Stunning initial speed makes for a dopamine rush and desire to tell the world (been there myself).

Finding out you can’t debug it anymore because you don’t have the mental model of what’s going on is less likely to be shared with the world.

Some people are super comfortable with the prompt-wait-test-reprompt loop - I personally am not. I want to understand every line of code and find it more fatiguing and less rewarding to review pages of LLM code.

The sweet spot for me is using an LLM to write a single function at a time. That’s the unit of work. Constrained, easier to understand and critically, no unexpected changes to other areas of my code.