My experience with AI coding agents: If the chunks of work they do are so large that you only have to trust that the model has written the right code and passes the tests, then you're doing something wrong. Slow down a bit, make sure you are in control of what the AI is doing.

Unit tests are good and give good coverage, but at the end of the day, you own the code, so even if you don't review every single line of code, you should still have a solid understanding of the structure and key components.

I absolutely hate it when people say, "Who cares if the code is ugly as long as it works." Code craft is still very valuable. In the long run, you always get to appreciate the value of well-crafted and architectured software.