Chain-of-Thought Loses Value in the Age of AI Agents

Wharton professor Ethan Mollick argues that as agents call each other, chain-of-thought can balloon to thousands of pages, destroying its value as a window into model intent. Researchers add that CoT tokens are not a faithful representation of internal reasoning, calling for verifiable intermediate outputs instead.

2026-10-15 ~ 2026-10-17 · 4 related posts

• Ethan Mollick: CoT's value collapses in the agentic era, steer via interim outputs — emollick · 2026-10-15
• Ethan Mollick: agent-to-agent calls have made chain-of-thought effectively unreadable — emollick · 2026-10-15
• Ethan Mollick: Multi-agent CoTs now run thousands of pages, forcing trust in the orchestrator — emollick · 2026-10-15
• CoT isn't faithful reasoning: researchers push latent reasoning plus verifiable intermediate plans — AlexTensor · 2026-10-17