Study: CLAUDE.md files grow unbounded; comments cut 99.3% excess instructions
omarsar0 · x · 2026-08-13
A study of 247,694 instruction lifetimes across 1,867 repos shows CLAUDE.md/AGENTS.md files grow without bound: appending is free, deleting costs exponential verification. Instructions more than tripled over lifetime, gaining 4.9 net per commit. Adding comments explaining rationale removed 99.3% excess instructions in verifiable settings and improved agentic instruction-following by up to 23.1%.
Related event: Study Reveals 'Ratchet Effect' Causing Infinite Growth in CLAUDE.md(2 posts)→
More from coding & agent
- A vector DB is not LLM memory: long thread maps write/read paths, failure modes and evals — techNmak · 2026-10-16
- OpenUI 1.0 launches open source, turning agent text answers into live interactive UIs — heyshrutimishra · 2026-10-16
- 10 GitHub repos to see what your AI agent really did, from Langfuse to claude-code-log — victor_explore · 2026-10-16
- Mosaic: multi-session agentic RL with hierarchical advantage estimation — LARK-Lab · 2026-10-16
- Recursive Environment Scaling tanks six models' success rates by 57–77 points — Lingxiao Jing · 2026-10-16
- Harvard's MPT: decentralized agent swarms learn self-orchestration via bidding — Harvard · 2026-10-16