WHAT I'M WORKING ON
Finding new ways to integrate RocketSim into AI
This week's article shines a light on this more, but I want to share a bit more insight into the journey towards the first preview of RocketSim 17.
I’ve been exploring integration with Claude Code, Cursor, and Codex for a while. There are options to do it natively, but none of them were built once, run everywhere. Adding another release pipeline is far from ideal, and I already have the CLI and Agent Skill.
As I’m sure many of you do, I often build web pages using AI. What’s great about it is that I can run and validate the website directly inside the browser in Cursor. It made me wonder: why can’t we do the same for app development?
I started streaming the Simulator, which I had implemented fairly quickly: RocketSim’s CLI allows agents to stream and create screenshots or videos.
At first, I figured I’d rebuild RocketSim entirely, so I started creating the side window in the browser. Quickly, though, I realized I was optimizing for a completely different flow: agentic development. You’re not looking for the same tools, because you could do that using RocketSim natively. Instead, it should be a way to follow along with what the agent is doing and to provide the agent with feedback to continue its journey.
That’s how I started to evolve into a Simulator stream, combining worktree and branch information with accessibility tree insights and design feedback. Agents will write the code; you will provide feedback on the app’s behavior.
It’s still early days, and I’m actively using this version myself, but I’m keen to see where this goes. I’d love to get your ideas on the future of app development, so don’t hesitate to hit reply and share your dreams with me.