One prompt line stops Claude from rewriting whole files for tiny edits
alex_verem · x · 2026-09-14
The post links to the official Claude Platform docs page "Prompting Claude Fable 5.1", summarizing behavioral differences and prompting patterns:
- File edits: The model prefers rewriting an entire file rather than making a small targeted edit — the result is identical but you pay for every token of the rewrite; one sentence in your prompt fixes this.
- Other behaviors: choosing the right effort level for latency/cost, requesting user-facing progress updates between tool calls, batching independent tool calls, and keeping conversation history append-only.
- Writing & formatting: default prose runs long and dense; density can be tuned via prompting.
- More: specifying what compaction summaries preserve, scoping changes/tests to the request, quoting retrieved sources properly, task completion, subagents, and vision.
Related event: Official docs offer prompt fix for Claude's whole-file rewrites(2 posts)→
More from coding & agent
- Open-source ComfyUI nodes bring relight video generation via MiniMax H3 — Emotional_Example_12 · 2026-09-14
- Ran an AI Agent for 2 Days and 300% of Quota to Fix My App — Sign-in Still Broken — jarrodwatts · 2026-09-14
- Hugo Bowne launches Master Agentic Data Science workshop with PyMC's Thomas Wiecki — hugobowne · 2026-09-14
- DeepSeek Harness hailed as the best multi-agent harness outside the labs — bookwormengr · 2026-09-14
- pi-vcc vs pi-blackhole: Sub-second Lossless Context Compaction or 3-LLM Recall? — mailto_devnull · 2026-09-14
- CS professor bans coding agents on assignments, students' reaction becomes a meme — CSProfKGD · 2026-09-14