Most relatable thing I've seen an AI say so far, to be honest.
@lizardbill Reminds me of this conversation I had with GPT-4o mini
@Jonas @lizardbill So, it pretends to run Python for the proof but then just fakes the output? (I see this correctly, don't I? It's a dictionary with multiple instances of the same key.)
@Jonas @lizardbill This isn't just "hallucinating the response", it straight up manipulated the output in order to gaslight you by adding an illegal index that would add to 3 ns. What the hell.