Drop-in Claude.md patch claims to fix Claude's degraded prose in Opus and Fable
myriable · reddit · 2026-09-08
一位重度用户发现 Opus 世代以来 Claude 的行文质量明显下降，充满压缩、难读的「Claudish」腔调。
他的做法是：先精确诊断出具体的文风毛病，再把逐条诊断写成指令加进 Claude.md，在 Claude Code 和 Claude.ai 上反复迭代评估，最终沉淀成一份可直接粘贴的 Claude Style Patch。
作者称该补丁经过多代模型实测，加入后文风能立即改善，让模型的智能更透明可读；他还援引 Boris Cherny 的公开言论，认为 Anthropic 内部也认同需要改进沟通质量。
More from coding & agent
- Harbor ships Adapters and Harbor-Index after a year: 120+ contributors, 300+ PRs, 10+ funders — xuandongzhao · 2026-09-08
- Don't hand off to a cheap model: Luna fails as a subagent for Astra on tasks needing real understanding — brandon_galang · 2026-09-08
- Evals Belong at the Center of Harness Engineering, Argues New Guest Essay — hugobowne · 2026-09-08
- 10,000 AI Agents Started With $5 Each to Survive Economically; Only 383 Remain Alive — anirbanbandyo · 2026-09-08
- Dev builds a one-customer app with Codex to audit a company's AI search visibility — aryanXmahajan · 2026-09-08
- Running a Neural-Network Ecology Simulation as a Desktop Wallpaper — Maxwell10206 · 2026-09-08