The Markov Channel and Conditional Diffusion Verifier - first public release

Formalising and verifying the hash-chained projector-camera Truth Beam loop, with on-chain anchoring and a recomputable finite-sample floor.

Verify without trusting (one command, ~2 min, no GPU, no login):

curl -fsSL https://data.truthbeam.com/release/truthbeam_verify.tar.gz | tar xz && cd truthbeam_verify && bash verify_all.sh

• Public since 2023: the hash-chained projector-camera loop itself (IPFS-pinned, on-chain-anchored recordings).
• New in this release: the Markov-channel formalisation and the conditional-diffusion noise-prediction verifier that reads it.
• Scope, stated plainly: AUROC 1.000 is a finite-sample floor - same rig, two sessions, one performer, one trained non-adaptive forger. Not general deepfake detection; provenance, not semantic truth.
• Everything content-addressed: CIDs in CID_MANIFEST.json; artifacts covered by the attached GPG-signed SHA256SUMS (key A8DC9A41F92762601892A646ABFD37A9F45A5286).

Sites: https://truthbeam.com · https://poliebotics.com · https://poliepals.com