Matrices —
Towards self-driving computers

We'll be honest: when we first heard about Matrices, we thought it was just another AI startup with a vague mission. But dig a little deeper, and you'll find something genuinely intriguing. This small team is building the infrastructure that could make "self-driving computers" a reality.

What Caught Our Attention

Matrices isn't trying to build another chatbot or image generator. Instead, they're focused on something much more foundational: training environments for AI agents that actually use computers like humans do. Think of it as a gymnasium for AI, where digital agents learn to click, scroll, and navigate browsers just like you would.

The company's tagline, "towards self-driving computers," might sound like sci-fi marketing speak, but it's actually a pretty accurate description of what they're after. They're building the training wheels for AI that needs to interact with real software interfaces.

Why This Matters Now

The timing here is spot-on. As AI labs race to build agents that can actually do useful work on computers, they need realistic environments to train these systems. You can't just throw an AI agent at a live website and hope for the best. You need controlled, repeatable scenarios where the agent can fail, learn, and improve without breaking anything important.

Founded in 2023 by John Qian and Leonardo Axel Setyanto, Matrices is riding the wave of interest in computer-use AI. With major players like Anthropic releasing computer-use capabilities, the demand for training infrastructure is only going to grow.

The Reality Check

Here's what we don't know: funding details, customer traction, or how their platform actually performs compared to alternatives. The company keeps a pretty low profile, which makes sense for a B2B infrastructure play targeting AI labs. But it also means we're working with limited public information about their progress.

Still, the core concept is solid. If you're betting on a future where AI agents handle routine computer tasks, someone needs to build the training infrastructure. Matrices seems positioned to be that someone.