Software development with AI is starting to feel like cooking steak
Points and comments are a snapshot, not live.
AI-assisted software development produces passable results but not reliable quality without deep understanding.
The author compares cooking steak to building software with AI: both require skill to achieve consistent quality. AI can follow recipes and automate repetitive tasks but lacks judgment, cannot see the user's vision, and often produces technically correct but wrong results. The article argues that to build good software with AI, developers must understand software fundamentals, define quality standards, and recognize when the machine is confidently wrong. The solution is learning the craft deeply rather than outsourcing judgment to AI.
What commenters are saying
Commenters largely agree with the analogy but split on implications. A [#1] top-level comment argues the real problem is business pressure to settle for 'just edible' software, since acceptable is cheaper and profitable. Others counter that many software needs don't require excellence-passable is sufficient for most use cases, and AI enables rapid prototyping otherwise impossible. A separate camp criticizes the article's use of 'we' to imply low standards are universal, calling for professional accountability similar to aviation safety.