Kaan · Article
2025-12-04
Six Cursor Meetups and Counting
Years of individual experimentation collapsing into a set of evenings. The edges of AI tooling get mapped fastest in a room of power users.

After having ran 6 successful Cursor Meetups,
I'm still amazed by how much we keep unlocking from that first Ottawa session. Early June,
- 40+ Cursor power users
- 2 demos set up
- a LOT of talking, a LOT of intensity, to multiply our coding efforts through Cursor
- having the Cursor ambassador from Toronto there just confirmed that we were onto something (latest Café Cursor Toronto was outstanding)
Two problems kept surfacing with ai assisted coding, and the solutions were simpler than expected:
Getting AI to respect your design system: The breakthrough wasn't better prompts, it was just dumping a critical mass of components into a tagged directory and telling Cursor to implement "using our established patterns." I figured out that models need a reference it can search when they work on design.
Database context devouring tokens: SQL migrations were too verbose and didn't expose schema relationships. Sukhpal Saini suggested Prism DB files in the repo, suddenly Cursor understood our data model without the models consuming more of our context window on schema dumps.
I built these Ottawa meetups to accelerate a community of builders and my own learnings. After 6 monthly events, the compression results are wild.
Years of individual experimentation collapsing into a set of evenings.
One thing I've learned: the edges of AI tooling get mapped fastest when you lock power users in a room and make them share what's actually breaking.
If you're building at these edges, you should probably be in the room. Connect if you want the invite.