Fair point. As a software engineer using Cursor, I’ve noticed it writes most of the code now. It’s easy to accept without review, which builds dependency. My role feels less like just coding and more like PM, tester, and reviewer combined.

We’ve started trusting AI like calculators. So, assuming it's right without checking. But LLMs can be confidently wrong, and once that habit sets in, even the “ChatGPT can make mistakes” warning fades into the background.