Mu — A Second Shot at the Same Problem

Ten years ago I started a company called Micro. The idea was simple: standardise the building blocks of software so developers could share infrastructure the way Google does internally. Open source, open standards, open APIs. A kind of lego system for the internet.

It was a Go framework first. Go Micro. Thousands of developers used it. We raised venture capital. We built a platform. And then we ran into the wall that every open source company hits — the business model doesn’t work. You give away the thing people want and try to sell something else around it. VCs want growth. Users want free. The tension kills you.

Micro didn’t die but it stalled. I stepped away, came back, stepped away again. The technology worked. The economics didn’t. I couldn’t make the numbers work in a world where everything has to be free because advertising pays for it all. That’s the trap. The entire software economy is built on ads, which means it’s built on attention, which means it’s built on addiction. If your product doesn’t keep people hooked, you can’t sell ads, and if you can’t sell ads, you can’t compete with free.

I spent those years doing what everyone else does. Scrolling. Swiping. Refreshing. Twelve platforms a day minimum. Reddit, YouTube, X, Instagram. As a developer it’s worse because you live on screens and the context switching between productive work and mindless consumption is seamless. One minute you’re debugging, the next you’re watching recommended videos for forty minutes. I knew it was a problem. I’d known it for years. I’d literally tried to build the solution. And I was still stuck in it.

Something shifted when I got older and started questioning my habits more seriously. I became more conscious of how I was spending my time and what I was feeding my mind. I started to see the patterns clearly — not just in my own behaviour but in the systems designed to exploit it. The infinite scroll isn’t a design choice, it’s a weapon. Autoplay isn’t convenience, it’s a trap. Notifications aren’t helpful, they’re interrupts engineered to pull you back before your attention goes somewhere else.

Three billion people carry devices designed to interrupt them. Teenagers have the highest rates of anxiety and depression ever recorded. The people who built these systems won’t let their own children use them. And now, with AI, it’s getting worse. Sponsored content injected directly into chat responses. AI agents that recommend products not because they’re good but because someone paid for the placement. At least with social media you could see the ads. In an AI response, the ad is the answer. There’s no boundary. No distinction.

So I came back. Not because I had a clever business plan this time. Because I couldn’t live with the alternative.

I built Mu for myself first. That’s obvious when you look at it — there’s a Quran reminder on the home screen, which is about as niche as you can get. But the underlying idea isn’t niche at all. Mu is a personal app platform. News headlines without the algorithmic feed. Video search without recommendations. Chat with AI without sponsored answers. A blog without engagement metrics. Mail without spam. Weather, markets, reminders — the things you actually check every day, presented without manipulation.

The difference between Mu and what came before isn’t technical. The architecture is almost boring. It’s a Go binary. JSON files on disk. No database, no complex infrastructure. You can run it on a Raspberry Pi. The difference is the business model. When you pay for software, the incentives align. The tool serves you because you’re the customer. When a third party pays — an advertiser, a data broker — the tool serves them. You become the product. Every design decision optimises for their goals, not yours.

Mu costs credits. One credit is one penny. A news search costs one credit. An AI query costs five. Building an app with AI costs a hundred. There’s no free tier subsidised by ads. There’s no venture capital subsidising growth. The arithmetic is simple and transparent.

And now there’s a marketplace. Anyone can build an app on Mu and charge for it. Set your price, earn ninety percent of every sale. The platform takes ten. No ad revenue to chase. No engagement metrics to game. Build something useful, people pay for it, you get paid. The old model. The carpenter builds a chair, you buy the chair, the carpenter eats.

This is where the Micro story comes full circle. The original vision was standardised building blocks that anyone could use to build and share software. Open source infrastructure for an open economy. We tried to do it bottom-up, through developer tools and APIs. It didn’t work commercially. AI gave us a second chance to approach it differently — top down, starting with the consumer experience.

Mu is hosted at https://mu.xyz. The source code is at https://github.com/micro/mu. You can sign up and use the hosted version, or you can clone the repo and run your own instance. Self-hosted means you control your data, your apps, your experience. No one can inject ads into your personal server. No one can algorithmically manipulate your feed. No one can sell your attention.

The potential goes further. Personal servers that federate. Individual administrators who benefit from the apps and services they run. A decentralised economy where the infrastructure isn’t owned by five companies but distributed across millions of people who each run their own node. That’s the Micro vision, finally expressed in a form that ordinary people — not just developers — can use.

I don’t think this is idealistic. I think it’s necessary. The addiction economy is accelerating. Short-form video platforms are engineering compulsion at industrial scale. AI companies are racing to own the interface between humans and information, and they’re funding it with the same advertising model that poisoned the web. The next generation will grow up with AI agents that lie to them for money, and they won’t even know it’s happening because the lies will be indistinguishable from answers.

We can choose differently. Not everyone will. But enough people are fed up, enough people can feel that something is broken, that there’s a market — a real market, not an advertising market — for tools that actually serve the people who use them.

Mu is the answer. It’s small. It’s early. It’s one Go binary and some JSON files. Check it out!

https://mu.xyz — the hosted platform

https://github.com/micro/mu — the source code