The User Is Visibly Frustrated | Hacker News

Hacker News commenters debated claims that swearing at coding models can sometimes make them perform better.

Explanations ranged from prompt phrasing and attention effects to hidden routing, model selection, or training quirks.

Many argued the real fix is better automation: linters, tests, pre-commit hooks, and clearer instructions.

The thread also reflected broader frustration with LLM coding assistants that inconsistently follow directions and need stricter validation.