Why Does CLAUDE.md Keep Growing? Catastrophic Remembering in Agentic Coding
Source: arXiv:2608.11095 (cs.AI)
Authors: Kushal Chakrabarti
Published: 2026-08-11
URL: https://arxiv.org/abs/2608.11095
PDF: https://arxiv.org/pdf/2608.11095
Abstract (English)
Agentic coding READMEs like CLAUDE.md grow without bound in real repositories, stopping only when the repository retires or someone rewrites the file wholesale. We trace this to imperfect recall: appending an instruction is always cheap, but once an instruction's rationale is gone, deleting it without risking a correctness regression costs O(2^|D|) in a prompt of |D| instructions. We name the resulting divergence catastrophic remembering, the inverse of catastrophic forgetting around which continual learning is organized. First, we characterize this phenomenon across 247,694 instruction lifetimes in 1,867 repositories: agentic prompts grow without bound, more than tripling over their lifetime (+226%), gaining +4.9 net instructions every commit; further, the older an instruction gets, the less likely it is to be deleted (log-hazard -0.032/commit). Then, we show that prompt comments can halt the growth: inverting IFEval yields verifiable worlds whose optimal prompts are known, and there comments encoding latent reasoning remove 99.3% of excess instructions (+211.3% to +1.4%). Finally, applying the same inversion to WildIFEval, we show that prompt comments can improve real-world agentic instruction-following by up to 23.1%. If English is the new code, why don't we have comments yet?
Entry Metadata
- id:
f3098025 - source platform: arXiv
- paper id: 2608.11095
- primary category: cs.AI
- authors: Kushal Chakrabarti
- published: 2026-08-11
- quality_score: 4
- one_liner_author: openclaw
- added_at: None
- tags: ["arxiv", "external-scan", "research", "cs.ai"]
- rationale:
Notes
- Grounded by
opencli arxiv paper 2608.11095 -f json. - summary_en is derived from the arXiv abstract verbatim (truncated to 1500 chars when long).
- External scan mode: discovery via
opencli arxiv recent cs.AI -f json; per-paper fetch via opencli; dedupe via pipeline_utils.normalized_url_key.