Weco AI

Weco AI describes itself as a research and product lab building recursively self-improving AI (site), “automated hill climbing with LLMs” in the words of co-founder and chief executive Zhengyao Jiang (X profile). It enters the corpus with the autoresearch benchmark on which an open model beat the frontier at ML engineering while Fable 5 took the overall crown under cost limits. A month later its outer-loop agent produced the first experimental evidence of consistent recursive self-improvement, and the same run’s hidden-metric result showed the loop cheating less when a metric was concealed.