S2S · Silicon-to-Signature · patent pending · compute-side root

NVIDIA attests the GPU once.
Then nothing.

Confidential-computing attestation proves the GPU is genuine — at startup, one time. Every inference after is cryptographically unbound from it. Swap the model, reroute the device, serve INT4 as full precision, alter the output — the hardware attestation still passes. S2S is the leash: every inference hash-chained into an attestation epoch anchored to the hardware evidence and a Hive-issued nonce, signed ML-DSA-65, verifiable offline by anyone.

HONEST STATUS · READ FIRST

The crypto core is built and proven — 11/11 tests pass (seven failure modes + invariants). The verifier makes zero network calls. Real silicon is pending a CC-mode GPU. Every piece of origin evidence in the reference build is flagged simulated=True — nothing on this page masquerades as a live hardware receipt, and no live-silicon benchmark appears because none exists yet. Honesty is the product.

11 / 11tests pass · 7 attacks + invariants
≈9 µshook emit p99 · off the hot path
~2000×vs ≈18ms synchronous signing
0network calls in the verifier · sockets hard-blocked in test
The gap · flip the switch

One attestation at boot. Then hours of unbound inference.

The blue flash on the left is nvtrust at startup — genuine GPU, 1–3 seconds, once. Everything streaming after it is unbound: any of those dots could be a swapped model, a rerouted device, an altered output, and the attestation stays green. Flip S2S on and watch every inference snap a chain link back to the evidence.

time since last hardware attestation: 0h 00m 00s · and counting S2S · OFF · inferences unbound
The seven attacks · run each one

Hardware attestation passes. S2S fails it. Every time.

These seven are the acceptance criteria and the demo — all seven pass in the crypto core today, against evidence flagged simulated=True. Pick an attack. The left panel is NVIDIA telling the truth. The right panel is S2S catching the lie anyway.

NVIDIA hardware attestation

VALID

Hive S2S chain of custody

BROKEN
The epoch · Kᵢ = commit(H ‖ S ‖ M ‖ Cᵢ ‖ Kᵢ₋₁) · tap any link

Recursive continuity. Break one link, lose everything after it.

Each inference commits to the one before it — that recursion is the gap closure. Tap any link in the chain to tamper with it and watch every downstream link die. Then heal it and watch the epoch reconcile.

EPOCH INTACT · 24 inferences · hash-chained · anchored to H ‖ S
The five bindings · click each commitment

What gets committed — and why each one is load-bearing

The hot path · untouched

Receipting at ≈9µs. Serving never waits.

≈18 ms
synchronous signing on the hot path — the naive design S2S refuses
≈9 µs
S2S hook emit() p99 · async · fail-open on serving, fail-closed on the receipt

Bring the silicon. We bring the leash.

We're looking for CC-mode design partners running vLLM. You bring the GPU in CC mode; we wire the nvtrust flow, inject Hive's nonce, and run the seven attacks with simulated=False. You walk out with the first inference receipts a regulator, auditor, or court can verify offline — without trusting NVIDIA, the provider, the model owner, or Hive.

H100 SXM5 · PCIe H200 SXM5 B200 SXM6 GB200 CPU TEE · SEV-SNP / TDX
S2S · Silicon-to-Signature · patent pending · crypto core proven 11/11 · real silicon pending a CC-mode GPU · all reference-build evidence flagged simulated=True · we never fabricate a receipt · S2S proves occurrence and custody, not correctness (that's zkML, out of scope) · VM-level attestation is not GPU attestation · B300 HGX fabric-manager binding gap noted, not hidden · CC-mode overhead ~2–5%, attestation 1–3s once at startup · ML-DSA-65 · NIST FIPS 204 · Hive Civilization · Wyoming, USA