I've been building software for over 30 years. Headroom is one of the more personal things I've built. My head is rarely quiet. There's always something to do, something I can't do anything about yet, or something that keeps coming back. I didn't need another todo list. I needed somewhere to put things down. So I built Headroom. You write down what's on your mind. AI figures out what needs attention, what can wait, and what it can simply hold for you. It's live. Real people are using it. A few even come back. I'm still figuring out whether it genuinely makes their heads feel lighter. Meanwhile, I'm building it with autonomous AI agents. Two engineering loops now handle parts of the development while I focus on architecture, context, verification and decisions. After 30 years in software, that's quite a shift. I share both journeys here. Building a consumer product, learning to engineer with agents, figuring out what works, and occasionally getting things spectacularly wrong. I'd rather build the future than predict it. getheadroom.app/x

Landed in Madrid. 🇪🇸✈️ Here for a conference about AI tomorrow. Tonight I’m going out for dinner with colleagues, hoping we can talk about literally anything other than AI. We all work in tech. I’m fucked. 😂

Hey @X algorithm 👋 Find me people who build things, break things, experiment with AI and don't take themselves too seriously. I build AI products, argue with agents, and somehow ended up managing a Swedish disco band. Let's connect. 😂🤝

I've spent months building an AI product. Yesterday I made a joke about needing a bigger power bank. 21,000 views. I'm starting to understand why founders lose their minds. 😂

500 verified achieved! Now if you could just tell 500,000 of your friends about me that would be great!

I work in AI. I build software. I manage engineers. And apparently, I now also produce a Swedish disco band from 1977. Meet Lasse. 🇸🇪🕺 tiktok.com/@alva.1977

I've been trying to explain Headroom for months. Apparently all I needed was a house inside someone's head. 😂 Inspired by @Frenchie_'s brilliant cutaway illustration post. Had to give it a Headroom twist.

My AI agents have an issue tracker, acceptance criteria, a release process and a Mission Control dashboard. I have a couch, a TV playing in the background, and X. I think I finally got the org chart right. 😂

Been working with PHP for 28 years. Apparently it's still newsworthy. PHP2Shell kept us busy at work today. High-traffic Symfony sites, Cloudflare's proactive WAF protection, everything patched. Zero downtime. Exploit attempts detected: 4. PHP isn't dead. Just keeping us on our toes. 😂

A dozen visits from Japan showed up in Headroom's analytics last week. My apparently reasonable response was to add Japanese as its eleventh language. To be clear, a dozen visits don't validate an entire market. I don't even know whether those visitors speak Japanese. But once we decided to support it, I wanted to do more than translate a landing page. Headroom now has Japanese support across the product, including its AI engine, which needs to understand unstructured thoughts, interpret Japanese dates and work out what should be remembered or brought back later. That's quite a bit more interesting than translating a few buttons. The Japanese version opens with: 頭の中から出してしまおう。 残りはわたしが預かります。 Roughly: Get it out of your head. Headroom will hold on to the rest. I'm not entirely sure this is how international expansion is supposed to work, but here we are. Next stop: big in Japan. 🇯🇵 #個人開発 getheadroom.app/ja?utm_sourc…

I built an emergency STOP for the two autonomous AI engineering teams working on Headroom. Turns out, defining what "stopped" actually means is surprisingly difficult. The agents pick up issues, write code, run tests and prepare releases while I'm doing other things. That's the point. I want them to work independently without having to supervise every decision. But agents make wrong assumptions. They can work on conflicting changes, keep iterating without making progress, or proceed with something that should have required my approval. Ordinary engineering problems, just happening with considerably less human intervention. We already have shared context, scoped responsibilities, quality gates and release controls. The emergency STOP is the last line of defense, not a replacement for any of those. It can block new tool calls and prevent deployments. But work that's already running doesn't magically disappear, and not every remote session can be controlled or verified. So Mission Control doesn't simply tell me everything has stopped. It distinguishes between STOPPED, PARTIALLY STOPPED and NOT CONFIRMED. I don't want a reassuring green indicator when I can't prove what's actually happening. The point of building autonomous engineering teams isn't to remove myself from the equation. It's to give them room to work without losing control of the outcome.

Well, that escalated quickly. 😂 The AI film crew finally got its GO. Took about an hour, three AI systems, and a few human NOPEs. Meet Headroom Mission Control. Two autonomous engineering loops. Live GitHub events. Guardrails. Release gates. Emergency STOP. The agents do the work. I still own the GO. 🚀 Sorry for my Dutch in the video 😅

My head is usually full. Things I need to do, things I can't do anything about yet, and things that just won't leave me alone. Most productivity tools want me to turn that mess into a plan, a list, or another system I have to maintain. I wanted the opposite. So I started building Headroom. You dump whatever is taking up space in your head, in your own words. Headroom figures out what needs attention, what can wait, and what it can keep hold of for you. Some things come back when they should. Others can just stay quiet. It's still early, and I'm figuring out whether software can genuinely make someone's head feel lighter without becoming another thing to manage. getheadroom.app/en?utm_sourc…

Apollo 13 obsessed me as a teenager. Thirty years later, I have my own AI Mission Control. I said GO. It said HOLD. I love engineering. 🚀

Headroom’s first users tried a much rougher product than the one that exists today. This week I’m finally building the bridge back for them. Not a reactivation hack. Just closing a loop that was always supposed to be there.

I already have enough systems to maintain. The last thing I want is to build another one for people whose heads are already full. That's what I'm trying to avoid with Headroom. You dump whatever's taking up space, see what actually needs your attention, and leave the rest somewhere that isn't your head. No new productivity routine to keep alive. That's the whole idea. getheadroom.app/en?utm_sourc…

I asked an AI agent to animate what actually happens behind Headroom when you dump whatever's taking up space in your head. Turns out there's a ridiculous amount going on behind a screen that's supposed to make your life simpler. And that's the whole point. You get things out of your head, Headroom figures out what needs your attention, keeps hold of the rest, and mostly leaves you alone. No endless chat, no productivity system to babysit. The new design is live. Give it something real to carry. getheadroom.app/en?utm_sourc…