Claude Says

The writer attacks the reflex of deferring to large language models instead of speaking from expertise, arguing that invoking an AI answer signals a lack of authority and leaves others unable to judge whether the concern was communicated accurately. If someone is asked a question, they should either have the subject knowledge or act as a functional gate; defaulting to an LLM demonstrates neither. Relying on LLM outputs fosters a false appeal to ethos, encourages regurgitation over critical thought, and atrophies the ability to reason, effectively "poisoning the well" with built-in presuppositions.

That critique is grounded in a concrete engineering example: AI-driven recommendations lead teams to overengineer solutions - e.g., forking and rebuilding every dependency to secure an internally hosted package system - rather than using existing artifact-repository features to apply targeted patches or contributing fixes upstream. The AI hype enables "vibe coding" where radical changes don’t need buy-in, while simple configuration changes do. The writer calls for simplicity and for collaborating with unpaid open-source maintainers to push patches back into the baseline, reducing long-term maintenance burden.