Make the cursor unlimited. What’s the point of limitations if the courses with their modules and the AI doesn’t really understand what is asked of it. I’m tired of scolding it and explaining it like a baby.
Hey, thanks for posting!
I feel your frustration. Generally speaking, you’ll find those kinds of difficulties prompting (scolding, explaining to it like a baby) aren’t usually down to the IDE, they’re down to the LLM on the far end. Those models are run by Anthropic, OpenAI etc. rather than Cursor.
We’re certainly all eagerly anticipating a future where those models achieve flawless performance, but the current cutting edge of codegen AI just isn’t there yet. It gets better with each new GPT or Claude model, but it means sometimes there’s a real art to guiding the models to the output you want. It definitely takes practice, and even more so the larger the codebase.
That being said, there’s definitely ways to improve your skill at handling them and get incredible results. I’ve been an end user of Cursor for maybe a year and a half, and other AI-assisted coding for a lot longer. I’ve definitely become much more adept at model-steering than I used to be.
You mentioned scolding the model – you might find some value in taking a look at Why Patience is a Virtue (or: How to Be Nice to Your LLM), there are some tips for why keeping your cool can actually improve model performance.
Were there any specific limitations you had in mind, beyond the fact that the compute power running the models costs money?
1 Like