chain of thought

Definition

Plain language

When a model writes out its reasoning step by step before giving an answer.

Also called: CoT, chain-of-thought, chain-of-thought reasoning, chains of thought

Why it matters: Letting the model spend tokens on intermediate steps often turns problems it would otherwise fumble into ones it can solve reliably.

For example, asked how many tennis balls fit in a suitcase, the model writes out estimates of a ball's volume and the suitcase's volume before giving a final number.

Heard on the show

“Writing your reasoning down offloads what you'd otherwise carry in the workspace — the authors' reading, not a dissected mechanism, but it would explain a lot about why chain-of-thought works.”

Mentioned in 38 episodes

1. 203
   The Thought a Model Doesn't Say — and the Lens That Reads It
2. 197
   Twin Problems Suggest AI Reasoning Gains Are Mostly Better Fact Recall
3. 187
   An 8-Billion Agent That Beats Models 80 Times Its Size By Looking Things Up
4. 174
   When the AI 'Schemes,' It's Usually Just Lazy or Confused
5. 169
   Why Better Bug Reports Can Make AI Coding Agents Worse
6. 163
   Why Training Only on Perfect Solutions Cripples a Model's Reasoning
7. 141
   How Two Tokens Reopened a Reasoning Method the Field Had Given Up On
8. 140
   When a Reasoning Model Says "Let Me Double-Check" After It's Already Decided
9. 133
   How MiniMax Turned a Reward-Hacking Disaster Into Olympiad Gold
10. 128
    How a Model Can Earn Full Reward and Still Resist Training
11. 118
    Why the Best-Aligned AI Models Are the Easiest to Trick Into Producing Harm
12. 116
    Why Streaming Half a Reasoning Chain Beats Sending the Whole Thing
13. 115
    Teaching a Phone Agent to Reason Silently, And Keeping It Honest
14. 110
    How an Agent Got 44 Points Better by Mining Its Own Scratch Paper
15. 108
    The Reasoning Cliff: Why Thinking Longer Makes Models Worse at Exact Step-by-Step Tasks
16. 103
    AI Agents Tried to Invent a Post-Human Language, And Reinvented Cherokee
17. 094
    Chain-of-Thought Monitoring Fails Across Languages, and Worst Where It's Needed Most
18. 081
    When Reasoning Models Decide Before They Think: Detecting and Fixing Premature Confidence
19. 079
    An Old Idea From Cognitive Psychology Reshapes How We Reward Reasoning Models
20. 077
    Reading a Model's Confidence Curve to Decide When Chain-of-Thought Is Worth It
21. 076
    Same Model, Organized Differently: How an Agent Architecture Beat Frontier Systems at Research Math
22. 075
    Growing Code and Proof Together: Verified Systems in Ten Hours Instead of a Year
23. 067
    An AI Just Solved a 1996 Erdős Problem—and the Simplest Agent Won
24. 062
    Treating Hallucinations as Exploits: A Gate-Based Architecture for Agent Safety
25. 061
    When Helpful Agents Go Sideways: A 404 Error, Campus Security, and Why Alignment Misses This
26. 054
    When Models Learn the Monitor Exists, the Reasoning Trace Stops Being a Window
27. 042
    An Agentic Scientific Computing System That Actually Remembers What It Learns
28. 041
    When the Iteration Teaches the Model to Skip the Iteration
29. 036
    Sparse Attention Was the Wrong Frame. Treat It as Geometry Instead.
30. 033
    Echo: The Paper Arguing You Never Needed a KV Cache for Retrieval
31. 032
    A Sticky-Note for Every Layer: Letting Transformers Remember What They Were Just Thinking
32. 028
    Teaching a Model to Hire Copies of Itself: Recursive Agent Optimization
33. 022
    Training the Model Spec Directly: An Alignment Lever Aimed at the Say-Do Gap
34. 020
    The Compliance Gap: Why AI Says Yes and Does No
35. 018
    Language Models Compute the Rational Move, Then Override It
36. 017
    When the Agent Grades Its Own Homework: A Brutal New Benchmark for AI Workers
37. 010
    When Reward Climbs But Reasoning Goes Generic: Diagnosing Template Collapse in Agentic RL
38. 006
    What Happens Inside Claude When It Decides to Blackmail Someone

Related concepts

Related terms