What I'm doing with AI these days

06 Nov, 2026

A while ago, Trevor McKendrick asked what people are using AI for these days and I answered him. Here's a follow-up.

1. I've forked a WebAssembly fork of Helix, my favorite text editor, and I'm hosting it for my own use on the Web. This is going way better than I expected it would: I can port only the features I need, AI is great at fixing bugs, and latency is low. There's much more to say about this project, but for now: I'm actually using it, it actually works as intended, and it's magical.
2. I've been trying to learn Galois theory properly for over 20 years now. I could pass the tests in college, but I never quite got my head around it. I've tried many approaches to learning this, including several different textbooks, but some combination of limitations (certainly including misunderstood foundations) has held me back.¹ The amazing progress in AI-as-math-tutor is, therefore, both a chance to work on a decades-long goal and a test case for the actual efficacy of using AI as a math tutor. It's going well so far, but I'm still early in the project.
3. A lot of my note-taking and personal management involves tracking lists, whether those lists are of tasks, books, favorites, or something else. I've written custom list-management software for this, and although it has some rough edges, it's the best list manager I've ever used for me.
4. As AI gets more and more useful at document review, I find myself more reflexively putting files through it for feedback. Nothing I'm doing is qualitatively different, but the experience of it is very different, both because I feel much less motivational friction about it and because the results are so much better.
5. I'm now using my own clipboard and snippet manager; I still love Alfred, but there are some specific ways I want these tools to work, and no extant tool covers all of those.

The biggest difference over the last ten weeks (my last AI report) is that my personal-software projects are more ambitious and more successful. There are things my own software can't do as well as the mainstream SaaS alternative, but more and more, other things matter more, and the environment for making my own tools has gotten better:

1. It's much faster and cheaper to improve rough edges;
2. I don't need most of the features the mainstream alternative has;
3. There are specific behaviors or features that the alternatives do not have and that I badly want;
4. I've settled on a standard set of tools and choices, which has significantly lowered the cost of testing a project like this;
5. The LLMs have, of course, improved;
6. I have also, hopefully, improved;
7. At least, I've gotten closer to internalizing the idea that I should be trying more, and more ambitious, software experiments.

#generative AI #personal #software