Another good Cory Doctorow article on AI business models, AI use cases, and AI-assisted humans vs human-assisted AI:
https://pluralistic.net/2024/04/23/maximal-plausibility/
Troubleshooting AI generated code with humans is much tougher than troubleshooting human generated code with AI. AI can churn out plausible but not quite correct hard-to-diagnose code that can take an expert significant effort to repair.
Those reverse-centaur human-assisted AI problems are exactly the sort of problems that someone might bring to to the forum.
AI might be useful in doing a code review on a program, but those aren't the issues folks bring to the forum looking for expert assistance.