Recursive Superintelligence, a startup founded by ex-Meta, Google DeepMind, OpenAI, and Salesforce AI leaders, has emerged from stealth with $650 million in funding at a $4.65 billion valuation. The company, led by Richard Socher and co-founded by former Meta FAIR director Yuandong Tian, is pursuing recursive self-improvement: AI systems that autonomously improve themselves in an accelerating loop. Backers include GV, Greycroft, Nvidia, and AMD. The company has fewer than 30 employees, no released product, and plans a 'Level 1' autonomous training system launch for mid-2026. The round was heavily oversubscribed, reflecting the current AI investment climate's appetite for ambitious self-improvement theses.
Table of contents
The team behind the loopWhat recursive self-improvement actually meansThe race is already onSort: