Three AI habits that stuck

I have tried every new AI tool that has launched in the past year. Most of them left my workflow within a week. A few changed how I work permanently.

The first habit is asking for a plan before code. I describe the thing I want to build, ask for a plan broken into steps, read the plan, push back where I disagree, and only then ask for code. The plan is almost always better than the code I would have written straight through.

The second habit is rubber ducking in writing. When I am stuck, I open a chat and write out the problem as though I am explaining it to a person. Half the time I figure it out before I hit send. The other half, the model points at something I had missed.

The third habit is asking for the opposite. If I just wrote something and I am proud of it, I ask the model what is wrong with it. It is very good at being a kind critic, which is exactly the thing I am bad at being for my own work.