I've been spending spare moments putting together a sort of musical experiment - it doesn't feel coherent enough to call it an instrument yet - written in Scheme using Spritely Goblins and David Thompson's Chickadee framework. I do not know if either of these technologies is the best for the job but I've been having fun anyway.
The vague question I started with was: if you wired up a bunch of neurons which could trigger one another, and put a beat into one of them, could they make interesting rhythmic patterns? And at the same time I was reading about Goblins and thought that the neurons could be modeled as objects. I picked Chickadee because one of the Goblins examples was integrated with its event loop.
The code is here but it's not very good and, as I expected, the timing is all over the place.
I haven't been daily-linking as much as usual but:
Phil Agre, "Toward a Critical Technical Practice: Lessons Learned in Trying to Reform AI" (1997) I'm getting fairly sick of LLMs but I recommend you give this essay from someone who grew disillusioned with an earlier generation of AI researchers a go.