July 2026 Update on Iron Arachne

Iron Arachne has been undergoing a lot of changes this year. It's the first year where I really started experimenting with using large language models for coding in general, and it was inevitable that I would try using them with Iron Arachne. The results have been mixed, as anything with AI is going to be. On the one hand, I have massively increased the number of unit tests for the code base, which uncovered quite a few bugs that I didn't realize were there. I was able to rewrite entire subsystems in an hour or two that would otherwise have taken me weeks. On the other hand, LLMs are really, really bad at writing prose, and I am still trying to find and replace all of the AI-generated garbage text with human-written text. "Go fast and break things," indeed. Of course, there are all the ethical and climate concerns about AI and AI providers. Balancing the need to keep current with software development technologies and the moral imperative to not, y'know, wreck the planet is difficult.…