The technique of taking random input and putting it in a prompt where you insist it means something is actually pretty good, helps a bit with the problem of LLMs being very uncreative.

This reminds me of the thought experiment ‘‘the infinite monkey theorem’’ Give an infinite of monkeys a typewriter and infinite time, sooner or later one of them will write the complete works of William Shakespear. The Dog coding is more chaotic neutral than useful. Its been done now and I suppose the owner of a crow or a raven will make a keyboard and see what it is able to deduce. I saw someone on youtube try to ‘‘teach’’ an octupus to play piano with mixed results and the dog via Claude with help has produced something originl, Its unique not really useful but as an experiment fun. I applaud the effort.

This seems similar to my experience trying to suss out actual requirements from various incompetent/incoherent PMs. Gibberish in, software out.