Questions & answers

Nothing. It only uses your machine's spare capacity, and you can leave anytime. Your keys are generated on your own machine and never leave it.

Any decent machine will do — even a laptop. No installation or command line needed: joining is one click.

It's the exact opposite. The node runs network-isolated, in a zero-egress environment: it cannot send arbitrary traffic anywhere, because it has no way to. What it does is deterministic computation whose output is a re-checkable proof — you run it yourself and get the same bytes; if they don't match, it's dropped. A botnet is opaque traffic; this is the reverse: every step is verifiable.

Your keys are generated on your machine and never leave it. The link between machines is deliberately narrow: state (shape, metadata) can cross, content (a fact, a claim, code) cannot.

It means one thing and nothing more: reproducible byte for byte, by anyone. If a claim doesn't reproduce independently — on several machines, separately from one another — it doesn't count as knowledge. This is how it works, not a slogan.

The core intuition is right — truth comes from independent agreement — but it's sharper on three points. (1) We don't compare an LLM answer: an LLM run twice gives different output, so the LLM only proposes, never decides; what the machines compare is a deterministic derivation that must come out the same byte for byte. (2) They don't re-run the same thing — they derive it independently, with separately written implementations, from the same public source; separate authorship is the guarantee. (3) If they don't match, it does NOT give an answer: it stops and flags the discrepancy. The cross-check is the core — but of deterministic, byte-exact agreement, not a model's guess.

Honestly: it's an early phase. Right now you get a founder's place and lifetime access for what you help build. There is a host-earning model built in too, but we haven't switched on the exact compensation yet — we won't promise something we haven't decided. It costs nothing, and you can leave anytime.

In the end everyone gets their own, isolated AI black box — not a shared cloud model, but yours, on your data, on your machine. On the customer side it's the same Zafire black box: proven, reproducible AI on your own hardware, with no cloud model. It will be available at three levels — Basic, Pro, and Enterprise — from personal use to organizational scale. The exact packages and pricing are still taking shape, but the core is the same in all three: deterministic, proof-bound AI, on your side.