My experience using Cursor
A personal breakdown of Cursor's annual statistics, 2 pet projects in two weeks, and several insights about the future of development against the backdrop of AI.
This image could not be loaded.Today Cursor showed me slides with annual statistics. I expected to see a chart of my titanic efforts, but instead I got a polite reminder that I've been actively using it for only 2 weeks. It was a perfect news hook. I'd long wanted to resume posting on the blog.
Individual infographic
The first slide showed the Cursor timeline for 2025: from version 0.45 in January to 2.1 in November. As a masterful finale, the developers placed an event dated December 1: You started Building with Cursor.
“Yevgeniy, 2025 was a big year. 16 versions, 26 models, bugbot, cloud agents, and 2.0 with composer 1. Ready to take a look back?”
– Cursor
305 million tokens in 2 weeks
The second slide showed a number: Tokens Used: 305,202,271 Next to it was a 2025 calendar where only weeks 49 and 50 (December) are lit up. They accounted for 43 and 143 million tokens
Realizing that all the activity fit into two weeks, was unusual. By feel, I'd been working with Cursor for a couple of months already.. It's probably somehow related to the fact that tasks that used to take days are now solved in an evening.
“You experimented with 1 models”. A solid red line.
The third slide is straightforward enough:
“You experimented with 1 models. And your go to changed 0 times”
– Cursor
First thought: “Am I doing something wrong?”. Cursor seems to be nudging me to use different models, and I'm missing out on something by using Auto mode.
In Auto mode, the communication style may change due to switching models if I understand correctly, but with a precise technical statement of the task, its role is secondary.
What did I manage to work on during these 2 weeks?
During this time I managed to work on two pet projects:
- An iOS app. Backend in Go, frontend in Swift. Coming soon to TestFlight. I finally got a developer account, which allows testing with real users.
- A major update to the ampleev.com website that you're viewing right now. I revived the archive blog, fixed authorization, update subscriptions, and comments. In one evening I created a fairly advanced article editor for this blog. A task like that would have taken a couple of weeks before.
Insights
Of course, everyone has more than enough of them right now, and I'm no exception.
- 100% test coverage is the new standard. For a serious product this is a must-have. It used to be expensive, now it's almost free.
- UX even in an MVP should strive for perfection. If the user winces — you've lost. AI helps bring UX to a perfection that has no limit, hence the original thesis "strive", and not just "should".
- There will be many highly specialized apps. Or maybe not even apps.
- Good generalist product managers will get rich. Programmers aren't really needed.
- We're turning into a “bee swarm”. AI tools are like a collective mind accessible to everyone.
Speaking of this theme, by the way, there's an excellent series, I recommend it:
Instead of a conclusion
For me, this article is a way to systematize knowledge and practical experience + record insights. I'll be glad if the content turned out to be useful to you ;)