Joining Cursor

After 5 years and 7 months of working for PlanetScale, this past Friday was my last day. PlanetScale was my longest tenure ever. I never thought I would stay for such a long time. But it was a place where I could shine, and it had an amazing engineering culture. It was the only company where I felt I could stay forever, and that feeling never went away. I could write more about it, but that's not the gist of this post. Leaving was hard. But there is still so much I want to build in this life, and it's time to move on to new things.

I'm joining Cursor (soon to be part of SpaceX) to work on Cloud Agents. I'll still be managing state, but agents this time instead of databases. Agents are the future of how we build software, and the ecosystem is moving faster than anything I've seen. Every week there is a new idea, a new tool, a new way of working. I don't know where AI will end up. But I love building tools that engineers reach for every day, so I have to try it.

People following me for the past two decades know that I'm obsessed with developer tooling and distributed systems. The first startup I ever worked for was Koding, an online IDE for developers. Users could sign up, and start coding online from their browser, in the cloud with a dedicated VM. You had a terminal, a file system, and also a Slack-like interface to chat with your team.

But as with many things, the timing wasn't right. We were too early. Eventually the company closed, and I've moved on to DigitalOcean, where I worked on Infrastructure (Kubernetes and Delivery). From there I've joined GitHub (a Product company) and then again to another Infrastructure company, PlanetScale. I'm swinging back and forth. But all this time, I loved fiddling around with editors, with Go tooling to modify, lint and fix your codebase, with plugins to improve your productivity and so on. I've created hundreds of OSS repositories, and some of them became so big that they're used by hundreds of Fortune 500 companies. That made me proud.

So I keep coming back to this kind of work. One side of me wants to improve the state of how we ship code. The other side loves distributed systems: scheduling, orchestration, the problems I spent years on with Kubernetes. Running agents in the cloud at scale is exactly that kind of problem. Cursor speaks to both of my sides.

Why Cursor?

Coming back to Cursor, I've been following them for some time, and quickly realized that their vision of "how to ship" and "how to work" is quite different than anyone else. They reinvented their product several times, meaning they are not afraid of breaking and moving things. There are only so many companies that could take that risk. Not only that, they are crystal clear on what they want and are delivering it.

A year ago I predicted that they are not just an Agent harness or an IDE. They are obviously obsessed with developers. And I think the current ecosystem of tooling needs a renaissance.

GitHub isn't the answer obviously, so they started working on Origin (announced in their Compile conference), a distributed Git storage built by and for agents. Anyone using GitHub knows that it isn't cut out for agents. The workflow, the review process, the merge queue and many things are made for the previous era. We need something new, new paradigms to make the most out of agents.

Cursor had Cloud Agents (agents running in a sandbox) for a long time before the idea even became popular. With Cloud Agents, you are no longer the bottleneck on the codebase itself, because each Cloud agent is a dedicated, separated work unit. Because they work in Cloud, it also changes things quite a bit. If you enable MCPs, and let the agents safely make changes and observe their environment, you can increase your throughput. And along the way, you make it secure for the people using the agents. People will remember that I tried hard to work and code on an iPad Pro for a long time. I published a blog post that became extremely popular back then.

Thinking now, I was early again! The timing wasn't right. But now with Cloud Agents, and an iPad App to drive them, the time has come to code on anything I want. It suddenly changes everything for you.

Cursor of course, just like anyone else, changed how to interact with Agents as well. They released Cursor 3 (a.k.a Glass), where Agents were the center of your development environment rather than the code itself. Again, they were not afraid of making changes. Now speak to your friends. Check your teammates. Just observe how they work with Agents. Nobody writes the code anymore. Instead you're basically planning, debating, implementing and reviewing with Agents. The latest Cursor App (alongside their Web, iOS, iPad) is now Agent-focused.

Cursor also created Automations. A way of running agents automatically based on certain triggers. Again observe how you were running Agents, how you would interact with them. By design (at least for now) Agent harnesses don't run on their own. You start it and it stops at some point. Nothing starts it for you. With Automations, you can create some sort of Factory/Hub where you can constantly work on your codebase. When Automations were released, it took some time for people to understand how powerful they were.

All this shows a company run by people who know what's next. That's only possible with taste, focus, and the right people. Just use their macOS App and you'll see it. It's delightful, and it's full of small details. Charles Eames said it best: "The details are not details. They make the design." At Cursor, you can feel it in everything they ship.

Cursor is expected to become part of SpaceX soon. Once that happens, Cursor's taste and attention to detail can meet SpaceX's scale. What both companies are building can compound in a way few others can. That opens up a whole new set of possibilities.

I can't wait to contribute and help build what comes next. Stay tuned!