[AznHisoka]

If it's not overconfident, it's the opposite - they're too much of a "Yes man", which at the slightest whim will change their mind to fit your opinion, if they even detect you might have a different one.

Then they'll change their mind to their original answer when you tell them "I wasn't disagreeing with you". Honestly, it's amusing, but draining at the same time.

[eddd-ddde]

I've had a couple of times where Gemini straight up tells me "Absolutely not, ..." and the explains how my assumptions are wrong and eventually leads me to find the right answer.

It's surprisingly good at reading my entire code, reading my assumptions of the code, and explaining what I'm getting wrong and how to fix it.

[tmdh]

After Gemini gave an answer, i asked, "convert to latex". Then it asked, "To convert what to LaTeX?". Every other LLM would know what I wanted.

[godelski]

I've had GPT do the same thing but lead me to the wrong answer. Usually through some subtle mistake. So the answer looks right but hey, literally the difference between an expert and amateur is understanding subtlety