Journaling sounded stupid to me, until I tried it; and then the whole “do you really think that?”, “is that true?”, “what about this?” and “why are you lying to a sheet of paper?” started happening and I was like, “oh I get it.”
Come to think of it, that’s a major reason why fully agentic coding doesn’t resonate with me and/or feels like I’m not growing or learning. I’m short-circuiting the “journaling” step where I mentally attack my own thoughts and assumptions.
You should try writing a design doc with the AI before you have it write the code. It will make random extrapolations from your first prompt, some good and some bad. Then you get a chance to argue back and forth with yourself with the robot as a helper.
Oh yes! I find an argumentative rubber duck to be one of the best uses of AI for myself!
“Are there any ambiguities, open questions, or hidden assumptions in spec? Please think deeply.”