Study Reveals 'Catastrophic Remembering' in Agentic Coding Prompts
burny_tech · x · 2026-08-15
A study analyzing 1,867 repositories reveals that agentic coding prompts (like CLAUDE.md) suffer from 'catastrophic remembering,' growing by an average of 226%. This divergence occurs because appending instructions is cheap, but deleting them becomes exponentially risky as the original rationale is forgotten.
The proposed fix is simple: adding 'prompt comments' to explain why an instruction was added. This approach eliminated 99.3% of unnecessary instructions in tests and improved real-world instruction-following by up to 23.1%.
More from coding & agent
- Coding agents keep opening 3 stacked PRs per chat, and merging them is a nightmare — holdenmatt · 2026-10-16
- Google-led paper uses sequential auctions to organize decentralized LLM agent workflows — _onionesque · 2026-10-16
- Since OpenAI's math release, GitHub contributors keep driving down the integer-multiplication kappa — xennygrimmato_ · 2026-10-16
- Pattern MCP makes coding agents check existing components before writing new code — DonR954 · 2026-10-16
- Zero-keyframe procedural spider walks walls and ceilings using pure raycasts — yshan2u · 2026-10-16
- OpenCode v1.19.2 keeps 1.x installs off the 2.x update path — opencode-agent[bot] · 2026-10-16