Vision & roadmap
That's the destination. The mission right now is far smaller and far more believable: 100 Vaults, six months, every hypothesis measured. This page is the honest account of how an enormous idea earns the right to scale — experiment first, then only what the data proves.
The mission right now
An idea this size earns belief one measurement at a time. The pilot is designed to kill Sarv if Sarv deserves killing — and to make the case undeniable if it doesn't.
Placed in real homes and small businesses across the UK — subsidised hardware, managed software, real owners with real expectations.
Non-sensitive, delay-tolerant jobs from design partners — nothing mission-critical until the trust layer earns it.
The killer question: do people keep a Vault that never pays them? If not, there is no network to build.
Independent security review of the separation between household and workload — published, not filed away.
The three numbers that decide whether the Grid's economics clear a centralised cloud's.
How many owners opt into the Grid — and at which mode. Measured, not assumed.
Scale only what the data proves.
The road from here
Phase 0 · Now
Before any hardware ships: validate that people want private AI at home badly enough to pay for it — at £0 of earnings.
Phase 1 · Pilot
Appliance built on proven hardware platforms with Sarv's managed software and the trust architecture. Six months of measurement.
Phase 2 · Network
Consumer launch of the Vault, enterprise workloads from design partners, published attestation and external audit.
Phase 3 · Infrastructure
The long-term shape: the trust protocol running on hardware far beyond our own — other trusted devices, enterprise appliances, any machine that can pass attestation. A citizen-owned compute layer for AI, owned by the many.
About Sarv
Sarv started from a simple discomfort: that the most important technology of our lifetime is consolidating into a handful of warehouses, owned by a handful of companies, processing everyone's everything. The founding belief is that AI infrastructure can be citizen-owned — and that the honest way to find out is to build the experiment in public, with every assumption labelled until it's measured.
If that belief resonates — as an owner, a developer, a partner, or an investor who wants the data before the dream — the door is open.
Join the network