One of the things I've been working on at @cursor_ai is beefing up Cursor Rules. We want Agent to be as powerful as the most knowledgable person on your team. Here's how we use them at Cursor. 🧵
Project Rules can be auto-attached to any composer/agent request based on filename globbing, or the agent can grab one based on its description. These rules are checked into your codebase so anyone can make use of them.
Project rules can be very detailed. The first time i had to write a custom editor UI it was difficult to figure out how to wire it up. But I put that knowledge into a rule and now i just ask agent "Make a custom editor UI that does X"
They can also be simple. We use SolidJS instead of react, but the AI would often give us React code. We can't use the original .cursorrules which get included with every request, especially since our dashboards _do_ use react. This simple rule gets included with all front-end tsx files, and i haven't seen unwanted react since.
Some of our code is auto-generated, and if the AI modifies a .proto file, it used to need human intervention to generate the code before it can continue working, with a simple rule added to all .proto modifications, it can just keep moving.
We have a _ton_ of ideas how to make these easier to author, easier to use, more powerful, but we're already seeing our agent do things it definitely couldn't before. More to come!
also we're hiring
anysphere.inc
Share this Scrolly Tale with your friends.
A Scrolly Tale is a new way to read Twitter threads with a more visually immersive experience.
Discover more beautiful Scrolly Tales like this.