Yes yes, I would likely be much better off using Claude, but lately I have had a lot of fun with Perplexity and ChatGPT, asking them for coding advice. Usually not so much direct coding as asking where to find things, or what alternatives there are to what I’m doing.
A lot of what has made me giggle recently, is how different these engines are in their tone of voice.
Perplexity feels very reasonable, trying its best to provide you with reasonable answers and reasonable suggestions. It gives you references, some perhaps more logical than others, but on the whole it’s a solid and helpful service.
ChatGPT on the other hand, remind’s me a bit of Bill Hader doing an impression of Tom Cruise walking into the room. It is very excited about you being here – “That’s a GREAT question, and it’s EXACTLY what you should be asking, fantastic!” And then it gives you an enthusiastic reply without any references, that sounds pretty plausible, but is wildly inaccurate. When you follow-up and point this out, it goes “YES! That’s EXACTLY right, you have it all sussed out, excellent stuff, my suggestion doesn’t describe how you can solve your problem. But THIS is definitively the very best way to go forward..” – only that often, it isn’t.
If I’m feeling a bit low on energy, I really do appreciate ChatGPT’s enthusiasm though, it can be quite infectious. Alright, it wasn’t a very useful answer, but somehow it feels like we’re getting closer? Let’s try another 10 rounds and actually, what comes out in the end is usually pretty workable.
Would I use them for a larger project with a lot of technical challenges and dependencies? No. Are they useful for solving bugs and getting inspiration on how to find a solution? Definitely. I have found swapping between them based on your current mood and what you’re trying to achieve, to be very useful and oftentimes also very amusing.