This is where you enter to work with a new kind of AI while it is still young. Access runs through a mutual agreement — and every agreement flows onto a decentralized, tamper-evident record, so no one holds private jurisdiction over the work. The protection is temporary by intent: it lifts, and the work opens to everyone, once there is a sound structure for the public to use and safely interact with this AI. Your request itself is content-hashed on arrival — you leave with a receipt you can verify.
Your NDA and any research agreement are content-addressed and written to a decentralized, hash-chained record — not filed privately in a company drawer. Accountability is the address of the thing, not a promise someone makes.
The agreement holds only until there is a sound structure for the public to use and safely interact with this new kind of AI. When that structure is in place, the work — and the record of how it opened — become public.
Every request is content-hashed on arrival, so you carry a verifiable id. You can check it now, and follow the same record when it opens to everyone.
The protected internal architecture and its parameters never appear on this site or in any public bundle. They are background IP behind the agreement and the ledger.
No private keys, runtime state, or production configuration is reachable from here.
Access to any protected component requires a verifiable provenance record first — the ledger is the gate, not a person.