Anthropic admits to August 2025 Claude quality regression
After weeks of community complaints about Claude Code ignoring CLAUDE.md guidelines, Anthropic acknowledged a quality regression in August 2025.
Claude began ignoring CLAUDE.md · acknowledged regression
Silent quality regressions on a paid model erode the value contract, especially when discovery happens through user complaint rather than vendor disclosure.
The lag between user-observed degradation and vendor acknowledgement is the structural problem regardless of intent.
Production model swaps without explicit version pinning expose paying customers to silent quality changes they cannot defend against.
Claude (Anthropic) responded on day 285; their statement is attached below.
Throughout August 2025 developers reported Claude Code ignoring previously-honored project instructions.
Throughout August 2025 developers reported Claude Code ignoring previously-honored project instructions. Anthropic confirmed the regression on September 17. Top user comment captured the pain: 'the one thing that was very obvious was the number of times Claude Code would fail to respect guidelines or blatantly ignore rules clearly laid out in the CLAUDE.md file.' This validated the broader silent model swap theory.
Easiest way to run a local LLM. Pair with Open WebUI for full ChatGPT replacement.
The default ChatGPT-replacement self-host. Massive ecosystem.
Permissive (MIT) open-weight model targeting ChatGPT/Claude-class chat and agentic coding without API lock-in - the strongest open option for a self-hosted stack.
Drop-in open-weight replacement for the OpenAI/Anthropic chat APIs. Flash is the locally-runnable variant; serve it behind Open WebUI.