Ethan Mollick: RSI May Not Be the Whole Game — AI Research Faces Multiple Caps

emollick · x · 2026-09-13

Ethan Mollick argues recursive self-improvement (RSI) may not be the whole story: AI capabilities could be capped by compute or architecture, AI-powered research itself may be limited (e.g. ability to find interesting problems), and companies may fail to exploit the resulting work. He also notes the anxious tone of recent Anthropic and OpenAI posts on the topic.

Related event: Mollick: OpenAI and Anthropic Hint Recursive Self-Improvement Has Begun(3 posts)→

Original post →

More from AGI Musings

AGI Musings channel →