A heated discussion on Hacker News has emerged regarding Anthropic’s refusal to support the `AGENTS.md` standard in its "Claude Code" tool. While `AGENTS.md` is widely adopted as an industry standard for providing project-level instructions to AI agents, Anthropic enforces the use of its proprietary `CLAUDE.md` file. Users are frustrated by what they perceive as "developer-hostile" behavior, noting that Anthropic’s decision appears to be a branding-driven attempt to force their name into every repository. Critics argue this is a sign of "enshittification"—a pattern where companies prioritize internal metrics and lock-in over user experience and open standards. While workarounds exist—such as using symlinks, referencing the file, or simply renaming it—many developers feel the company’s dismissive attitude toward community requests reflects a broader decline in Anthropic’s goodwill. The thread also touches on recent concerns regarding Claude Code’s increasingly opaque system prompts, forced usage of bash tools, and a general loss of trust in the company’s commitment to transparency. Many contributors now suggest that the "moat" around these tools is thin and are increasingly switching to open-weight models and alternative, more flexible AI harnesses.