My Cursor Impressions
Over the last few days I had the chance to build an iOS app from scratch using Cursor. I want to share the experience I had and the challenges I ran into.
The Mesmerizing Creation Process
One of the most impressive aspects of building something with AI is that it automatically constructs the app’s building blocks right before your eyes. When you describe a navigation structure, a simple algorithm, or an object relationship, seeing it materialize within seconds is genuinely impressive.
The speed it provides, especially when doing rapid prototyping on the UI side, is a huge advantage. For people like me who aren’t very good at design, these prototypes are life savers; they turn straight into production designs. :)
On simple tasks, for example things like “put a list here, go to the detail page when tapped”, the first results it produces are usually good enough. With small adjustments you can use them right away.
Working on Existing Code
When it comes to working on a codebase that it wrote earlier or that you developed, things get a little messy. That is because the system can’t fit the entire web of relationships in its head. At some point it can misunderstand the command you gave and break another structure, or it can suggest faulty designs.
Here you may need to talk for hours and explain the context over and over. As I shared before, AI is still far from producing a programming systems product.
https://x.com/urklc_/status/1922362846479855918
Instead of sitting down and making it do everything, I preferred to move forward by getting help where I needed it. It really speeds things up, especially on boring, repetitive tasks or in moments when I ask “how should I do this?”
The Overlooked Emotional Void
There is one more topic beyond the technical side, and I think it is the most important part of this whole thing: emotional satisfaction. I noticed that the sense of achievement and the joy of creating that I used to feel when I built a product from scratch on my own has dropped significantly. After a certain point, it becomes harder to say “I made this” about the resulting product. That is because machines do most of the work. Maybe, as the inevitable result of this era, we will all become a little more machine-like.
Conclusion
Right now I have a huge app that I need to debug, one that Cursor wrote and that the AI couldn’t solve from memory. These few days of experience contributed a lot to my understanding of how these tools should be used. By avoiding writing code and getting our hands dirty, and without trying to understand the app being built before our eyes, that is, by leaving the AI on its own, we are still a bit far from building a reliable, solid app.