We had a group of our people using Cursor. I wanted to be agnostic — that's my approach. I didn't want to dictate to the team "you must use this tool" because at the time, and this is two years ago, GitHub Copilot was still pretty popular, there was Cursor starting to gain momentum, Windsurf wasn't out yet.
There were a few on the team using other tools and my stance was "You can use whatever tools you want but I want to see you implementing things".
We run a brown bag every three weeks. We get together and we go, "Hey, what's cool that we've been doing from a tech perspective? What's changing? What are we working on?"
So not a product related discussion. It's very much tech-related. For a few of those, it was all about "Show us what you've learned about Copilot. Show us what you learned about Cursor."
I had a couple of mates saying they've been able to rebuild legacy code bases in new stacks purely using prompts. So one of our brown bag sessions we ran with that.
We extended the session. We spent a half a day coding where I had everyone — we've got a student-focused team, a university-focused and an enterprise-focused team — grab a section of legacy code and told them "Use whatever tool you want, try single-shot, multi-shot, high context, low context, long prompt, short prompt, just try things and see what we get."
And the results were pretty amazing.