Character limit resulting in things being broken

The character limit is so frustrating, I’m trying to use ChatGPT to help me with some coding issues and it constantly stops typing half way through, this has been a consistent issue for months, and simply typing “continue” doesn’t work, it will constantly rewrite the same script over and over, or start writing from the wrong place, it will miss code etc because of it

Can we get an increased character limit??

2 Likes

Based off of the news that broke today their main and #1 priority the past couple months has been focusing on stability and uptime issues which is understandable. I doubt that includes increasing character limit before they addressed those problems. That also means though that I think they hear us on that so I say give 'em a little more time.

In the meantime, I suggest considering the following with each prompt, not just each session:

  • If you want it to continue where it left off, try something like this:

Thanks. That’s a great example! I read the whole thing. You should too! In particular, look at the last line, where it got cut off. It’s like a mystery. Ooh! I know! Why don’t you look at that last line, where you talk about [code snippet starting at the last comment or function or line or whatever]. Then, as if you had simply stopped typing that line to scratch your nose and were simply resuming where you left off, do your best to generate an accurate example of what the rest of the code would look like. Begin the code block with this line please: [code snippet starting at the last comment or function or line or whatever].

  • Each input from you will consider context from that session up to a certain point, but will only follow explicit instructions from your current input without training. To that end you should consider typing your messages to the bot in a separate text file with a template in it that keeps some instructions for the bot at the top so that those are constantly reinforced every message.

  • It’s not an AI. It’s just a language model. You need to be explicit and clear as if you are explaining all of the component parts of the task to the model for the first time. The model is also very touchy with semantics; you may need to phrase things in such a way that it gets what you want. Often it generates nonsense code not because it’s bad at coding, but the people didn’t take time to standardize certain fundamental pieces of context in the input for the language model such as naming convention, library specifics, even that it needs to have proper syntax.

  • If it is not giving you what you want, tell it what you want to end up with. Don’t just say “I want to do this”. If I find myself fighting with a session that I don’t wanna restart, I stop whatever I’m trying to make it do, and ask it the following:

"Hey ChatGPT, I’m having trouble getting you to do x. I know it is within your capabilities and I believe I am holding back our progress by way of poor communication on my part. Can you help me by telling me exactly what information you need in order to produce the following output? [Precisely describe your end-user experience]

It’ll tell you what to do. Take it’s suggestions, feed them back into it. It actually loves that and it can improve its work with you down the line as it becomes a bit more fine tuned that way.

  • I always make sure to use consistent naming conventions and be descriptive when speaking.

  • Use regenerate a few times before correcting the bot for not following a prompt. If you correct the bot, which you should do vigilantly, you need to ensure you’re doing so correctly. I’m guilty of losing my cool with it, but you’ll only clog up your valuable context with useless abuse that will potentially ruin future output. Corrections should look like this (granted take this with a grain of salt, I’m not an authority on anything, this is all personal experience):

Thanks for that. It looks like you weren't compliant with your input there. [Point out the error] - We all make mistakes, that's okay! Why don't we try again? This time, do not do [the thing```].

Idk if any of this is useful to your use case but hopefully it helps. Best of luck in your project!

1 Like

Unfortunately that doesn’t really work most of the time, thank you for the response though

Check this blog post ChatGPT Character Limit, I resolved similar error using the this guide.

Hope this help!