I just found about him and his work through this article. Thanks for mentioning that… Also, this piece would be hard to miss though. I got too excited I guess :)
What a post, and a serious amount of homework for me to do. Cursor and Claude Code are priorities for me to learn (and fast)
Question for Tal and Aman (don’t think I can tag you) - does the approach you outline work if I’m also running Claude Code inside Cursor? Or do you recommend keeping them completely separate?
Thank you Nick! This experience was tested mostly in Cursor, but most should work in CC. You'd want to use Ctrl+O (Cmd+O on mac) to make sure that Claude Code is being verbose and showing the reasoning and the tool calls. Claude Code, of course, only uses entropic models as well. So it will be harder to get a sense for different models.
I also prefer Cursor because of the highlighting of the files and the ability to see the changes. It's just better UX that I prefer for pairing. That's a subjective reference.
I think Cursor is a nicer interface for seeing the elements of AI agent threads much more clearly and visually. If that's your goal, and all things being equal, I would definitely choose Cursor.
This "article" creates a whole new paradigm for learning: interactive, memorable, personalized.
Actually going through the steps in Cursor left me feeling like Neo in The Matrix, "Now I know kung fu [AI]."
(Except now I'm learning to create agents instead of kill them...)
Given how much I've learned from you about teaching and making things memorable, interactive, and personalized, this is really meaningful Jack
Excellent work Tal and Aman 🙌
Dude Hamel that means a ton coming from you, we're huge fans of how you operate.
How do you manage to raise the bar every week? That’s really impressive.
Great content! Thanks
@Elena | AI Product Leader — in case you missed this one. :)
Thank you, Nihal! I love @Tal Raviv writing and experiments so much! 🙏
I just found about him and his work through this article. Thanks for mentioning that… Also, this piece would be hard to miss though. I got too excited I guess :)
100 hours — respect!! 😍Bookmarking it to give it the time it deserves. Thank you!!
What a post, and a serious amount of homework for me to do. Cursor and Claude Code are priorities for me to learn (and fast)
Question for Tal and Aman (don’t think I can tag you) - does the approach you outline work if I’m also running Claude Code inside Cursor? Or do you recommend keeping them completely separate?
Thank you Nick! This experience was tested mostly in Cursor, but most should work in CC. You'd want to use Ctrl+O (Cmd+O on mac) to make sure that Claude Code is being verbose and showing the reasoning and the tool calls. Claude Code, of course, only uses entropic models as well. So it will be harder to get a sense for different models.
I also prefer Cursor because of the highlighting of the files and the ability to see the changes. It's just better UX that I prefer for pairing. That's a subjective reference.
I think Cursor is a nicer interface for seeing the elements of AI agent threads much more clearly and visually. If that's your goal, and all things being equal, I would definitely choose Cursor.