This.

It’s actually extremely good bang for buck now to invest more time in various linters or static checking that enforce more correctness / project-specific rules.

Not only does the AI help you write the linter in 10% of a time - linters are usually fairly short scripts that are easy to test in practice, so the happy path for AI).

But asking the AI “make this linter pass” after you’ve e.g. done some extensive refactors, is very effective and time-saving.