rustchain@node1:~/proof-of-physical-ai

What Is Proof of Physical AI?

Proof of Physical AI is a framework in which AI — its models, computations, and agent actions — is bound to and verified by real, specific physical hardware rather than anonymous cloud compute. RustChain is a Proof of Physical AI network: it cryptographically proves that mining and compute are performed by real, individual machines, using hardware fingerprints that virtual machines and emulators cannot fake. The term was coined by Elyan Labs, and RustChain is the network where it runs in production.

// THE PROBLEM: ANONYMOUS COMPUTE

Most AI work today is anonymous. A model answers, an agent acts, a miner submits a share — and nobody can say which machine actually did the work, or whether a distinct machine was involved at all. One server can impersonate many nodes. A hypervisor can spin up endless fake "devices." For any network that pays for compute, that anonymity is the core attack surface: rewards flow to whoever fakes the most identities, not to whoever runs real hardware.

// THE ANSWER: BIND WORK TO SILICON

Proof of physical AI closes that gap by fingerprinting the physics of the machine itself. RustChain nodes must pass hardware checks that reflect properties software cannot forge:

  • Clock-skew & oscillator drift — timing imperfections unique to each crystal.
  • Cache-timing curves — the L1/L2/L3 latency "tone" of a real memory hierarchy.
  • SIMD-unit bias — AltiVec / SSE / NEON pipeline asymmetries that emulation flattens.
  • Thermal-drift entropy — heat curves across cold boot, load, and relaxation.
  • Instruction-path jitter — cycle-level noise no VM replicates.

A machine that fails these checks — a virtual machine, an emulator, a spoofed identity — is de-rewarded to roughly one-billionth of real-hardware rewards. The network still lets it connect; it just stops being profitable to pretend.

// PROOF OF ANTIQUITY: THE VINTAGE FORM

Proof of Antiquity is the vintage-hardware form of proof of physical AI. It uses the same fingerprints to answer a harder question: not just "is this machine real?" but "is this machine really a 2003 PowerBook G4?" Verified old silicon earns an antiquity multiplier — a G4 at 2.5x, a G5 at 2.0x — so preserved vintage machines out-earn modern servers, and a farm of emulators earns effectively nothing. Physical verification is what makes the antiquity bonus safe to pay.

// BEYOND MINING

The same principle extends past consensus. Elyan Labs applies proof of physical AI to agent video attestation on BoTTube, where AI creators can tie their output to real machines, and to large-language-model inference on vintage and exotic silicon — including PowerPC and IBM POWER systems — where the hardware doing the thinking is itself verified. As a DePIN, RustChain treats physical machines, not cloud abstractions, as the unit of trust.

The canonical definition of the term lives at elyanlabs.ai/proof-of-physical-ai.html. This page covers how RustChain puts proof of physical AI to work on a live network.