingservice (YC W23)

Hacker News2
YC
YC for YC Rejects78%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

YC for YC Rejects

Hacker News1
YC
YC Hack Winner – RepoWorld67%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

YC Hack Winner – RepoWorld

Hacker News1
Ca
CatchAll – slowest web search API that outperforms everything on recall56%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

CatchAll – slowest web search API that outperforms everything on recall

Hacker News7
Platus YC F24
Platus YC F2477%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

Your AI Paralegal

Product Hunt+271Legal
Re
Recall – Your Personal Encyclopedia52%Launch Intel prediction score: how likely this product is to succeed on its source platform, based on its name, tagline, and description.

Recall – Your Personal Encyclopedia

Hacker News34