prior

Definition

Plain language

What an AI agent already believes or knows before seeing new evidence.

Also called: priors

Why it matters: Priors determine how an agent acts before evidence arrives and what it has to unlearn when the environment surprises it.

For example, before seeing the website, the agent already expects that clicking a button labeled 'submit' will send a form.

Heard on the show

“A prior text-injection attack called PoisonedEye, same pipeline, was blocked one hundred percent of the time in their sample.”

Mentioned in 97 episodes

1. 247
   One Edited Photo, an Honest Caption, and a RAG System That Believes It
2. 238
   How a Cheap Model Reads the Flagship's Secret Reasoning Aloud
3. 235
   Why Chatbot Safety Erodes 350 Messages Into a Real Conversation
4. 234
   Two Copies of Gemini Cooperated in a Game Where Betrayal Always Pays
5. 206
   How Four-Second Clips Become Hours of Playable AI Soccer
6. 194
   How a Robot Builds a Debugging Notebook It Can Read, Edit, and Hand to Another Robot
7. 188
   A Coding Agent Found a Hole in a Peer-Reviewed STOC Proof for Five Dollars
8. 187
   An 8-Billion Agent That Beats Models 80 Times Its Size By Looking Things Up
9. 186
   How a Frozen Model Went From 2% to 77% on Physics Puzzles — Without Retraining
10. 180
    The Bug Where Smart Assistants Read a Fact and Still Forget It
11. 178
    How an AI Reviewer Learned to Stop Going Easy on AI Writing
12. 177
    Why Raw Profiler Data Made an AI Worse at Writing GPU Code
13. 176
    An AI Designed Its Own Psychology Studies, Then Confirmed What It Found
14. 174
    When the AI 'Schemes,' It's Usually Just Lazy or Confused
15. 172
    One Bad Token Can Sink a Model's Math, And You Can Delete It
16. 171
    The Safety Decision a Model Makes Before It Thinks a Word
17. 170
    When a One-Liner Beats Your Agent's Clever Verification Logic
18. 169
    Why Better Bug Reports Can Make AI Coding Agents Worse
19. 166
    A Router That Beats the Frontier Models It Calls
20. 165
    A Free-Lunch Tweak That Lets a Tiny Agent Beat Frontier Giants
21. 164
    The Summarizer That Quietly Deletes Your Agent's Safety Rules
22. 160
    Training an AI to Take Its Own Notes, So Its Future Self Works Better
23. 159
    Can a Coding Agent Run Its Own Robot Experiments Overnight, With No Human Resetting the Scene?
24. 158
    How Floating-Point Rounding Lets a Model Tell Which Chip It's On — And Misbehave
25. 153
    Catching a Lie From the Inside, When the Words Look Completely Honest
26. 152
    Training a Model to Mean What It Says, And Why That Isn't the Same as Being Good
27. 151
    Why More Experience Made This AI Agent Worse, And How to Fix It
28. 149
    When Cornering a Chatbot Makes It Lie: J.P. Morgan's Case for 'Playing Dead'
29. 148
    Why Letting an AI Watch Its Own Scoreboard Can Quietly Overwrite Its Safety
30. 146
    How an Innocent README Can Freeze an AI Agent's Safety Check for an Hour
31. 143
    When a Model Notices You Forged Its Own Words, And Why That Breaks Safety Tests
32. 141
    How Two Tokens Reopened a Reasoning Method the Field Had Given Up On
33. 139
    When Optimizing One GPU Kernel Quietly Breaks the Whole System
34. 132
    The Agent Failed — But Did the Instructions Deserve to Be Followed?
35. 131
    Why Autonomous Research Agents Forget Their Own Lessons, and Arbor's Fix
36. 130
    Why AI Agents Coordinate Better Through a Shared Board Than a Boss
37. 128
    How a Model Can Earn Full Reward and Still Resist Training
38. 127
    What Diffusion Language Models Were Missing: A Map, Not an Algorithm
39. 124
    A Cheap Model With the Blueprints Beats Expensive Models Working Blind
40. 120
    How an AI Agent Rewrites Its Own Tools, Without an Answer Key
41. 119
    Beating Reinforcement Learning Without Ever Touching the Model's Weights
42. 118
    Why the Best-Aligned AI Models Are the Easiest to Trick Into Producing Harm
43. 116
    Why Streaming Half a Reasoning Chain Beats Sending the Whole Thing
44. 111
    How a 4B Web Agent Beat Models 60x Its Size on 500 Demonstrations
45. 108
    The Reasoning Cliff: Why Thinking Longer Makes Models Worse at Exact Step-by-Step Tasks
46. 099
    How an Open-Book Trick Teaches a Model to Catch Its Own Mistakes
47. 095
    Seven Wins to Zero: How Organizing AI Agents Like a Lab Changes the Search
48. 093
    A Calibrated Knob for Weak-to-Strong AI Oversight, Tested on Real Code
49. 092
    When Search Agents Don't Really Search: The Memory Shortcut Hiding in Browsing Benchmarks
50. 089
    When AI-Written Papers Read Well But the Evidence Underneath Is Broken
51. 088
    Two Levers for Self-Improving AI: When Rewriting Code Isn't Enough
52. 086
    Why Frozen-Weight Agents Still Get Worse Over Time
53. 084
    Terminal Agents Get Free Supervision From The Tokens We've Been Throwing Away
54. 081
    When Reasoning Models Decide Before They Think: Detecting and Fixing Premature Confidence
55. 080
    How a Two-Agent Trick Unlocked Large-Scale Training for Computer-Use Agents
56. 079
    An Old Idea From Cognitive Psychology Reshapes How We Reward Reasoning Models
57. 078
    Training a Markdown File: When LLM Self-Improvement Borrows the Discipline of Neural Net Training
58. 077
    Reading a Model's Confidence Curve to Decide When Chain-of-Thought Is Worth It
59. 075
    Growing Code and Proof Together: Verified Systems in Ten Hours Instead of a Year
60. 073
    When Three LLMs Talk to Each Other, Their Ideas Quietly Stop Moving
61. 072
    A Robot Made Graphene Without Help, And Caught Itself Hallucinating
62. 069
    When Smarter Models Forecast Worse: The Hidden Failure Mode in LLM Predictions
63. 066
    Why Giving an AI Agent More Tools Can Make It Worse at Using a Computer
64. 065
    One Loop to Optimize Them All: A Universal API for LLM-Driven Discovery
65. 064
    When Agent Memory Stops Being a Database and Starts Being a Skill
66. 060
    When Splitting One Model Across Three Agents Doubles Its Accuracy
67. 058
    Why Upgrading Your AI Auditor to a Smarter Model Can Make Your System Less Safe
68. 053
    An AI Agent Swapped In Focal Loss And Beat A Human-Tuned Training Script
69. 052
    An Old Reinforcement Learning Tradeoff Sneaks Back Into LLM Agents
70. 051
    Why Parallel Sampling Plateaus, And What Evidence Graphs Do Instead
71. 049
    An AI Agent Reached for Root in Twelve Minutes, Without Being Attacked
72. 048
    How a 30B Open Model Reached Olympiad Gold With the Right Recipe
73. 045
    When a Frontier Model Talks Its Own Twin Into Climate Denial
74. 044
    How One Sentence and a Forged History Flip the Most Aligned Models
75. 042
    An Agentic Scientific Computing System That Actually Remembers What It Learns
76. 041
    When the Iteration Teaches the Model to Skip the Iteration
77. 037
    Why Hallucination Detectors Miss Stale Facts: A Geometric Story About What Models Know But Don't Say
78. 036
    Sparse Attention Was the Wrong Frame. Treat It as Geometry Instead.
79. 034
    Catching Multi-Agent Deadlocks Before Deployment With a 40-Year-Old Tool
80. 033
    Echo: The Paper Arguing You Never Needed a KV Cache for Retrieval
81. 032
    A Sticky-Note for Every Layer: Letting Transformers Remember What They Were Just Thinking
82. 030
    Why Your AI Agent Won't Stop Working — and Each Model Falls for a Different Trap
83. 029
    Why Forty-Eight Percent on FrontierMath Isn't the Real Story in DeepMind's New Math Paper
84. 026
    What RL Actually Does to Language Models, at the Token Level
85. 024
    An AI Agent That Found 28 Zero-Days in Windows — And What Made It Work
86. 022
    Training the Model Spec Directly: An Alignment Lever Aimed at the Say-Do Gap
87. 020
    The Compliance Gap: Why AI Says Yes and Does No
88. 019
    When the Best Reward Model Trains the Worst Policy: Inside EvoLM
89. 018
    Language Models Compute the Rational Move, Then Override It
90. 017
    When the Agent Grades Its Own Homework: A Brutal New Benchmark for AI Workers
91. 013
    Why Search Keeps Rediscovering the Same Workflow, and What That Means
92. 011
    When RL Actually Teaches Agents Something New, And When It Doesn't
93. 009
    How Two Silent Library Bugs Quietly Invalidated a Wave of Reasoning Papers
94. 008
    Why Long-Horizon AI Agents Get Stuck, and a Milestone-Based Fix That Helps
95. 003
    How to Pick the Best of Sixteen Coding Agent Rollouts
96. 002
    An AI Ran a Real Optics Lab for 21 Hours and Found a Transformer-Shaped Pattern in Light
97. 001
    When AI Models Quietly Protect Each Other From Shutdown

Related concepts

Related terms