Code quality still matters
Key takeaway: Review AI-written code against the same bar as your own: the right problem, the simplest design, and edge cases handled.
AI writes almost all my code now, and code quality still matters. AI code fails the same ways human code does, with unclear names, tight coupling, and missing tests. It just fails faster.
When code shows up instantly and looks reasonable, it’s tempting to trust it and move on. But more volume means bad code arrives faster too, and without checks you pile up technical debt at machine speed. That’s the trap from Ship as fast as possible, but not faster, and it doesn’t change because a model is typing.
Caring about clear, simple code isn’t nostalgia. Someone, or some agent, has to understand the code later, and dense code makes that harder. Every unneeded abstraction is a maintenance cost. And the structure of the code decides how easily it can change and, with agents working in parallel, how many changes can land at once.
That makes code review the main quality gate. When a model writes a hundred lines in seconds, the person reviewing them is what stands between that code and production. Ask whether it solves the right problem, whether it’s more complex than it needs to be, whether it follows the patterns already in the codebase, whether it handles the edge cases, and whether someone will understand it in six months.
Don’t spend that review on syntax nitpicks. Review the outcome, not the output. Writing it fast is fine. Reviewing it carefully and keeping it maintainable is still the job, same as it was before.