I do wish the big llm companies would put even a modest amount of effort into teaching the basics of usage. Fairly sure the parts of this agent thread we aren’t seeing will show he made an (incorrect) assumption, asserted it to the llm, then asked it to explain. Hence hallucination.
If only they could structure things so that the user who trips ends up falling into a pit of success 😅
It’s so hard tho given how they work, tho who knows with a different system prompt. The thing that people don’t understand is their sycophantic nature, and to a lesser extent how they are bad at anything deterministic.