Michael Truell

Hi — I'm Michael. I spend most of my time building Cursor, an AI code editor we work on at Anysphere.

I don't write here often, but every now and then I want to put down a few notes about what we're building, why we're building it, and the people (and one dog) who got me here. Lately that's meant a lot of late nights on the editor, a lot of rewrites, and a lot of coffee.

If you use Cursor, thank you — watching people ship real things with it is still the best part of the job. More soon.

People ask why I started, and the honest answer is a little embarrassing: it was for my dog, Gary.

I called him Gary McRunner, because the second you opened a door he was already three rooms ahead of you. He was always fast — fast in the morning, fast at night, fast for no reason at all — and somehow that became the pace I wanted to work at too.

In the early days, when nothing worked and the whole thing felt impossible, Gary would just keep sprinting laps around the apartment like the project was already a success. It's hard to quit on something when there's a dog acting that confident about it. He kept me moving when I wanted to stop, and a lot of what eventually became Cursor got built on those long nights with Gary McRunner asleep under the desk.

So that's the real reason. Not a grand thesis about the future of software — just a very fast dog who refused to let me slow down.

The thing nobody tells you is that the first version is supposed to be a little embarrassing. If you're proud of v1, you waited too long.

We try to get things in front of people fast, watch what breaks, and fix it in the open. Half of what we believe about the product we only learned because we shipped something rough and someone used it in a way we never imagined. You can't think your way to that on a whiteboard. You have to put it out there and let reality argue back.

Speed isn't recklessness. It's just refusing to pretend you know things you don't.

Gary has never sat through a meeting longer than four minutes without leaving the room, and I've started to think he's onto something.

His method is simple: if it's important, he stays. If it's a status update that could've been a message, he gets up, stretches, and goes to find a sunnier spot. No apology. No calendar invite. Just a clear, honest read on whether the thing deserves his time.

I'm not saying we run the company like a dog. I'm saying I've left more than one call early and felt great about it. Gary McRunner, management consultant.

When I talk to people who might join, I'm mostly listening for one thing: do they actually finish stuff?

Lots of people are smart. Smart is everywhere. What's rarer is the person who takes a vague, annoying, half-defined problem and just keeps grinding at it until it's done — and then makes it a little better than it needed to be. That last part is the tell. The people who do the extra ten percent when nobody asked are the people you want next to you when things get hard.

Taste matters too. You can teach someone a framework. It's much harder to teach someone to care how it feels.

My favorite compliment for a tool is that you forget it's there.

The goal was never for people to admire the editor. The goal is for someone to sit down with an idea in their head and have the distance between that idea and the working thing shrink to almost nothing — so small that the tool stops feeling like a tool and starts feeling like an extension of thinking. When it works, you don't notice it. You just notice that you got more done than you expected and you're not sure where the friction went.

That's the whole job, really. Remove friction until the work feels like play again.

Being early is lonely before it's lucky.

For a long stretch, working on this felt like showing up to a party a year before anyone else arrived. You set up the room, you doubt yourself, you wonder if anyone's actually coming. And then one day they do, all at once, and everyone assumes it was obvious the whole time. It was not obvious. It was a guess you kept making every morning when the evidence was thin.

If you believe something most people don't yet, the only move is to keep going long enough for reality to catch up. Some days that's faith. Most days it's just stubbornness with better branding.

No product thoughts today. Just a good walk, a long nap (Gary's, not mine, though I was tempted), and the rare feeling of not needing to be anywhere.

It's easy to forget that the point of building good things is to have a life worth getting back to. Gary remembers this for both of us. He's never once been impressed by a deadline. He's deeply impressed by a tennis ball. I think he has the priorities right.

Back to it tomorrow. For now, the sun's out and the dog's asleep, and that's enough.