Things to know about developing with AI

AI Use Disclosure

The ideas here are my own. I use AI tools in various parts of the writing process, including summarizing my notes and prompts, improving grammar, tightening wording, and checking whether an argument is clear. These posts are as much notes for myself as they are for anyone else who is interested.

Assuming you have a state-of-the-art model - the thing to always keep in mind is minimalism.

The pi coding agent works best at the moment.

Developing software has turned the whole job into two critical skills: knowing what questions to ask & managing tech debt by reducing complexity.

  1. First, don’t just ask the model to do something; ask how to do something or what options there are to do something After it’s answered go tell it to do the thing.

  2. The next step is to always ask if there is a simpler alternative with fewer dependencies.

Anything with a too magical UI is going to give you the wrong mental ‘model’ what is happening and how to think about the LLM tool.

  1. Define as much of your problem in terms that can be checked. Give it a test and tell it the result it should reach before stopping.

  2. A bit of a hidden aspect many people seem to miss, the tone you take with the model is very important. Ask a bunch of cooking questions before having it write JavaScript or propose a tech stack is going to get you different results.

  3. Combine that all into the semi-obvious hack (which ‘subagents’ do automatically) - use the model to talk to another instance of itself. The model can translate your ramblings into coherent specs in the right tone. Use that to start new session gets you the good results. It’s also part of why the “first write a plan” works because it fills the context with the right tone and clear instructions.

First written: 2026-06-05
Last edited: 2026-06-05