Keenable, a stealth‑mode startup founded by former Yandex search lead Andrey Styskin and AI researcher Matthias Petri, emerged with $26 million in seed financing led by Accel, with Conviction Partners and angel investors also participating. The duo argues that today’s web search infrastructure, tuned for human attention spans, is ill‑suited for AI agents that can ingest far larger corpora and therefore need a search layer optimized for machine consumption.
The company claims to have constructed a search index exceeding 100 billion documents and already serves its API to several AI labs and inference providers during both model training and runtime. While customer names remain undisclosed, Keenable has announced a partnership with voice‑AI firm Gradium to enable live information retrieval. Styskin emphasizes that the core innovation lies in task‑specific index structures that prune the search space rapidly, avoiding the prohibitive cost of scanning the entire web for each query. An upcoming product, WebQueryLanguage, will let AI systems compose answers by stitching together fragments from multiple sources even when no single page contains a complete response.
Accel’s Zhenya Loginov notes that major search operators are restricting access to their public APIs, leaving few web‑scale alternatives for AI developers. Drawing on his experience at Amazon and Yandex, Styskin sees this gap as an opportunity to sell a purpose‑built retrieval layer that can be licensed to AI platforms needing up‑to‑date, machine‑readable web data.
Why this matters
Source fact: Keenable’s index exceeds 100 billion documents and its API is already in production use by AI labs and inference providers. Inference: Such a machine‑oriented search layer could lower latency and cost for retrieval‑augmented generation pipelines, potentially shifting the economics of AI‑driven applications that depend on up‑to‑date web knowledge.
Share this article
Found this insightful? Share it with your community on Reddit, X, or copy the link.