I have been a religious Cursor + Sonnet user for like past half a year, and maybe I'm an idiot, but I don't like this agentic workflow at all.
What worked for me is having it generate functions, classes, ranging from tens of lines of code to low hundreds. That way I could quickly interate on its output and check if its actually what I wanted.
It created a prompt-check-prompt iterative workflow where I could make progress quite fast and be reasonably certain of getting what I wanted. Sometimes it required fiddling with manually including files in the context, but that was a sacrifice I was willing to make and if I messed up, I could quickly try again.
With these agentic workflows, and thinking models I'm at a loss.
To take advantage of them, you need very long and detailed prompts, they take a long time to generate and drop huge chunks of code on your head. What it generates is usually wrong due to the combination of sloppy or ambiguous requirements by me, model weaknesses, and agent issues. So I need to take a good chunk of time to actually understand what it made, and fix it.
The iteration time is longer, I have less control over what it's doing, which means I spend many minutes of crafting elaborate prompts, reading the convoluted and large output, figuring out what's wrong with it, either fixing it by hand, or modifying my prompt, rinse and repeat.
TLDR: Agents and reasoning models generate 10x as much code, that you have to spend 10x time reviewing and 10x as much time crafting a good prompt.
In theory it would come out as a wash, in practice, it's worse since the super-productive tight AI iteration cycle is gone.
Overall I haven't found these thinking models to be that good for coding, other than the initial project setup and scaffolding.
I think you’re absolutely right and I’ve come to the same conclusion and workflow.
I work on one file at a time in Ask mode, not Composer/Agent. Review every change, and insist on revisions for anything that seems off. Stay in control of the process, and write manually whenever it would be quicker. I won’t accept code I don’t understand, so when exploring new domains I’ll go back with as many questions as necessary to get into the details.
I think Cursor started off this way as a productivity tool for developers, but a lot of Composer/Agent features were added along the way as it became very popular with Vibe Coders. There are inherent risks with non-coders copypasting a load of code they don’t understand, so I see this use case as okay for disposable software, or perhaps UI concept prototypes. But for things that matter and need to be maintained, I think your approach is spot on.
Have you found that this still saves you time overall? Or do you spent a similar amount of time acting as a code reviewer rather than coding it yourself?
Do you have any Cursor rules defined? Those tend to control its habit of trying to go off the rails and solve 42 problems at once instead of just the one.
What worked for me is having it generate functions, classes, ranging from tens of lines of code to low hundreds. That way I could quickly interate on its output and check if its actually what I wanted.
It created a prompt-check-prompt iterative workflow where I could make progress quite fast and be reasonably certain of getting what I wanted. Sometimes it required fiddling with manually including files in the context, but that was a sacrifice I was willing to make and if I messed up, I could quickly try again.
With these agentic workflows, and thinking models I'm at a loss.
To take advantage of them, you need very long and detailed prompts, they take a long time to generate and drop huge chunks of code on your head. What it generates is usually wrong due to the combination of sloppy or ambiguous requirements by me, model weaknesses, and agent issues. So I need to take a good chunk of time to actually understand what it made, and fix it.
The iteration time is longer, I have less control over what it's doing, which means I spend many minutes of crafting elaborate prompts, reading the convoluted and large output, figuring out what's wrong with it, either fixing it by hand, or modifying my prompt, rinse and repeat.
TLDR: Agents and reasoning models generate 10x as much code, that you have to spend 10x time reviewing and 10x as much time crafting a good prompt.
In theory it would come out as a wash, in practice, it's worse since the super-productive tight AI iteration cycle is gone.
Overall I haven't found these thinking models to be that good for coding, other than the initial project setup and scaffolding.