Claude Is Not Your Architect. Stop Letting It Pretend | Hacker News

Hacker News debated a post criticizing Claude as a weak architect and the broader problem of trusting AI like an authority.

Commenters argued that LLMs should be treated as fallible assistants, not confident experts, especially in technical work.

The thread weighed whether polite, deferential language helps users or increases anthropomorphic trust in AI.

A key concern was that confident but wrong outputs can mislead developers more than clearly tentative answers.

The discussion highlights a design question for AI products: how to balance helpfulness, confidence calibration, and reduced misuse.