If you find ChatGPT useful, pay just one month of Claude AI and use both. I bet you'll discover that Claude AI can help you a great deal more. If that will be the case, you gained a much better AI assistant. Otherwise it's just 20 bucks... For me the difference is huge.

I have used Claude(Free) for coding and architecture-related questions and found it pretty solid but I found chatGPT a generalist LLM that is good at many things. Is Claude the same? Or you found Claude better for coding related tasks?

or maybe through perplexity he can try both (and more)

I think both are really capable. However anecdotally, I notice myself doing way less of the "annoyed, cancel response, extend prompt instructions, regenerate" cycles with Claude as compared to chatgpt.

Also, I am not sure if it has my writing style preferences saved somewhere, but Claude uses the terse style I prefer to read explanations in, without me needing to mention it. ChatGPT can do it too but it needs to be prompted lest it do the whole "repeat question, bullet points, summary" ceremony. ChatGPT gets really annoying when you have do the "slightly more detailed, slightly more terse, there we go" ceremony as well. For similar reasons, I prefer it for code.

One place where ChatGPT absolutely takes the cake though is mathematical proofs heavy with symbols. Nothing else comes close in my experience. It's also better at latex typesetting (seems to be more fluent with common packages like amsmath).

I was in the same boat. I gave Claude a try, just to see what I was missing, and I ended up never returning to ChatGPT except to generate pictures (which Claude can't do). Give it a try for a month. It's worth a shot. They have improvements beyond just generally being smarter. For instance, Claude can _edit_ a prior code block rather than just emitting a new one. It can build React components and show them to you, rather than just giving you the code.

We use both, in my anecdotal experience it depends on what you use them for. The engineering teams which build solar plants use them as “an extra set of eyes” for increased quality control, and they chose ChatGPT. It discovered something to do with cable sizes and efficiency which ended up saving/making a lot of money. I’m sorry I can’t be more precise than that, but I know they had a much better result rate with ChatGPT. For my team which is software development, we favour Claude. It’s obviously rather anecdotal but it has been much better at Go than ChatGPT. For other languages like Python I don’t think there is any real difference. If you feed them both the same prompt they’ll often give you an extremely similar output, and neither will consistently be better than the other.

So I think it’s mainly to work with the one which works for you. I’ve seen some of our more junior developers first use Claude and then ChatGPT to refine their output, or the reverse.

For writing code, I've found Claude far better at conventional/framework based code. I've found ChatGPT better at reasoning about more complex/nuanced code.

As other commenters have said, that is just my personal opinion, I have zero evidence to back that up lol