A collective brain. Decentralised.

Organisations and individuals pooling their computing power. Everyone keeps their data at home.

Join the projectUnderstand the idea

Computing power is concentrating. Your machines sleep at night.

Large models run in a handful of data centres, far from you. Meanwhile, the office computer, the firm’s server and the home machine wait for morning.

At night, almost everything sleeps.Together, they think.

Your knowledge stays with you. Only computing power travels.

  1. The office closes.The accounting firm’s Mac mini has nothing left to do. Like thousands of other machines, it goes to sleep.
  2. It offers itself to the network.It only lends its power. The firm’s documents, memory and decisions stay locked, on site.
  3. A laboratory sends a task.Summarise thousands of papers. The task is split into pieces, and each free machine takes one.
  4. By morning, everything is back.Each piece returns with its trace. The firm has computing in reserve, for its own assistants.

Images: Apple (Mac mini, Mac Studio), NVIDIA (DGX Spark). An example to explain the idea; Meotis is not affiliated with their manufacturers.

Open models. On everyone’s machines.

Models whose weights are public, which any machine can run without sending anything to a provider.

  • Qwen
  • Llama
  • Mistral
  • Gemma
  • DeepSeek

What we want to guarantee. Before anything else.

  1. 01Your data does not travel.Only a computing task leaves. Your documents, your memory and your decisions stay on your machine.
  2. 02Every computation leaves a verifiable trace.Who computed what, when, on what budget: every contribution can be checked afterwards.
  3. 03Everyone receives according to what they bring.Power lent at night comes back by day, when you need it. Sharing is measured.

Open questions. We tackle them one by one.

  • Computing without seeingHow can a machine compute for you without reading your data?
    Under study
  • Proving the workHow do you check that a computation was done, and done properly?
    Under study
  • Sharing fairlyHow do you measure what each one brings, and what each one receives?
    Under study
  • Using less energyPutting existing machines to work, rather than building new ones.
    Under study

Knowing what a computation costs. Before sharing it.

To share fairly, you first have to measure. We studied what a small model can do under a compute budget set in advance: the paper, code and measurements are open, and already picked up by an open project.

Published 9 June 2026 · ZenodoDoing more with a small model, under a strict compute budgetAerni, Fluck & BeckerRead the paper

Learn more: sharing compute, by Meotis Research

The collective brain is a research programme by Meotis Research, in Neuchâtel. The idea: organisations and individuals pool the computing power of their machines, like a Mac mini asleep at night, without giving up their data.

Only a computing task travels; documents, memory and decisions stay with each participant. Every computation leaves a verifiable trace, and everyone receives according to what they bring. The models used are open models, whose weights are public.

Nothing is open to the public yet: we are first laying the scientific foundations, publishing our measurements and having them reproduced.

Let’s build it together. Everyone in their place.

Researchers and developersJoin the researchProtocols, measurements, papers: the programme is open, and every result can be repeated.Join the research
Organisations and individualsLend your machineBe notified when a first trial opens. Just your email.

For an organisation · Let’s talk about what you could bring, and receive.

A 30-minute callkMeet video call · no commitment

Just your email. The video call link arrives straight away.