Hello,
I am completely new to Cursor AI. I am starting to use it and it looks amazing.
However, I am a bit worried about this: is CursorAI able to handle large projects? When the project gets bigger, does CursorAI cope with complexity or does it generates more and more hallucinations/errors?
I am asking this because LLMs have sometimes a hard time to deal with large contexts, right?
I’m using cursor on a 460k line .net application with two different .NET languages, xUnit, JSON, images etc. It’s a monster.
I was surprised but cursor can index the entire project and analyze it very accurately .
I use the paid Anthropic Sonnent 3.5 API access when I need cursor to dive deep in the code and it works great.
Cursor can handle huge projects, I was very surprised, it works better than I expected. It’s a huge time saver and it tells me what parts of the application do when I’ve never even seen the code. I’m new to the codebase and cursor has been a massive help.
I recommend it for large projects but I would use Anthropic OPUS or Sonnet or even the preview open AI o1 model when digging deep into large projects looking for specific logic.
Generally the free models that come with a cursor subscription are enough but if you’re a professional developer paying $2-$3 for fast access to the most accurate models is worth it. You can toggle the advanced paid models on and off as you work to save money.
I’m just a customer but I highly recommend cursor even for huge codebase. It’s well worth it.
Hi @pland ,
If you are new to AI coding assistants, I would add to also be mindful of the tips pointed out in this video (not specifically related to Cursor):
The video covers things like:
Principal 01 - Set realistic expectations for AI
Principal 02 - Embrace iterative development
Tip 01 - Prepare your coding assistant for tasks
Tip 02 - Specify what you don’t want the assistant to do
Tip 03 - Fix bugs by showing the assistant the error
Tip 04 - Work in small, focused units
Tip 05 - Always review code updates
Tip 06 - Choose and master a particular AI toolset
Tip 07 - Think ROI when considering AI costs
Personally, I stick to trying to complete small tasks, one at a time.
If I try and do more than that (which I sometimes do), I end up regretting it because there are just too many places in the code that I have to mentally review to ensure I have taken the right approach. This can lead to being overwhelmed because you are literally doing 10 things at once and possibly introducing code and conventions that are not as ‘perfect’ as they need to be to ensure congruency with the rest of your codebase. You then have to go back and undo things to get back on track and more carefully introduce code that sits well with the conventions in the rest of your codebase.
And here are some Cursor-related usage tips that might be of assistance when first starting out:
Thanks a lot for your quick and comprehensive answers! Very helpful!