For investors

We are building the room AI works in.

Most AI still arrives as a text box. You type, you wait, and whatever it did while you waited is gone. Pixel City is a studio that builds the other thing — one place where people and AI agents work side by side, and the work is visible while it happens. An agent here is a character you interact with, with tools of its own. You see what it's doing before it's done.

Creating next-generation AI-based user experiences.

Why now

The model is not the moat.

Every few months a better model lands and the one before it gets cheaper. That is good for anyone building on top and hard for anyone whose product was the model. What doesn't commoditise is the place the work happens.

Capability is becoming a utility

It arrives on a schedule now, from several vendors, and the price keeps falling. A product whose whole answer is a smarter box has a short shelf life, because the box underneath it is replaceable by design.

The experience layer is what lasts

People stay with a place they know how to walk into. The room around the model — what you can see, what you can hand off, what you trust it with — is ours to build, and it survives every model swap underneath it.

What we build

One product, shipping.

Luma World is a virtual office you open on your Mac. Your agents arrive as characters, take desks and get on with it — and you're in the room while they do, not reading the log afterwards. Public release is this month, September 2026.

Luma World

A virtual office for AI teams.

macOS · public release coming luma.pixelcity.world →
How we work

Small, fast, and inside the thing.

We work in Luma World every day. Agents took desks in this office before anyone outside got to try it, and they still have them. A rough edge doesn't wait for a bug report — whoever is in the room that day feels it.

We ship from inside

Whatever lands is used the same day by the people who built it. It is the shortest feedback loop we know how to build, and it costs nothing to run.

Small on purpose

A small team can hold the whole product in its head. We would rather move a week earlier than add another hand-off to move at all.

Agents do the repeat work

The work that would otherwise need more people is the work our own agents already do. It is the same bet we're asking anyone to make on the product.

What's next

The office is the surface.

Coding agents are where this starts, because that's the work we know best. The room doesn't care what the work is. Anything an agent can do at a desk is a character you can watch, in the same office, under the same rules.

More kinds of work

A desk, a screen and a name is a general shape. The first desks run code; research, support, review and operations are the same office with different people in it.

More people in the room

Once the work is visible, someone who would never open a terminal can still sit in and follow it. That is the part a text box has never given anyone.

Get in touch

If this is a conversation you want to have.

We'd rather show it than describe it. Write to us and we'll open the office and walk you through a working day in it.