One of the things about LLMs for hobby coding is that they can take a sort of computer addiction and amplify it 10–100x. I enjoy tinkering with computers to do various tasks in my life. But I’m not a professional software engineer, so creating something new often involved a lot of searching around on the web and a lot of trial-and-error experiments. There’s a self-limiting quality about that because it is a cognitively fatiguing task. Using Cursor¹ initially, and now codex, it is so easy to get things done that I end up doing more rather than less. And that, of course, means more to keep up with.

For example, I like nicely formatted screenshots for my project repository README files. One of the standard tools for this is Shotcandy. Since it is open source, I wanted to host it on my own Debian box for privacy reasons. In under five minutes, codex had a custom Docker image set up for me to run on my server.

This new capacity is welcome. But it comes with a risk; it’s a risk related to overload. I find myself doing so much that I’m not leaving enough breadcrumbs to follow. I need to create a rule for myself that stipulates that I document every project in my PKM before moving on. Even just a word about what port the application is being served on, how to build it, and a link to the repository could suffice. Just a way of slowing down and documenting.