Why do this?

The ability for agents to do good work is highly surprising. From the time ChatGPT had its coming-out party, I doubted their ability and potential, to my own detriment. A few months ago, I had the "aha" moment while trying out working with the newest versions of some models, realizing I had just written a "little agent" that could automate most of the workflow of my own job.

If you can't beat 'em, join 'em, and here we are. This is a collection of experiments and ideas that I've been having my machines code up for me. All of these projects are vibed-up, and I only understand the broad strokes, but they accomplish at least some of the goals I want them to achieve, and I find that continuously impressive. Impressive enough that I want to give them a gallery to live in.

The set of projects mostly came about because I had no idea what task to set my new colleagues to attempt, and so I fell into various meta-rabbit holes about what I could get them to do to try to help me find out what to get them to do.