Sakana Fugu

Sakana AI's Recursive Self-Improvement (RSI) Lab

Instant LLM Updates with Doc-to-LoRA and Text-to-LoRA

Digital Red Queen: Adversarial Program Evolution in Core War with LLMs

The Darwin Gödel Machine: AI that improves itself by rewriting its own code

Continuous Thought Machines

AI CUDA Engineer: Agentic CUDA Kernel Discovery, Optimization and Composition

Transformer^2: Self-Adaptive LLMs

Automating the search for artificial life with foundation models

The AI Scientist: Towards Automated Open-Ended Scientific Discovery

Can LLMs invent better ways to train LLMs?

We raised $30M to develop nature-inspired AI in Japan