Post

I'm beginning to suspect that a key skill in working effectively with coding agents is developing an intuition for when you don't need to closely review every line of code they produce. This feels deeply uncomfortable!

In my experience the main problem I find with LLM driven coding is that you never get as deep into the problem space as you do when you write everything from scratch and it’s not possible to replicate that depth when just reviewing output 🤷🏻

I think problem space needs to be defined here, I have a feeling you are talking about different things.

Quite possibly, I don’t think I chose my words very carefully. I guess I’m talking about what might be called achieving “flow state”. For me at least that means 👇🏻

You’ve worked deeply into the problem/solution enough that you can visualise what every line of code does, why it exists, where the tradeoffs are, what the next 10 steps are, exactly where the uncertainty lies etc… I have a model visualised in my head that I can manipulate, zoom in, zoom out etc…