Ex-Meta Researcher Launches 'Recursive Superintelligence' AI Startup
About This Episode
Tian Yuandong, a former Meta researcher, has launched a new AI startup called Recursive Superintelligence, which focuses on 'recursive self-improvement' AI systems, with a $4.65 billion valuation and $650 million in funding from major investors including Nvidia and AMD. This development signals a significant investment in self-improving AI, with implications for the compute arms race between the US and China.
Our Take
A Chinese-born ex-Meta researcher just raised $650M to build AI that improves itself — and the US chip industry funded it: this is what the race to superintelligence actually looks like, and the AGI timeline just got a lot more concrete.
Continue Reading on Unscarcity
AGI Timeline 2026: What Altman, Hassabis, and Amodei Predict
Recursive self-improvement is the core mechanism most AGI timelines hinge on — a $4.65B bet on it is a concrete data point that every major CEO prediction now has to reckon with.
China's AI Strategy: Rival Architect, Not Villain
A Chinese-born researcher raising hundreds of millions from US chip giants to build self-improving AI is a perfect case study in China's role as a rival architect of the AI future — neither villain nor ally, but a parallel builder with its own agenda.