ChatGPT is a language model, it’s not intended for code and you’re using it “off label” at your own risk. It can produce working code, which is impressive in itself, but in order to know if it’s decent code you still need to be competent with that language. I had someone run a few prompts for me a while back, it ignored central parts of the query, and its output was basically like a very junior developer - fair enough, but not great or even that good.

Potentially useful, but if you expect it to be more than one part of the “process”, you might be setting yourself up for trouble.

Edit: just like it’s not a coder, it’s not a search engine or knowledge base, either. It just knows language and what seems like it ought to follow a given phrase. Be very aware of this difference, because sometimes it spits out 100% falsehoods with the same level of confidence and authority as the true stuff.

I have built several programs with ChatGPT 4 by now. From very basic Python scripts to Python webscrapers and C# in combination with Unity3D.

In the beginning it was much better than it is currently. At the moment context is severely hampered no matter the limit and you’ll be bashing your head against circular arguments and it straight up ignoring stuff you just posted two messages ago.

Trying to troubleshoot code it wrote a few days ago will be a slog and like dragging yourself over nails at times. Here’s what I have found to help and make life better:

1. Be very, very, very precise in your instructions. And keep them saved, so you can reuse them later (point 4)
2. From the very start plan to build your project with small functions that interact (good policy anyway) which makes troubleshooting and changing these functions much easier and will prevent you running into message limits.
3. If it fails to work for you the way you need it, you’ll might have to scrap your entire code and start over with ChatGPT -> Again, the reason for point 2 being very important. Scrapping one functin is much less painful than an entire tool.
4. Start new chats when you feel the quality degrading. Sometimes it helps and since the context is garbage at the moment anyway, it doesn’t matter much.
5. Post the code it is supposed to fix every single time. It will inevitably refer to other code, code hallucinations, etc. otherwise. Again why point 2 is important.

I agree with the other comments that ChatGPT isn’t really that good for programming, it hallucinates often and you end up working too hard just to try and figure out what it got wrong. However, I have found a good AI engine, phind.com, that has started to replace my google searches. It’s just a wrapper for ChatGPT, but it cites its sources so you can verify or dig deeper, provides search engine results in a sidebar and has upvote/downvote options to help it improve. So it feels like a personal google “agent” that runs off and googles something for you and comes back with a concise report.

I quite like GitHub Copilot and use it a lot, but I find ChatGPT not all that useful.
For actual coding, it feels like describing what I want it to do is more complicated than doing it myself.
I can see some uses as a search engine, but I’ve had a lot of bad luck where it suggested code that was plain wrong or not working and often did not even compile, so most of the time, I’d rather look on GitHub, Stackoverflow or sites like that.

I use chatgpt a lot when coding. It’s pretty good and the code is typically usable. But sometimes it messes up hard and it can take a while to realize that. Net benefit though and I’m sure the technology will improve over time.

I’ve used it a bit to get the framework or boilerplate I need. It’s not a one-click solution!
What I will do is ask it to generate code for a purpose and then iterate over the suggested code adding and/or modifying specific areas until i have something usable… Depending on the complexity of the code/feature this can take quite a number of iterations. You need to understand the code it gives you!
Once I feel satisfied I will copy the code to the editor for tweaking and adjusting to my environment.

I will admit that I’ve been surprised sometimes by the suggestions I’ve been given. Sometimes in a good way, sometimes bad.

Remember that whatever you feed it will feed the LLM, so don’t give it anything specific to yourself or your organisation.

It is a conversational tool that can generate decent code if properly prompted but it lacks for the most part enough context. For it to be really useful it has to be able to be trained on my entire project that I’m working with, not just a single file or function.

What I miss is the ability to “chat with my project”. I.e. have the whole project in the trained context, and then reason about architectural changes, pros and cons, have suggestions for refactoring, help with complex renaming schemes and moving code, etc.

It would be super interesting to be able to give instructions like:

• Organise my files by dependency and the logic they implement.
• Or something like, create web components from common input elements in my html pages.
• Where is the user auth code implemented?

Things like that.

I’ve used ChatGPT to answer questions relating to Python. Notably, I asked it how to use QtNetwork to send and receive requests with authentication, as the application I was using did not use any non-standard modules I was more accustomed to like requests but did have PyQt. Not only did it gave me working code snippets but explained it in a way that I was able to understand. No, it’s not perfect. But man it’s better than hunting Google for that one StackOverflow post.

I have heard it trips up on certain less-used programming languages like Swift though, so depending on your use case YMMV. I’ve also not used Codex but a friend of mine has. Apparently it really liked to mention this one specific GitHub profile.

For shits and giggles I asked ChatGPT a while back to represent a Pokemon with a Python class, and it gave me working code. Google Bard would trip up and not use the class when I told it to.

Its good for basic Python scripts and stuff, but not so good for complex programs.

While it’s pretty decent at coding, it’s often (in my experience) either giving you an overcomplicated way, an outdated way or just completely wrong code.

What I do really like is that when you ask for a snippet you can easely ask for variations, like: “make it a bit shorter”, “add comments” or “don’t/do use library X”. This allows you to quickly get a few variations that allows you to come up with something of your own.

On a slightly other note, I’ve been impressed by github Copilot on this subject. While still often wrong, with small things it almost feels like it’s reading my mind while programming.

It’s certainly helpful when you want to try out something new.
For example I recently wanted to make a Firefox AddOn, which is something I hadn’t done before.
So I asked OpenAI how to do it and it talked me through it step by step.

Basically it allowed me to google less, because I could just ask ChatGPT, as it was faster.
Some info is outdated or wrong, so you still have to know what you are doing and still have to use Google.

Also, I wanted it to help me get some data from the DOM, but that was a rather difficult job for OpenAI, it never “understood” what I wanted and just gave me code that didn’t do what it was supposed to do, and even after explaining the problem with the code and telling it what I want, it wouldn’t understand the problem and just give me other bad code that changed nothing about the problem.

So it’s important to understand, that this is just one more source of information/help you have as a dev. It is not a standalone solution that can do your work for you. It merely can help you. The same way as googling, stackoverflow or reading the documentation helps you

ChatGPT 4 is pretty decent at checking code for any mistakes, and it can generate pretty good code if you can describe what you want very well. But sometimes it does give you code with a slight mistake or two, so what I normally do is give it the code in a new chat and get it to check it itself

Used chat gpt 3.x to assist in Godot and gdscript logic for our game… Nearly always wrong but often gets me thinking in a different direction than the one I was stuck in. So certainly value there.

I use it occasionally to write specific functions in python, for example, a compound interest calculator. I don’t trust it to write a full app yet.

Not at all. It often gives bad answers, or workarounds rather than working code. It’s not useful to me if I have to fix its code, when I can do it more efficiently and quicker by my own.

With some hand-holding, it’s quite decent at reading and documenting functions, which is what I use it for since I’m too lazy to document them myself. :P