Y combinator Portfolio

▤ Summary

LLM Stats / ZeroEval introduces a new tool for building more reliable AI agents through continuous evaluation. The product focuses on calibrated LLM judges that improve over time as they learn from production data and labeled failures. It also adds Autotune, which runs evaluations across models and supports prompt optimization from a small set of human samples. The team positions the launch as infrastructure for companies that need better measurement and reliability in AI products. The article also highlights the founders’ background and invites teams with production agents to book a demo.

▦ Linked Companies