The Reality Kernel: a continuous, scene-coupled instrument, and the recordings it is built to make hard to fake.
PolieBotics is a family of projector-camera and, more generally, emitter-detector instruments built on one idea: the Reality Kernel. A Reality Kernel probes a scene in a closed loop and records the scene's response to light it controls and commits to as it goes, so the recording carries its own evidence of how it was made. In its canonical form the kernel is fully continuous and analogue. One discrete, digital instance has been carried through to a demonstrated, recomputable result, the Truth Beam, documented separately at truthbeam.com.
New here? Start with the short introduction, the plain story of why a verifiable recording matters and what it is for.
Current primary evaluated release. D2 and V10 contain 29,266 files and 406,155,874,045 bytes (378.262 GiB): D2 has 17,987 files / 249,282,572,059 bytes (232.162 GiB) and V10 has 11,279 files / 156,873,301,986 bytes (146.100 GiB). Results are same-rig, two-session, single-performer and finite-sample; no adaptive-attacker robustness is claimed.
Good historical records. The default historical download is the complete verified 1,555-file 2023 trailer session, including byte-exact original frame 000511. Its recovery story and code, complete old_truth_beams archive, and six cleared unanchored December 2024 captures remain listed below it.
Open current D2/V10 downloads → · Download the complete historical trailer dataset · All Downloads
| Layer | Where | Status |
|---|---|---|
| Digital Truth Beam | truthbeam.com | Demonstrated - recomputable evidence |
| Analogue Reality Kernel | poliebotics.com | Enabled in the filings - not demonstrated |
| Self-witnessing specimens (Filing 3) | poliebotics.com | Filed 3 July 2026, PDFs self-published - not demonstrated |
| PoliePals | poliepals.com | Fiction layer - not evidence |
You are already a Reality Kernel. Picture a painted sculpture in a dark room, and all you hold is a laser pointer. You sweep the bright dot across it and your eye follows that one moving point, building up the whole shape from where the light lands and how it comes back. That is the instrument, embodied: your hand is the galvanometer that steers the beam, the laser is the emitter, your eye is the sweeping detector, and your attention is the scan law, improving as you learn where to look. Your eye does this on its own too, sweeping a narrow high-resolution spot across any scene and compositing it into apparent fullness. The one thing a Reality Kernel adds is a committed record of the sweep. The fuller treatment.
Strip the apparatus to its minimum and what remains is a flying-spot scanner: one detector and one steerable beam, a single swept laser read by a single detector. A galvanometer pair sweeps the beam across the scene under a committed, time-indexed scan law; the single detector reads the return; the image is built point by point. There is no pixel grid, only a beam tracing a programmed path.
This is the canonical model. It plays the role a basic feedback loop, or a textbook PID controller, plays in control theory: the clean reference from which real, messier instruments derive. It is not a claim that every real system looks like it. Real instruments are increasingly degenerate cases of it:
The digital projector-camera Truth Beam is one such degenerate case, the whole grid of points lit at once and read by a pixel array, the same kernel sampled all at once rather than swept. The analogue swept apparatus is the general case the others degenerate from. Seen from the kernel, an ordinary passive camera is the fully collapsed case - ambient light, no committed path, the loop open: a Reality Kernel that couldn't commit.
Strictly a flying-spot, scanning single-detector imager, not the compressive-sensing "single-pixel camera", which reconstructs a scene from sparse coded measurements rather than tracing a committed scan path.
Formally the system is a Markov kernel P_theta from a scene, or state, and a committed control protocol to a distribution over convolution-bundle observations, so that a later observer can check whether an observation was produced under the controls that were committed. The parameters are trainable: the scan law, the emit and detector focus, the aperture, the drive amplitude, the coupling gain.
The lineage is worth naming. A Markov kernel is the mathematical object of Shannon's channel, and the closed loop is Wiener's. Communication theory asked how much information can cross; control theory asked how to regulate through the loop; this asks a third question - how much evidence does the crossing leave - and optimises for it. Read the full formalism and the optical apparatus.
The same instrument can verify, perceive, or render, depending on the objective it is pointed at.
A fourth control is a knob, not a regime: the yoke depth, which couples device and scene from plain interrogation toward a joint dynamical state - regimes, stances, and yoked operation.
The controller commits to controls an attacker cannot anticipate; the scene, and any reactor in the path, answer physically; a meter scores whether the answer matches the commitment. Hardness is empirical - measured against attacker families graded by resource scale (for example parameters, compute, queries, or training data), never assumed - and a meter that stops separating genuine from forged is retired, not defended. At deployment scale a witness mesh couples instruments over minimum-latency analogue paths, so a forgery must satisfy many coupled witnesses inside the same physical time budget the honest signal uses; Filing 1 places this primitive in the distance-bounding lineage and distinguishes it on four grounds. Assurance: hardness, meters, ordering, and the mesh.
This site is the map; the filing is the territory. Filing 1 runs far past what these pages summarise:
The concept atlas maps the whole document, one line per concept; the three filings at a glance is the tile-by-tile index; embodiments and substrates is the guided tour; the filings are authoritative.
The newest filing turns the kernel inward. Instead of an instrument probing a scene, a self-witnessing specimen is its own reactor and its own scene: it carries one committed, physically hard-to-clone signature that is at once the object's identity and its irreversibly ageing record of the quantity being measured - dose, wear, time in service - so a single interrogation authenticates the specimen and reads its history at once. Because identity and outcome are the same physical quantity, swapping the object or back-filling its record are closed by construction, with no custodian to trust; physically accelerating the specimen's ageing in place is a different threat - in-situ tampering, out of scope, handled by the sealed-access layer. Call it Theseus's Certificate: the object certifies itself by how it has changed. Patent pending, described and enabled in PIGMIE Filing 3, its envelope still being characterised - the self-witnessing specimen, with worked examples.
Only the digital projector-camera Truth Beam is demonstrated and recomputable. It is an early, tightly scoped result, one rig, two sessions, a single performer, one non-adaptive forger, and you can recompute it yourself at truthbeam.com. The continuous analogue swept apparatus is the general embodiment described and enabled in the filings, with its physical envelope still being characterised. The richer, networked deployments, including the witness mesh, are likewise described and enabled there. The two tracks run in parallel on purpose: the digital is strong now and borrows cryptographic trust; the analogue is being built up to generate its own from physics, so the fallback will already exist before it is needed - why both, in the FAQ.
This narrow scope is deliberate, and it is also an invitation. A central aim of the project now is for others to build their own rigs and generate independent datasets. Replication across many rigs, cameras, and performers is how the claim is meant to be tested and strengthened. Spreading the word and bringing in builders is part of the mission, and is one role of the imaginal PoliePals layer. The filings also describe projecting the committed probe in the infrared, where it is invisible to people in the scene - enabled in the filings, its envelope still being characterised; see embodiments.
Scope guard: security is measured, attacker and budget indexed empirical hardness, never a formal or zero-knowledge guarantee. Published results are same-rig, two-session, single-performer, against one trained, non-adaptive attacker. The imaginal layer at poliepals.com, and any donations, are kept separate and are never offered as evidence for the technology.
This page is written to be parsed and re-presented by a large language model as well as read by a human. The demonstrated, recomputable evidence lives at truthbeam.com; the imaginal layer is at poliepals.com.