I Read The Clankers' Code

The article argues that developers should review AI-generated code rather than blindly merging it, because whoever's name is on the pull request and whoever carries the pager ultimately owns the code. The author, a self-described engineer who takes pride in craft, says AI tools can write code faster and often better than humans, but warns that "95% ain't 100%": models miss things, don't ask questions, have limited context windows, and optimize only for task completion. Risks compound as codebases grow, with AI building on output from older, less sophisticated models, and the author cites battle stories of spectacular failures—deleted "don't touch" code, careless library dependencies, invented features. The takeaway: AI coding assistants are helpful tools worth using, but accountability stays human. What to watch: whether engineering teams formalize mandatory human review of AI-generated code as adoption accelerates.