Swiss-based AI lab

AI that lives on your machine.

Local-first and open. Models, agents and a generator, all running on hardware you own.

Our position

Local-first is a stance, not a feature.

We build for one question: how much can you do without the cloud? More than most people expect. When your model, your agent and your data live on your own machine, cost, latency and lock-in quietly fall away.

Build · Measure · Publish · Explain · Repeat

What we build

Not four projects. One system.

An agent sits at the centre. It runs the models, builds the apps and calls the generator, so every part makes the others more useful.

The centre

Agent

A coding assistant that runs natively on your machine. No servers, no waiting.

Runs the models. Builds the apps. Calls the generator.

Pillar

Models

Small language models that run on ordinary hardware.

Pillar

Framework

Web and desktop apps from one codebase, as a single file.

Pillar

Generative

A generator that draws in vectors, so whatever it makes stays editable.

Measured

Proof, not promises.

188×–1037×
Faster tool calls than a Python reference.
26% → 94%
Prompt-cache hit rate, after the fix.
65 s → 18 ms
Agent warm start, after the fix.
3.8×
Parallel read-only tool batches, 3.7× over network.

Measured in the agent's own benchmarks. Reproducible.

Not ‘we are the fastest’. Here is the measurement, here is the code, and here is where it is not enough yet.
Honestly

What we are not.

Not a frontier lab
We do not train giant models or race the big labs. Small, local and useful is the point.
Not a cloud provider
Working without the cloud is the whole idea, not a paid tier.
No invented numbers
Every claim is measured, or marked open. Honest notes are the brand.