A debate surfaced recently in open source: should projects accept contributions regardless of whether they were AI-assisted, as long as the output quality is high? It's a fair question that touches something real about how we work now. But in practice, it misses a few critical things.
The Maintenance Burden Doesn't Disappear
If a human writes unreadable code, that's a hiring or training problem. If AI generates unreadable code, you've got a different problem: the person reviewing it might not fully understand it either, and neither might the AI that generated it next time. Code review becomes theater instead of verification. "Quality" Needs Definition Earlier
When you scope a piece of work with a human, you can say "make it maintainable, document your reasoning, follow the style guide." Those are instructions a skilled person understands. With AI-generated work, quality often means "passes tests and runs", not "someone else can confidently modify this in six months." The goalposts shift without you noticing.
The Real Question Is Ownership
If something breaks in production, who debugs it? If a team member leaves, who explains the design? If requirements shift, who refactors it safely? These aren't abstract. They determine whether a contribution actually reduces work or just defers it. The honest take: AI contributions are fine. But the bar for review should go UP, not stay the same. You're not just checking that it works; you're checking that it's maintainable by humans who might not know why it was written that way. That's extra work upfront, and it's worth pricing it that way.