AI coding instruction files grow 226% on average; study proposes 'catastrophic remembering' and comments as fix
rohanpaul_ai · x · 2026-09-07
A study of 1,867 GitHub repositories reveals that AI coding instruction files (e.g., CLAUDE.md) grow unboundedly, with instruction count rising 226% on average, and older instructions become harder to delete. The researchers term this 'catastrophic remembering,' the inverse of catastrophic forgetting. Key findings: instruction files more than triple over their lifetime, gaining +4.9 net instructions per commit; deletion likelihood decreases with instruction age (log-hazard -0.032/commit). The proposed fix is to attach comments recording the rationale behind each instruction, hidden from the executing model. In controlled IFEval tests, informative comments reduced excess prompt size from +211.3% to +1.4% without affecting constraint satisfaction. The paper asks: 'If English is the new code, why don't we have comments yet?'
More from coding & agent
- Building a neutral Codex-class harness requires running the original inner harness, dev says — joshalbrecht · 2026-09-07
- Dev Builds Timestamped Summary Site for All MLOps Community Talks, With an MCP Server — strickvl · 2026-09-07
- Shipping an Obsidian plugin with GPT-6 Astra: scan limits, SVG icons, review delays — vista8 · 2026-09-07
- Grok Build 1.0.22 ships subagent messaging, first-party MCP server and safer Auto mode — mark_k · 2026-09-07
- Agents on a 16GB MacBook Air M5: Qwen 3.5-9B 4-bit can't yet build a working app — Fluid-Author-9566 · 2026-09-07
- Remastering old i2v videos with MiniMax H3: a depth-map + reference workflow — WaitAcademic1669 · 2026-09-07