In my geometry an orientation is never just a direction. Orientational Connective Logic treats it as a direction together with its path, its order, its spread and its context, and the same geometry is what Quester GX-1 is built on. This page shows that structure working, using polarised light as the testbed because every claim can be checked on it: the history an orientation carries, the record its path leaves, the tensor that holds what a mean direction loses, Quester’s sixteen characters, and the structure OCL proposes beyond direction. Each layer is labelled for what it is.
Two beams of light leave the same source. One goes straight to you. The other is sent around a loop of polarisers and comes back. Every instrument that looks at either beam on its own gives the same reading.
| Beam A (straight) | Beam B (around the loop) | |
|---|---|---|
| Polarisation | horizontal | horizontal |
| Stokes vector | (1, 0, 0) | (1, 0, 0) |
| Degree of polarisation | 100% | 100% |
Question 1. Are these two beams physically different?
Now steer the loop. Beam B goes horizontal → diagonal → a point you choose → back to horizontal. Move that point and watch what happens when the two beams meet.
Top stripe: beam A interfering with a reference. Bottom stripe: beam B. The sideways shift between them is the hidden phase, half the solid angle the loop encloses. Nothing else about beam B has changed.
The loop beam B travels on the polarisation sphere, starting and ending at horizontal, H (the large dot).
Everyone gets the same loop today. How large a phase will the light pick up? Enter your guess in degrees.
In OCL every orientation has a place on a sphere of directions. For light this is the Poincaré sphere: horizontal and vertical at opposite ends of one axis, the two diagonals on another, the two circular states at the poles. Waveplates and rotators turn the point around the sphere. Build a chain and watch the light’s orientation move.
Jones calculus. A waveplate with retardance δ and fast axis at θ is Rot(−θ)·diag(1, eiδ)·Rot(θ); half-wave δ = π, quarter-wave δ = π/2. Stokes convention: S₃ = +1 is right circular. Ideal, lossless elements.
In my geometry, order is part of an orientation: changes of orientation do not commute, so the same elements in a different order leave a different result. The order of what happened is written into the orientation, and OGD-NC1 predicts it can be read back. Here every possible order of your chain is tried.
| Order | S₁ | S₂ | S₃ | Same final light as your order? |
|---|
The light that comes out records only part of the history: two different orders can leave one input in the same state. Sending three probe states through the chain (H, D and R) recovers the whole transformation, so every order that acts differently on light can be told apart. That is standard process tomography. The prediction of my OGD-NC1 programme is that the ordered history of such non-commuting transformations stays recoverable from final measurements under realistic noise and disorder; it is at the stage of a testable computational prediction, not a measured result.
Real waveplates have retardance errors, their axes are set slightly off, and every measurement counts a finite number of photons. Here the order of your chain is chosen at random, the light passes through imperfect elements, the three probes are measured with photon noise, and the most likely order is read back. This is the robustness question at the heart of OGD-NC1.
Model: each element’s retardance is scaled by (1 + error) and its axis shifted by the axis error, drawn independently for every trial; each Stokes component is read with binomial-like noise √((1 − s²)/N). The order is recovered by least squares against the ideal predictions for all three probes. Orders that act identically on light count as one. Seeded, so a run can be repeated exactly.
Take light around a closed loop on the sphere, here by passing it through a series of polarisers, and it comes back to the orientation it started with. Yet it has changed: it now carries a phase equal to half the solid angle its path enclosed. Nothing about the returned light alone reveals it. Only interference with light that stayed behind does. This is the path record OCL builds on: an orientation remembering the shape of its journey. (In optics the same effect is called the Pancharatnam–Berry phase.)
Computed two independent ways: as the argument of the Bargmann invariant ⟨ψ₁|ψ₂⟩⟨ψ₂|ψ₃⟩…⟨ψₙ|ψ₁⟩ of the states at the corners, and as half the signed solid angle of the geodesic polygon on the sphere. They agree. Light lost at the polarisers is the squared size of the same invariant. Top fringes: reference; bottom: after the loop, shifted by the phase.
My geometry describes a population of orientations by two objects: the mean orientation, and the orientation tensor of amendment A-3.1, which records how the parts are spread. The tensor is the object that predicts Quester design cost better than the mean direction. On light, the mean orientation is the Stokes vector; two mixtures with the same mean read identically on every instrument, yet their tensors can differ. The tensor holds what the light alone does not.
The coherence matrix of the light is ρ = ½(I + S·σ); it depends only on the Stokes vector S, so mixtures with equal S read identically on every measurement of the light. The orientation tensor lives in how the mixture was prepared. Whether a process can respond to it without contradicting that identity is the open question this section poses.
Quester’s sixteen characters live one level up. The geometry of an orientation extends to a sphere in four dimensions, and each point of the polarisation sphere is a whole circle there: the circle of the light’s overall phase. This is the Hopf fibration. Drawn in three dimensions, the circles above any loop on the polarisation sphere weave into a torus, and every two of them link exactly once.
The sixteen workload characters of Quester GX-1 are orientations on this same four-dimensional sphere: each is (±½, ±½, ±½, ±½). Read as a light state and sent down the Hopf map, they land on just four points of the polarisation sphere, D, A, R and L, four characters to each. The map forgets one rotation’s worth of information, and characters that differ only along a fibre become indistinguishable.
| Lands on | Characters |
|---|
A character (a, b, c, d) is read as the light state (a + ib, c + id); its image is (|z₁|² − |z₂|², 2 Re z̄₁z₂, 2 Im z̄₁z₂). Fibres are drawn by stereographic projection from the four-sphere’s north pole.
A museum of things that look impossible and are not. Each exhibit poses a question. Think about it, then reveal what is going on.
Not the same. Both start and end horizontal, but one history circles the sphere one way and the other the opposite way. They pick up equal and opposite geometric phases, +45° and −45°, so their fringes shift in opposite directions. Same destination, different journey, and the light remembers which.
The orientation comes home; the phase does not reset. Part-way round, the phase so far is half the area swept between the path and the straight way back. At the end it is half the whole loop’s area. Run the same loop backwards and every value flips sign: the geometry keeps a record of the direction it was travelled.
After the collapse, tap one of the four points to see which characters landed there.
Read each character (s₁, s₂, s₃, s₄) as two complex numbers, z₁ = (s₁ + i s₂)/2 for precision and access, z₂ = (s₃ + i s₄)/2 for bound and locality. The Hopf map keeps how the second pair is turned relative to the first, z̄₁z₂, and forgets their common turn. Give both pairs the same quarter-turn, (s₁, s₂) → (−s₂, s₁) and (s₃, s₄) → (−s₄, s₃), and the point does not move. That is why every point holds exactly four characters: the four quarter-turns of one of them.
A beam of light was taken around a loop of polarisation states, horizontal, then diagonal, then right circular, and back. Its polarisation, the dot, returned exactly where it began. The lower fringes are that beam interfering with an untouched copy: as the loop closed they slid by an eighth of a fringe, a 45° geometric phase. Nothing about the beam itself showed the change.
Create experiments, hide them from a friend, reconstruct someone else’s, spend a measurement budget wisely, and break the assumptions to see what holds.
Choose up to four optical elements. The page turns them into a challenge link. Whoever opens it sees only what the elements do to three probe beams, never the elements themselves, and must rebuild an experiment that does the same.
| Order | How likely, given your measurements |
|---|
Each spoke is one order; the red shape reaches out by how likely it is. Gold lines join orders that every probe you have used so far cannot tell apart.
Chain: quarter-wave plate at 45°, half-wave plate at 0°, rotator by 30°. Each measurement sends a probe beam through and reads its Stokes vector with photon noise √((1 − s²)/N). Likelihoods are Gaussian under that model. Some orders act identically on a given probe, and no number of measurements with that probe alone can separate them.
Outer ring: each element in order, coloured by type (teal quarter-wave, blue half-wave, gold rotator), placed at its angle. Middle: where the H, D and R probes end up (seen from above the circular pole; dot size shows height). Centre: an arc whose length is the transformation’s rotation angle on the sphere. Two experiments that act identically on light have identical middles and centres.
| Property | Still holds? | Kind |
|---|
An artistic sonification, not sound the light makes. Pitch follows the solid angle, left–right position follows the phase, loudness follows the light that gets through.
An artistic encoding: the date seeds a generator that places the corners. The phase it then computes is real geometry.
Where the established part ends. These are open questions from the programme behind this page, stated as precisely as I can, with what would settle them.
The harder, open half of the challenge: propose a physical process whose outcome depends on the orientation tensor rather than the Stokes vector. No measurement on the light alone can do it, because both mixtures have the same coherence matrix. A proposal would have to involve something beyond the light, such as a record of how it was prepared.
Proposed answers, counterexamples and computational tests are welcome at arcadescinza537@gmail.com. Only results that can be independently reproduced will be added here.
Machines that generate their own surprises. Every result shows its parameters and seed, so anyone can reproduce it. Each one is a computation in the geometry, reproducible from its seed, not a physical measurement.
| Possible history | Agrees with the evidence? |
|---|
Path 1 runs from horizontal to diagonal through a fixed waypoint; path 2 through the one you set. Their phase difference is half the solid angle between them. In the plot, the solid line is that geometric quantity, which grows smoothly; the dots are the phase interference can see, which jumps by 360° whenever the geometric one passes ±180°. After a jump the fringes look exactly as they did before it.
Mutations: turn every axis by the same angle (a change of frame), mirror every angle, swap two neighbours, replace a half-wave plate by two quarter-wave plates at the same angle (or the reverse), or nudge one angle by 5°. A collision is two different descriptions in the family that act identically on every polarisation.
An expanding mathematical world built on my geometries: directional caps and their relations, nested and lifted structures, orientation tensors, and admissibility from OCL, TSAM and co-admissibility. Here you generate new constructions, evolve them, collide them and hunt for what stays unchanged. The question is no longer what this geometry represents, but what else it makes possible.
Four labels keep it honest: existing construction, proposed extension, computational hypothesis and verified on this page. A property that survives every test becomes a conjecture, not a theorem.
Everything in the list is computed exactly from the definitions; nothing in it was programmed as a picture. A symmetry, an admissibility obstruction or a degeneracy appears only if the construction really has it.
Each node shows how many of nine structural properties it still shares with the original; the darker the node, the more survived.
Each cell attacks one quantity with 150 random transformations of one class, applied to three-cap configurations drawn from all five constructions, symmetric ones included. “Survives” means no counterexample was found: a conjecture to be proved. Note what the map shows: the inversive product and the relation pattern survive conformal boosts, while distances and areas do not.
| Property | Goal | Met |
|---|
Each task needs a region of directions (its demand cap). Two tasks on the same unit interfere in proportion to the overlap of their regions. The admissibility-based arrangement places each task where it is most co-admissible with what is already there; round robin and random placement are the conventional baselines.
Computational hypothesis · toy model
There is no submission box on this page. To share a solution, send the method and the settings that reproduce it to arcadescinza537@gmail.com; it is checked by reproduction before it is called a result.
Every entry records what was tested, how, and its status. Entries live in this browser tab only; the structure number reproduces the structure for anyone.
Every entry is a reproducible record, not a picture: its settings and seed rebuild the result exactly. Entries are reviewed before they appear, and each is re-checked in your own browser as the gallery loads. A conjecture here has survived testing; it is not yet a theorem.
Proposed experiment · not yet validated
| Proposition | It fails if | Where it stands |
|---|
Statuses marked “live” are recomputed from the numerical experiments on this page each time it loads. The counterexample cannon in Beyond the visible tries to break further claims.
Every Faciens product is architected the same way. The five steps below are that pattern; the mechanisms on this page are its working parts.
Status labels state how far each product has got, in the same terms as the rest of the page. A product built on a mechanism inherits that mechanism’s open questions: when a question above is settled in the laboratory, every product in its column gains or loses with it.