Co-founder Llion Jones launched a dedicated research lab focused on recursive self-improvement. This approach aims to bypass the massive compute requirements dominating US frontier labs. While Sakana AI bets on iterative self-optimization, Anthropic warns of associated control risks. Practitioners should monitor if this method effectively reduces the hardware barrier for high-performance model development.