Sakana AI launches RSI lab to break compute arms race

Sakana AI has launched a research lab focused on recursive self-improvement. The startup aims to move beyond the compute arms race by developing AI systems that can iteratively redesign and improve themselves. Sakana highlights projects like LLM-Squared and Darwin Gödel Machine as evidence of progress. The AI Scientist system wrote a peer-reviewed paper, published in Nature in 2026. Sakana outlines a four-phase roadmap toward self-improving AI systems.