Saw this post - How I Use “AI” by Nicholas Carlini through Antirez’s post

I use LLM very similarly, and these are my (thoughts) on his bullet list.

  • As a tutor — (I have used to define and clarify concepts)
  • To get started — (The bootstrapping problem; great for starting projects in frameworks and languages that you are not a regular.)
  • To simplify code — ( ask LLM to rewrite something)
  • For boring tasks — (I know how to do it, but too lazy or busy to write it all myself. ask LLM to write the first draft, and improve it)
  • To automate tasks — (so much easier to just ask LLM to write scripts than writing one yourself. See some of my experiments in writing automation scripts below)
  • As an API reference — (all the time. eg: show me an example of using urlencode in Python)
  • As a search engine — (all the time!)
  • To solve one-offs — (this is similar to “To automate tasks” above)
  • To teach me — (this is same as “As a tutor” above)
  • Solving solved problems — (this is similar to “to simplify code”, especially useful rewriting gnarly stuff)
  • To fix errors — (all the time. “explain this error” and solve it.)

My automation experiments:

And some programming uses of LLMs:


This article — q What do I title this article? is another hands-on illustration of using a local script to localize (if you are a programmer, you are likely spending a lot of time on the terminal) the use LLMs for coding adjacent tasks.

One from Simon Willison — django-http-debug, a new Django app mostly written by Claude