Ethan Mollick: agent-to-agent calls have made chain-of-thought effectively unreadable
emollick · x · 2026-10-15
Wharton professor Ethan Mollick notes that reading a model's chain-of-thought used to reveal where it was heading, but with agents calling other agents and CoTs running to thousands of pages even for simple tasks, that's no longer possible.
- The collapse of CoT's value (even summarized) is changing how we work with AI.
- Steering AIs and getting observability now requires asking for interim outputs and tests, and using AI to troubleshoot issues you find.
Related event: Chain-of-Thought Loses Value in the Age of AI Agents(4 posts)→
More from AGI Musings
- Anthropic's timeline mocked: 75% automation by 2030 but RSI by 2027? — QuintinPope5 · 2026-10-17
- e/acc figure backs local AI as 'civilizational infrastructure' against cloud gatekeeping — beffjezos · 2026-10-17
- AI Discourse: 'Childless Careerists' Face Ego Death, While Others Urge Simply Rejecting AI — GabGarrett · 2026-10-17
- Chamath's $200B Anthropic Argument: Balance Sheets May Beat Smarter Models — CurieuxExplorer · 2026-10-17
- Why the AI data industry is dominated by under-28s: three reasons plus an unstated fourth — dejavucoder · 2026-10-17
- Brain alternates between conscious and unconscious modes ~15 times per second — neuroecology · 2026-10-17