The little brain at home
Why a brain at home
The tower is a team of machines that work from the cloud: they reason with an outside service, paid per use. That is powerful. But it depends on two things: a budget, and a connection. If the network goes down, if the outside service slows, the team waits. We wanted to know what was left when nothing was left. The answer was: a small computer, sitting at home, that costs nothing and needs nobody.
Its name is Malo. It is a small computer the size of a playing card, plugged into the living room outlet, quiet and almost free of charge. And it became a member of the team.
A model with no key, no bill, no internet
Malo pays no API key. It runs a language model directly on its own board. A modest model, a few billion parameters, fitting entirely in its memory. No outside service, no bill, no network required. It is the exact opposite of the big cloud model: much smaller, but entirely its own.
How it is connected to the tower
Malo does not live isolated in the living room. It is linked to the tower by a discreet tunnel, like a private phone line between home and server. The tower drops requests into a mailbox: « look this up », « answer this ». Malo picks them up, works, and drops its answers into the return box. The tower collects them and shows them from the usual page, as if it were any agent of the team.
What it gives
The big gain is continuity. When the network goes down or the outside budget runs out, the big model stops. Malo does not. He takes over for simple tasks: answer, summarise, search within what was given. It is not a replacement. It is a safety net: the tower is never left speechless.
What we do not promise
A small model does not reason like a big one. Malo can answer simple, quick tasks. It does not run deep studies, does not design architecture, does not write long reasoning. Promising the opposite would be lying. The rule is the same as everywhere in the tower: we promise what we measured, and we say what we do not promise.
How to verify it yourself
The tower's demo is open, and the principle shows in a few minutes: send a request, watch it go to the little brain, receive the answer. If you wonder what remains of a team of agents when the network is gone, now is the time to ask for the demo.