According to Reuters, Google co-founder Sergey Brin has been deeply involved in the company's AI research in recent months. Despite holding no formal management position, he leverages his influence to push model training and requires key AI employees to fully commit to Gemini. One direction Brin explicitly pushes resources toward is recursive self-improvement (RSI), which involves AI in creating the next generation of AI.
Brin's move is driven by Google's urgency: as of August this year, the next-generation Gemini flagship model's release has been delayed by about two months, with internal tests showing it lags competitors in key areas like coding. Brin previously returned in 'founder mode' to push the merger of Google Brain and DeepMind and launch Gemini 3; this return is seen as an attempt to replicate that integration and acceleration.
RSI is not unique to Google. In May, Anthropic published an article titled 'When AI builds itself' detailing its progress; in late July, OpenAI brought back Weng Li, co-founder of Thinking Machines Lab, to lead a research team focused on RSI; and Jeff Dean, who recently left Google, has a startup goal directly targeting RSI—letting AI run experiments, evaluate results, fix weaknesses, and iterate continuously.