HN Discussion: 'LLMs Learn What Programmers Create, Not How Programmers Work' — 37 Points, Practitioner Insight

Hacker News·low signal

A Hacker News post with 37 points and 14 comments argues that LLMs trained on code artifacts learn the output of programming but not the process — the debugging, design decisions, false starts, and architectural reasoning that produce good software. The discussion highlights a growing practitioner consensus that AI coding tools optimize for code generation speed while missing the cognitive workflow that makes code maintainable. For builders evaluating AI coding tools, this frames the key question: are you automating output or understanding process?