Sakana AI's MASS: A 27B Open Model Improves Its Own AI Team — and the Blind-Spot Risk
Posted on 12th Oct 2026 12:05:44 in Artificial Intelligence, Machine Learning
Tagged as: Sakana AI, MASS, Self-Improving AI, AI Agents, AI Research
Sakana AI and UC Berkeley researchers show how a single 27B open-weight model can run, judge and improve its own multi-agent team — lifting research-task efficiency up to 1.6x per output token, while coding benchmarks slipped and the authors flag a shared blind-spot risk at the heart of self-improving AI.
