Chat GPT Coderesults better with procedural languages or with functional ones?

[qikInNdOutReply]

With tests written first, or with specifications written first. Is using the most popular ecosystem a garant for the highest chance for success? Or would it make sense to use the prefered language concept of your field of expertise?

I see this promptr thing and it blows my mind.

[dtagames]

I can only speak to my experience with Copilot, which I use more for daily coding that ChatGPT. But there, I'm on an edge case environment: Typescript with web components and Lit. And, I write in functional style as much as possible.

Since Copilot looks at your own code as well as its LLM, it's very good at copying your style after you have opened some files for it to dig into. I'm amazed at how much it offers to complete or copy my functional style, despite many examples that must be in the corpus that don't look like mine.

I would say never let an AI force you to deviate from your style. The most important thing is that you understand the code. This is even more important than having it run. The reason for that is it might not always run, and then you really want to be able to understand it.

[verdverm]

I would look to the distribution of languages it was trained on. The more the better, which from my understanding, ChatGPT does best with python & JS/TS.

Again, this will be a model dependent answer.