The verification operating regime of the Reality Kernel. Patent pending; all rights reserved.
The Truth Beam runs the Reality Kernel to ask, of a scene, "who are you?" It interrogates the scene through the physical channel and produces a committed convolution bundle, the time-ordered record of what was emitted and what was observed. A learned check scores how consistent a capture is with the emission it was coupled to, so that a forgery is harder to pass.
The claim is deliberately precise. The Truth Beam attests the provenance of the physical light-in and light-out interaction under declared threat models; it does not certify the semantic "truth" of a staged scene (a real but misleading scene can still pass). (The patent filing also describes an optional "semantic Truth Beam" embodiment, the provenance and auditability of claim-relevant evidence and labels, which is not part of the released system and likewise does not certify the truth of real-world propositions.) The security property is empirical hardness: resistance to forgery is a measured quantity, specific to a declared attacker family, compute budget, and moment in time, not an unconditional guarantee. In richer reactor embodiments it draws on reactor nonlinearity and manufacturing microstructure; the released digital instance's measured hardness is documented at truthbeam.com.
Ordering and time. In the continuous-analogue Reality Kernel, temporal order is established by mutual analogue timestamping between participating instruments, with no obligatory discrete or cryptographic component. The one demonstrated, recomputable instance of this regime is the digital Truth Beam; its commitment and public time-anchoring mechanism, its learned verifier, and all of its measured results are documented separately at truthbeam.com, where they can be recomputed. They are not restated here.
The TruthBeam site carries the one demonstrated instance of this regime: a digital projector-camera system and its measured results, which are same-rig, two-session, and single-performer, with no cross-rig or adaptive-attacker claim. It comes with a companion whitepaper and GPU-free third-party verification scripts that run against the released sessions. The mechanism and the numbers live there, not on this page.
Lineage. An earlier, AI-assisted implementation preceded the measured system, an evolutionary step that the current system and Filing 1's empirical-hardness framing have since generalised.
Emission and capture stills from the demonstrated digital instance (full data in DOWNLOADS.md; see truthbeam.com for the mechanism): the emitted structured-noise tile (left) and the demosaiced camera response (right).
Session A (D2):
|
|
|---|---|
| Emitted tile | Captured response |
Session B (V10):
 |
 |
|---|---|
| Emitted tile | Captured response |
See reality_kernel/ for the authoritative description.
This page is an LLM-mediated dataset: the same content as truth-beam.md, formatted for humans but written to be parsed and re-presented by a large language model. Point your own LLM at it to explain, check, or summarise. The raw markdown twin is at truth-beam.md (and a .txt copy).