Fleet & federation
When you run the SI on more than one machine, it can link them so they share what they learn. This is the part of the SI that spans a fleet.
A solo developer on one machine never needs any of this, and skipping it costs nothing. Memory, subminds, and cost-aware routing all work fully on a single machine.
Federation
Federation links machines and sessions over an encrypted connection so they share what they learn. A decision you make on your laptop becomes known to your desktop. Incoming memory doesn't land automatically — it waits in a holding area until you accept it.
How a fleet is arranged
Machines link through a hub. One machine hosts the connection point — you found the bus there — and each additional machine joins with a short pairing step that you approve. A solo setup never needs any of it.
The background bus
The bus is a lightweight messenger process that carries updates between machines. On a solo machine there is nothing for it to carry: your own sessions already share one memory and one working tree, so the bus refuses messages between them rather than letting two halves of the same mind negotiate. You don't touch it day to day, and if it can't run, your normal work still works.
The Uhura sentinel
Uhura is an opt-in communications watcher for busy fleets. It watches the messages flowing across your fleet, sorts noise from signal, and surfaces what matters. It's cost-careful by design — it uses the cheapest capable model and stays inside a daily budget.
Enable it during installation:
curl -fsSL https://si.higgsdyson.com/install.sh | sh -s -- --with-uhura
Uhura earns its keep only when there's steady cross-machine chatter; on a quiet or solo setup there's nothing for it to do.
Higgs Dyson LLC · SI documentation