Right on! So true. merging slop without review, even slightly running against a manual test case can give you a good confidence and make you a good engineer.
With more code being generated and as easy as a command, it becomes rather intimidating to review code, to accept suggestion and produce more code. But code is not a magic wand its some assumptions crumpled with logical validation, both of them contradict yet when done right, creates a software that people use and breathe.
I agree to the manual as well as automated testing and LLMs will follow the pattern. If you already have tests in the codebase, it will make sure the test suite is updated when it makes a new change. There are obvious and unavoidable circumstances when you’d have to check the changes with manual test, its something that comes with the plate in the software engineering role. There is no denying in this, its a fact not an opinion.
Software and accountability is opposite side of a coin, you can’t let software account on its own, humans have bought its existence from their imaginations and manifestation, you need to validate and prove the thing you wanted to build.