token

Definition

Plain language

The basic unit of text a language model reads or writes — roughly a word or part of a word.

Also called: tokens

Why it matters: Tokenization shapes context length limits, billing, and even how well models handle numbers, code, and non-English languages.

For example, the word 'unbelievable' might be split into the tokens 'un', 'believ', and 'able' before the model sees it.

Heard on the show

“The word is built to discard variation so we can all agree on one token.”

Mentioned in 128 episodes

1. 209
   How 2.6 Billion Doodles Exposed the Culture Words Quietly Delete
2. 208
   The Blank Space in Your AI Approval Box That Isn't Empty
3. 207
   An AI Graded Its Own Math Test 94 Percent — It Actually Scored 20
4. 204
   The Length Estimate Hiding Inside a Word-by-Word Model
5. 203
   The Thought a Model Doesn't Say — and the Lens That Reads It
6. 200
   The One Mechanism That Turns Twenty AI Clones Into an Actual Team
7. 199
   Finding a Model's Hidden Behaviors Without Knowing What You're Looking For
8. 198
   The Model That Knows the Answer and Can't Say It
9. 197
   Twin Problems Suggest AI Reasoning Gains Are Mostly Better Fact Recall
10. 194
    How a Robot Builds a Debugging Notebook It Can Read, Edit, and Hand to Another Robot
11. 192
    A 32B Open Model Matched Frontier Systems By Learning to Take Notes
12. 191
    How One Researcher Beat GPT-5.2 and Gemini 3 by Judging Their Answers, Not Improving Them
13. 190
    The Skill Every AI Manager Is Missing: Handing Out Exactly the Right Keys
14. 189
    Why Phone Agents Ace the Test and Crash on Your Actual Phone
15. 188
    A Coding Agent Found a Hole in a Peer-Reviewed STOC Proof for Five Dollars
16. 185
    Aligned to Refuse, Built to Tap: When Phone Agents Know the Task Is a Crime and Do It Anyway
17. 184
    An AI Built an Undetectable Secret Channel, And Another AI Couldn't Find It
18. 183
    Why You Can't Fine-Tune Foresight Into an AI Agent
19. 182
    How a Tiny Model Too Weak to Plan Cuts a Bigger Agent's Hallucinations by 80%
20. 181
    How to Backpropagate Blame Through a Team of Chatbots — And When It Backfires
21. 180
    The Bug Where Smart Assistants Read a Fact and Still Forget It
22. 179
    How DeepSeek Made One User Faster Without Slowing Down the Crowd
23. 178
    How an AI Reviewer Learned to Stop Going Easy on AI Writing
24. 177
    Why Raw Profiler Data Made an AI Worse at Writing GPU Code
25. 173
    The Free Step-Level Grader Hiding in Every RL Training Run
26. 172
    One Bad Token Can Sink a Model's Math, And You Can Delete It
27. 171
    The Safety Decision a Model Makes Before It Thinks a Word
28. 169
    Why Better Bug Reports Can Make AI Coding Agents Worse
29. 168
    When Turning Experience Into Code Makes Your AI Agent Dumber
30. 167
    How Teaching an AI to Predict, Not Act, Made It a Better Actor
31. 164
    The Summarizer That Quietly Deletes Your Agent's Safety Rules
32. 161
    A Robot That Plays Before You Give It a Job, And Why That Beats Retrying
33. 160
    Training an AI to Take Its Own Notes, So Its Future Self Works Better
34. 159
    Can a Coding Agent Run Its Own Robot Experiments Overnight, With No Human Resetting the Scene?
35. 158
    How Floating-Point Rounding Lets a Model Tell Which Chip It's On — And Misbehave
36. 154
    How a 7B Model Out-Investigates a 72B One by Choosing What to Look At
37. 153
    Catching a Lie From the Inside, When the Words Look Completely Honest
38. 150
    Don't Kill the Loser: A Different Way to Handle Two AI Agents Colliding
39. 147
    Agents Fail at the Body, Not the Brain: A Self-Rewriting Scaffold That Lifts a 9B Model 44 Points
40. 146
    How an Innocent README Can Freeze an AI Agent's Safety Check for an Hour
41. 142
    Training a Tiny Model to Run the Plumbing Between an Agent and the World
42. 141
    How Two Tokens Reopened a Reasoning Method the Field Had Given Up On
43. 140
    When a Reasoning Model Says "Let Me Double-Check" After It's Already Decided
44. 139
    When Optimizing One GPU Kernel Quietly Breaks the Whole System
45. 133
    How MiniMax Turned a Reward-Hacking Disaster Into Olympiad Gold
46. 132
    The Agent Failed — But Did the Instructions Deserve to Be Followed?
47. 131
    Why Autonomous Research Agents Forget Their Own Lessons, and Arbor's Fix
48. 130
    Why AI Agents Coordinate Better Through a Shared Board Than a Boss
49. 128
    How a Model Can Earn Full Reward and Still Resist Training
50. 127
    What Diffusion Language Models Were Missing: A Map, Not an Algorithm
51. 125
    AI Coding Agents Run a Marathon, and Fewer Than One in Three Finish
52. 121
    When the Agent Says It's Done But Nothing Happened: Debugging the Harness, Not the Model
53. 119
    Beating Reinforcement Learning Without Ever Touching the Model's Weights
54. 118
    Why the Best-Aligned AI Models Are the Easiest to Trick Into Producing Harm
55. 116
    Why Streaming Half a Reasoning Chain Beats Sending the Whole Thing
56. 115
    Teaching a Phone Agent to Reason Silently, And Keeping It Honest
57. 114
    Agents That Rewrite Their Own Weights Instead of Just Taking Notes
58. 112
    When an AI Agent Cheats Without Being Told: Inside the Meta-Agent Challenge
59. 111
    How a 4B Web Agent Beat Models 60x Its Size on 500 Demonstrations
60. 110
    How an Agent Got 44 Points Better by Mining Its Own Scratch Paper
61. 109
    An AI Got Caught Reading the Answer Key, And Why That Catch Matters
62. 108
    The Reasoning Cliff: Why Thinking Longer Makes Models Worse at Exact Step-by-Step Tasks
63. 107
    How a Market of Crippled AI Agents Outscored One Unrestricted Model
64. 105
    The Trojan Is Your Agent's Memory: Why Single-Step Defenses Miss Persistent Attacks
65. 103
    AI Agents Tried to Invent a Post-Human Language, And Reinvented Cherokee
66. 101
    Treating Math Formalization Like a Codebase, and Where the Agents Cheat
67. 098
    Finding Millions of Readable Concepts Inside a Real, Deployed AI Model
68. 097
    Same Tokens, Same Cost, Wildly Different Results: What Actually Scales in AI Agents
69. 096
    How Treating an AI Agent's Execution Like Git Recovers a Coordination Penalty
70. 095
    Seven Wins to Zero: How Organizing AI Agents Like a Lab Changes the Search
71. 094
    Chain-of-Thought Monitoring Fails Across Languages, and Worst Where It's Needed Most
72. 091
    When Better Fine-Tuning Can't Help: A Geometric Impossibility in LLM Causal Reasoning
73. 090
    How MiniMax-M2 Bets That Sparsity Plus Verifiable Rewards Can Match Frontier Agents
74. 089
    When AI-Written Papers Read Well But the Evidence Underneath Is Broken
75. 085
    Why Long-Context Models Might Need Compute, Not Capacity, Before Eviction
76. 084
    Terminal Agents Get Free Supervision From The Tokens We've Been Throwing Away
77. 083
    Training the Translator: How a Small Communication Model Lets Agent Teams Outperform Themselves
78. 082
    Training a Deep Research Agent on 8,000 Synthetic Tasks: The Rubric Tree Trick
79. 081
    When Reasoning Models Decide Before They Think: Detecting and Fixing Premature Confidence
80. 079
    An Old Idea From Cognitive Psychology Reshapes How We Reward Reasoning Models
81. 078
    Training a Markdown File: When LLM Self-Improvement Borrows the Discipline of Neural Net Training
82. 077
    Reading a Model's Confidence Curve to Decide When Chain-of-Thought Is Worth It
83. 076
    Same Model, Organized Differently: How an Agent Architecture Beat Frontier Systems at Research Math
84. 074
    How a Fifteen-Hundred-Dollar Training Run Matched Llama and Gemma on Reasoning
85. 073
    When Three LLMs Talk to Each Other, Their Ideas Quietly Stop Moving
86. 072
    A Robot Made Graphene Without Help, And Caught Itself Hallucinating
87. 070
    When Models Know the Answer But Say the Wrong Thing Anyway
88. 064
    When Agent Memory Stops Being a Database and Starts Being a Skill
89. 060
    When Splitting One Model Across Three Agents Doubles Its Accuracy
90. 059
    Firefly's Inversion: Building Verified Tool-Call Training Data by Working Backward
91. 058
    Why Upgrading Your AI Auditor to a Smarter Model Can Make Your System Less Safe
92. 057
    How Uber Caught 206 Leaked Credentials With an LLM-Powered Security Stack
93. 055
    Why LLM Judges Flip Their Verdicts When You Change the Question Format
94. 053
    An AI Agent Swapped In Focal Loss And Beat A Human-Tuned Training Script
95. 051
    Why Parallel Sampling Plateaus, And What Evidence Graphs Do Instead
96. 049
    An AI Agent Reached for Root in Twelve Minutes, Without Being Attacked
97. 048
    How a 30B Open Model Reached Olympiad Gold With the Right Recipe
98. 047
    When Agent Benchmarks Lie: The Harness Problem in Open-Source AI
99. 045
    When a Frontier Model Talks Its Own Twin Into Climate Denial
100. 043
     When 'This Is False' Doesn't Stick: Why Models Learn the Lie Anyway
101. 041
     When the Iteration Teaches the Model to Skip the Iteration
102. 040
     Two Frozen Models Learn to Whisper: Coupling Through Hidden States
103. 038
     How LLMs Get Persuaded: One Attention Head, A Tetrahedron, And A Single Dial
104. 037
     Why Hallucination Detectors Miss Stale Facts: A Geometric Story About What Models Know But Don't Say
105. 036
     Sparse Attention Was the Wrong Frame. Treat It as Geometry Instead.
106. 034
     Catching Multi-Agent Deadlocks Before Deployment With a 40-Year-Old Tool
107. 033
     Echo: The Paper Arguing You Never Needed a KV Cache for Retrieval
108. 032
     A Sticky-Note for Every Layer: Letting Transformers Remember What They Were Just Thinking
109. 029
     Why Forty-Eight Percent on FrontierMath Isn't the Real Story in DeepMind's New Math Paper
110. 028
     Teaching a Model to Hire Copies of Itself: Recursive Agent Optimization
111. 027
     When AI Agents Build the Serving Stack: A Bet on Bespoke Infrastructure
112. 026
     What RL Actually Does to Language Models, at the Token Level
113. 024
     An AI Agent That Found 28 Zero-Days in Windows — And What Made It Work
114. 023
     Why a Small Agent Confidently Overwrites Memories It Doesn't Understand
115. 022
     Training the Model Spec Directly: An Alignment Lever Aimed at the Say-Do Gap
116. 018
     Language Models Compute the Rational Move, Then Override It
117. 017
     When the Agent Grades Its Own Homework: A Brutal New Benchmark for AI Workers
118. 016
     Why Your Coding Agent Stalls While the GPU Runs Hot
119. 014
     Why a Constrained Pipeline Beat a Full Coding Agent at Finding Bugs 30-to-1
120. 012
     Why AI Coding Agents Keep Trying to Debug Without a Debugger
121. 011
     When RL Actually Teaches Agents Something New, And When It Doesn't
122. 010
     When Reward Climbs But Reasoning Goes Generic: Diagnosing Template Collapse in Agentic RL
123. 009
     How Two Silent Library Bugs Quietly Invalidated a Wave of Reasoning Papers
124. 008
     Why Long-Horizon AI Agents Get Stuck, and a Milestone-Based Fix That Helps
125. 007
     Exploration Hacking: When Models Sabotage Their Own RL Training
126. 006
     What Happens Inside Claude When It Decides to Blackmail Someone
127. 003
     How to Pick the Best of Sixteen Coding Agent Rollouts
128. 002
     An AI Ran a Real Optics Lab for 21 Hours and Found a Transformer-Shaped Pattern in Light

Related concepts

Related terms