I tried Claude, but went back to ChatGPT

So, I got a bit burned out on all the long bullet lists and the ‘and honestly?’ ceremony by ChatGPT, and switched to Claude for a bit.

At first it felt like a breath of fresh air; shorter responses, natural conversation. Claude pushes back more, asks questions to guide you into considering things. It genuinely felt impressive.

As I started to use it more though, especially for technical discussions, I noticed Claude hallucinates way more than ChatGPT.

So how how usually interact with ChatGPT is that whenever I ask something, I make sure I check "Web search". It then reliably consults online resources, first. It’s a strong mode I can switch ChatGPT into. With Claude, "Web search" is already on by default but Claude uses it *way* less. Completely unreliable during a session.

So I was discussing the nuances of some DDD topic, and was put completely on the wrong track a bunch of times. I had to correct Claude (telling it explicitly to look it up online), instead of _it_ correcting _my_ assumptions. And that’s only because I already knew some of the details from earlier research. I never feel like that with ChatGPT.

ChatGPT may be a complete bore and antagonising at times with the sycophancy and the comfort-ceremony, but I can count on it to be more sceptical of its own knowledge, as it defaults to online resources more often. It’s slower though (especially in thinking mode), but so much more useful to me.