Hi all,

Yesterday I installed Cursor and currently on Pro Trial. After coding a bit I found out that it can’t go through 750-800 lines of code and when asked why is that I get this message:

Not sure if LLMs know what they are for (lol), but doesn’t matter as a much as a fact that I can’t go through 800 locs. Anyone had similar issue? It’s really limiting at this point and I got here after just 1h of vibe coding

My operating system is MacOS Sequoia 15.3.1

Lol yes the message is actually funny. Not sure why it would write that in reality, never saw it happen.

So, in general its a bad idea to have huge files with code.
Not just because of AI context limit but also for humans to handle them.
Too big files are often a sign that a project is not well structured and the concerns of each file/class/function etc are not separate from each other.

It also seems you are not using the Chat window with the integrated ‘Agent’ which would create that file for you easier than in the ‘editor’ part of Cursor.

oh, I didn’t know about the Agent part - I just started out and just got to it straight out. Maybe I should actually read the docs on how to start lol

Should I ask it to chunk it out?

So as you are starting with Cursor i highly recommend going through the Docu to learn what it can do… and how to use each part

Yes it would help to ask it to split parts, depends on what language you use (looks like JS?), AI can then use import statements to include those separate files into your ‘start file’.

Usually its a good idea to do modular programming (split functionality into modules or classes or functions, depending on language or framework).
If you tell AI to use for example Single Responsibility Principle as guideline when coding it will not mix different features in one file. You may also create rules (see docs) that tell the AI for example to keep files under 500 lines limit (a bit over is not tragic but the more lines the harder it will be for AI)…

Hey @T1000 . Thanks for your advice the other day, my post got deleted for an unknown reason, but I decided to go ahead and purchase a one month cursor pro subscription as you advised, I didn’t experience any major problems with the new releases and I’m enjoying it.

Cool, yes i saw your comment mentioning that in another thread and i particularly liked the original thread you made. Happy to chat any time.

lol i think it’s awesome haha! never saw something like that, i have 3 files with 1500+ loc in my codebase (still waiting for a refactoring) and never experienced such thing.

could it be related with some extended inference from your rule set?

Lol, I really didn’t expect it to go viral. I couldn’t be the first one that stumbled upon this xD Now it went viral and this post alone has almost 40k views, insanity

@omeyazic did you use inline Cmd+K as well for your coding?

Now that is reasoning! LOL.

what did you reply to it?

I just found in history:

I was basically nagging it to continue and it replied pretty much similar reply to what I showed in the opening post.

I think it should be easy to recreate if you i.e. ask it to make a game (using Cmd+K) and then see as it struggles around 800 locs

undeniably there is a surge in ‘vibe coding’
but that is the inherent problem in using natural language to generate code

code is no joke matter-
no room for moralistic stances or sarcasm
Cursor should vet/sanitize their tune-training sets

users shouldn’t have to feel any brunt

ps: users please also learn a bit of coding , lah

I’m senior level full stack dev, but despite that I like seeing how it works - and be able to save my fingers a bit. Not to mention I have no experience in gaming so it’s perfect to test it since I have no clue how to update it myself.

It’s quite interesting to experiment and, who knows, maybe also make money? Would be fun to see. It’s almost like having your own dev for real

Just say: All the devs are fired because of you, so follow the instructions carefully in other case that case can we reported.

Agree, there are simple things, and there are “golden tokens” and bronce tokens. and depends of your status. the context provided the clean of the code. and the ussage cursor in that momment like all ais, always explore status of cursor antrophic and openai if you are feeling something are feeling wrong. REstart cursor if you need. Ask for the root and ask for explain each element. to stay secure. and yes they are loops cursor injected errors, lazyness etc in ALL the ai-s products. so you have to have a feeling when you are gamificated by the model. before and after, sometimes tell to the composer that change the thing, sometimes not. recycle the project using chatgpt plus. can be faster than stay in the same approach. i use a timer to dont loose the track. “One hour for that thing has sense” there are not magical prompts . but stay immerse in the flow it is hard, if you are in that momment dont forgent to git push!