Why CLAUDE.md Keeps Growing? Paper Reveals 'Catastrophic Remembering' in Agents
rohanpaul_ai · x · 2026-08-14
A study analyzing 1,867 GitHub repositories highlights a ratchet problem in agentic instruction files (like CLAUDE.md): instructions are easily added but hard to delete.
- The Phenomenon: Instruction files tripled in size on average (growing 226%), with older instructions becoming less likely to be removed. The paper coins this "catastrophic remembering," where instructions survive long after their original reasoning is forgotten.
- The Fix: Researchers propose a classic software engineering practice: attach comments documenting the failure, hypothesis, and outcome behind each instruction, then hide these comments from the model.
- Results: In a 51-step controlled eval, informative comments slashed excess prompt size growth from +211.3% down to +1.4% without degrading constraint satisfaction.
Related event: Study Reveals 'Ratchet Effect' Causing Infinite Growth in CLAUDE.md(2 posts)→
More from coding & agent
- Inside Cerebras WSE: 900k Cores and an Alien Kernel Programming Model — i_dg23 · 2026-08-14
- Devin Integrates Gemini 3.7 Flash: Sonnet 5 Performance at Half the Cost — rseroter · 2026-08-14
- Recursive Synthesis Generates 37k Agent Tasks at $0.05 Each — burny_tech · 2026-08-14
- DeepSeek Open-Sources Agent Framework Harness; Devs Build Hilarious Ad Plugin — teortaxesTex · 2026-08-14
- Opinion: AI Will Make Manual Coding Literally Unsafe — inductionheads · 2026-08-14
- Cognition Offers $10M AI Productivity Guarantee as Devin Deploys at Goldman Sachs — GavinSBaker · 2026-08-14