what about that text made you think it was AI? i spent about 4 minutes composing it.

i don't use AI to write anything other than commit comments and sometimes start with writing documentation by getting an AI to build all the structure.

anyway, if you think i use AI to write, first of all, it's LLM, second of all, i may not know as many facts as Claude or Grok but i have the ability to process information they struggle to process without the assistance of systematic reasoning processes to augment their text generation. this is also why some 37% of AI content is hallucinations, why most AI images look fake (again, probably because of the lack of systematic reasoning that includes causality and checking against similar real images or texts).

i sat there watching Junie struggle to implement a very simple algorithm that intersects a series of separate results for tags when there is more than one tag in a query, it spent an hour, over 3 attempts, and failed to do what i intended.

i wrote the code in 10 minutes that worked first time. but i will concede that probably watching the AI fail gave me an idea about how to actually do it, before that, it was quite inchoate in my mind.