What Apple showed
According to heise online, on October 14, 2026, an Apple AI research group published a paper about the Minions software:
- A central server sends AI tasks to Macs with little or no use.
- Each Mac runs Ollama and returns the result.
- At Apple, waits of more than ten hours for a GPU dropped to minutes.
What is still missing
- It is a research prototype, not a product.
- It only serves organizations that trust each other: isolation between users does not exist yet.
- heise could not find the project's code.
Who it is for
- Hospitals, universities and public bodies with many Macs idle at night.
How it works
You describe the case
How many Macs, which models, and when they are idle.
We study it with your samples
We measure whether your Macs run the model you need, with your own samples, before any contract.
We build the project
If the study shows that it is worth it, we build the project and connect the result to the same API.
Good to know
- It is a project proposal: Apple's software is still a prototype and does not isolate one user from another.
- On demand: the study starts when an organization asks for it.