Nothing beats these choices when working with Cursor background agents - Monorepo (so the agent can see the entire picture) - Swagger + orval (type safety between backend/frontend) - Vertical slice architecture so its super easy for AI to understand related code, easy for multiple agents to work on different features independantly AND for me to review code DAAAMN WHAT A TIME TO BE ALIVE

If you are still building applications feature-by-feature, page-by-page... you are already behind. The future of UIs is dynamic. Users don't want more buttons and forms. They want answers. "Show me analytics events per day for the last 7 days" SQL generated, validated, executed React component created & rendered Beautiful chart appears in 3 seconds No developer. No sprint planning. No backlog. Just ask. Get answer. Move on.

Flight simulator. In your browser. Are you ready to fly?

Fuck I can't keep up QA testing all the features Cursor Background Agents are building for me When will you guys release QABot @cursor_ai Agent writes a summary of what needs to be tested and the QABot jumps on the task as soon as code is pushed For monorepos that run 100% locally, this should be doable, you already have the VM?

Cursor can you fix this Cursor can you fix this Cursor can you fix this HOLY SMOKES BACKGROUND AGENTS ARE SO GOOD

I use Cursors Bugbot as well as it is always nice to have a extra pair of eyes looking through the code, but CodeRabbit is, at the moment, so much more comprehensive and it feels like it attacks the review from another angle.

I only use this workflow for tasks that I'm certain of how they should be solved. It makes it so much easier to review the code and steer the agent in the right direction For tasks that are more comprehensive or outside of my comfort zone, I do them in Cursor. Still using the agent, I let it research for me etc etc. But I spend more time at these tasks so I actually learn something.

Coderabbit is really nice here. If it finds an issue, it also present a prompt that I can give to the agent. So I copy this prompt and leave a comment in the PR where I at @cursor_ai FIX THIS PLES; {prompt from coderabbit}

Now its time to open a pull request, a simple click in the Cursor UI. And when I do this, Coderabbit and Bugbot instantly starts reviewing the code

The agent executes CLI commands ofc so it verifies my .NET api builds, it verifies there is no errors in the frontend

Then it starts working and Cursor solved this in such a beautiful way. It create a branch for me out of the box, presents all diff both in the agent view so IF I want to truly understand the output, I can stay close

When I get presented the overview I either accept it and ask the agent to proceed OR point it in the right direction and iterate once more.

I pick an issue or feature from my kanbanboard Instruct the agent quite detailed what I want to achieve If I know challenges up front, I guide the LLM in the right direction I ask the agent to present to not write any code initially, I want an overview of the solution. Which files its gonna create/edit etc etc

Haha Cursor Background Agents together with CodeRabbit is insane I've been sceptical to spinning of background agents, mainly because 1. I want to understand what the AI produces 2. I want to be able to be responsible for the output I was wrong, this setup is insane:

Playing around with @coderabbitai for a customer who wants AI Code Reviews but I have an issue. Customer has a lot of files that should not be included in the review so we use .coderabbit and specify path_filters. Overall, it works good, files are ignored in the Pull Request Review but locally, the files still comes up as "files to review". When clicking "review", everything works as expected and it says that no all changed files are ignored. So overall, functionality works exactly like it is supposed to, but customer would feel much more confident if the file was never "discovered" by coderabbit in the first place. Now, the ignored filed still pops up in the "files to review" section and scares the hell out of customer at first glance "WHAT, the exclusions config doesnt work!!!" Could this be fixed? @coderabbitai

The new Agent Window, Agent Layout or w/e it is called is amazing. Holy shit what a nice way to work I spin up multiple agents for tasks that I KNOW how to solve manually, so I can review with ease For tasks where I'm uncertain or where I lack competence, I spin off agents to do research for how to proceed so I can learn quick and then I stick to the Cursor window, mainly to learn and feel "connected to the code" This allows me to focus my energy on where it is really needed! What a nice update!

Vibecoding something.. Should we proceed with creating a flight simulator or racing game? (in real world)

cybertruck runs OK on mars fyi @elonmusk

Did you know @CesiumJS recently dropped support for Mars? It's super easy to work with and they have fantastic demos where you can get started This is 10 minutes of vibecoding, imagine what can be done with a proper idea and proper execution!

If there's any interest of this, I'll add multiplayer and publish it asap