Hacker News Reader

Comment by ngruhn | Hacker News Reader

I noticed that. If I don't write the code myself I only develop a very shallow mental model of what's doing. But I guess that always has been the product managers perspective.

View on HN

I don't think you need to write the code to develop a deep mental model of what's going on, but you do need to think about it a lot and intensely to develop that model and coding forces you to slow down, spend a lot of time thinking about the problem, and generally trying out different ways of looking at it.

Coding in this way is like having a personal Socrates to help walk you through the problem and achieve enlightenment.

This is the way I feel as well. I think there is an ethical way of using AI as an assistant: have it either generate a high level plan for what the architecture of your solution should look like (and decide whether the choice looks correct for the domain and actionable), or let it generate the code and then walk yourself through each line, documenting the solution as you go. I will go as far as to do the latter and then delete the code, re-implementing it to make certain I understand the relevant concepts.

The same thing happens with third party tools and frameworks/libs. The documentation very rarely help you develop a sound mental model of how it work, so your only option is to get your hands dirty - and often also burn your fingers in the process.

Reply on news.ycombinator.com