o11 (YC W26) builds AI for financial firms: investment banks, private equity, hedge funds and asset managers.
They use Valyu to power SEC filing, company, and news search across those agents, with our Vercel @aisdk integration.
Full case study below 👇
ChemRxiv is now searchable inside Valyu.
Thousands of chemistry preprints, organic to materials to catalysis, queryable in plain English through our Search and DeepResearch APIs. Full text, figures included, structured the way agents need them.
It joins arXiv, PubMed, bioRxiv and medRxiv.
Now agents can move from biology to method to chemistry effortlessly, on the newest results, often months before they reach a journal.
Built for drug-discovery, materials, and synthesis-planning agents.
our slack had a graveyard of “can someone look into this?” messages
nobody ever does. by the time the answer matters someone's already pasted it from X, a customer email, or a competitor's changelog
so this long weekend I wired the @ValyuOfficial DeepResearch into Slack 👇
AI Demo Night 2.0 was 🔥
We co-hosted with @DustHQ last week at Encode Hub and the demos did not disappoint. Shoutout to the teams from @lightfern_ai , @coloopai and @ValyuOfficial - some seriously cool tools being built right now.
If you're building a startup, go check them
Today we shipped the world's most accurate DeepResearch API.
Built on the search infrastructure we have been shipping for 12 months. Already running hundreds of thousands of research jobs a month for hedge funds, banks, biotech, government, consulting, and frontier AI labs.
This small London startup just benchmark-mogged the entire AI search market. 🤯
Valyu DeepResearch just launched. 72.7% on DRACO (Perplexity's own benchmark). ahead of Perplexity DR, Parallel Ultra8x, Tavily, and Exa.
Their entire engineering team apparently fits in a single
Most published benchmarks aren't shipped with their underlying scripts - "trust me bro" benchmarks.
Ours don't. Datasets, run scripts, raw outputs, and judge evaluations are public and reproducible: