Thoughts on AI in programming (June 2025)

I’ve been pretty indifferent to how AI usage has been promulgating across my profession over the last two months. AI raises the floor, not the ceiling of quality (I believe I operate closer to the ceiling), environmental usage concerns (heavily in debate), general distance from where the hype is happening (I got off twitter 6 months ago), and honestly the field was moving too fast for me to believe I could catch up1.

I was wrong to take this stance

Earlier this year I decided to work on my perceived reliability and I feel like I’ve been doing a pretty decent job. Recently, I’ve been thinking about speed. It’s undeniable that AI is really good at writing boilerplate, and sometimes my job is lots of that. Lately, I’ve also been switching contexts much faster which it helps with. This means I can finally get to that long list of side project ideas that are mostly pipes.

I really enjoyed thinking about problems tech could ease when I was younger, and somewhere along the way that joy died because I struggled with the tech too much. Maybe with my curiosity on how things work comes back, some more of my ability to think about problems that can be solved with tech will come back too!

Next steps

  1. I hear agents coding AIs are a thing now? I need to set one up to run locally and explore MCPs
  2. I need to get better at reading code. I don’t know what metric to measure this by - help? I don’t know how I fare but I feel like I can always get better.
  3. I want to practice TDD by default - it builds better apps, and gives better prompt responses.

Footnotes

  1. Look at the number of updates on Simon Willison’s blog! ↩