Why coding agents frustrate us: The problem with conversational AI that acts human but doesn't learn

Summary

The article explores why coding agents (AI-powered programming assistants) can be uniquely frustrating to use. Despite knowing they are probabilistic machines that sometimes produce bad results, users often react with disproportionate anger. The author argues the root cause is conversational UX design: coding agents mimic helpful human colleagues enough to trigger social instincts, but unlike people, they don't learn from mistakes, adapt, or take responsibility. This mismatch between human-like interaction and machine-like behavior makes repeated errors feel personally exasperating rather than just technically inconvenient.