@shaoruu - create dedicated "playgrounds" to iterate on the ui/ux or logic with the ai, kinda like a free-roam space until you get something you like, and then integrate it back to your main app - use @ pr and @ commit to review changes & undo things you don't like

@ryolu_ i use cursor to… make prototypes + try crazy ideas and they usually are way easier than i imagined

@baltaaazr i think the most underrated / hidden feature that i really like is resume from here, like the little + buttons after a codeblock i use it when the agent gets stuck on lints, so i can resume from its first attempt which is usually 80% of the way there and continue from there

@ajhofmann18 - don't ask agent to do something until you understand the problem and solution that you want to see. If you need help understanding the problem and solution, iterate in ask/codebase chat first - if agent is consistently one shotting your requests, try aiming higher (do more work in one request). If it's consistently getting stuff wrong, break the problem down into smaller requests - reference the correct patterns that exist in your codebase as the basis for implementing new work

@danperks_ - new chats very often, almost every 5 minutes probably, cmd+n is my most hit shortcut - lean into the linter. setting up a linter to flag what you want the AI to see, but ignore what you don’t can be v helpful for the agent fixing its own code - be scrappy with it. i frequently clone a repo from GitHub, use ask to find out how one area works, then throw it into my own repo to implement

@ericzakariasson - let cursor generate tool it can execute instead when it's faster - setup common workflows with rules that the agent can execute. e.g resolve merge conflicts - be very explicit and provide as much context as possible with @ symbols