AI made writing code cheaper, but did it make software engineering better?
A green CI pipeline proves that tests passed. It doesn't necessarily prove that the right things were tested.
I'm increasingly interested in the gap between passing tests and actual confidence in a system.
How do you approach that distinction when reviewing AI-generated code?
commit b60bf2f
I don't think writing code was ever the real bottleneck in software engineering. Writing code faster only means we hit the real bottlenecks faster.
Building feature-wise helps in reviewing the tests. Problem I am facing is opposite, it creates too many unnecessary tests. So a clean up ticket to revist test. Of course, I have seeder for each flow in the app to visit via claude-in-chrome when required.