Second unexpected side effect of agentic coding: I have never been more focused on correctness.
When Claude writes code, I’m biased toward assuming it’s incorrect. So I am spending most of my time asserting its correctness
...
emphasis on correctness because assume AI code is incorrect