The most useful reframe in any conversation about AI-generated code is a blunt one: quality is still a choice. It was a choice when you wrote everything by hand, and it remains a choice when a machine writes it for you. The tooling changed. The accountability did not.
This should be clarifying, not discouraging. If quality is a choice, it is achievable regardless of what stack you use. The team that builds disciplined review into their AI workflow ships better software than the team using the same tools without it. The AI is not the variable. The discipline is.
What changes with AI is the cost of the cleanup cycle. Manual refactoring used to be expensive enough that technical debt accumulated past the point of feasibility. AI makes large-scale cleanup fast and cheap - but only if you are willing to audit first, name the problem, then direct the machine. That sequence matters: audit, identify, direct. Skip the audit and you are not using AI to raise the floor. You are using it to pile up debt faster.
The teams skipping that step are not doing so because AI makes inspection unnecessary. They have decided that good software is not worth the friction. That is their decision. They own the consequences, not the tool.
Thoughts? Find me on Bluesky.