You can't unit test for taste

Category: Pulse
Status: discussion
Sources: post, HN
Summary: A 246-point thread debates the limits of automated tests in an era of AI-generated code, arguing that correctness checks cannot capture design taste and judgment. It is an opinion essay rather than an implementation writeup.
Comments: Commenters split between treating taste as tacit senior-engineer judgment that review must catch and arguing that much of what is called taste reduces to teachable, testable constraints.