Appearance
PSDG: Three Lessons for Physics — From Cauchy Surfaces to Superposition
What a solved children's game might suggest about themes in 20th century physics
Lens only — not physics
This page is an informal essay for readers coming from physics. It is not a theory of spacetime or quantum mechanics, not new predictions or equations, and not a claim of formal isomorphism between PSDG and GR or QM. The only exact, checkable claims for the project remain the benchmarks and oracle on the rest of this site (technical report, snapshot). Linking here is intentionally only from the FAQ.
Twentieth century physics learned three hard lessons. Each took decades. Each required a revolution. Each arrived through different machinery — general relativity, matrix mechanics, quantum theory — and so each was filed as a separate discovery belonging to a separate subfield.
A small deterministic dice game called PSDG (Philosopher's Stone Dice Game) suggests they may rhyme with a common pattern when viewed through one window: when commitments bind relative to when consequences become evaluable, and which summary of the world you treat as sufficient.
PSDG is a two-player tabletop game. Six dice, a small mat, roughly ten minutes to play. After a random setup, every decision is deterministic with perfect information. The state space is small enough for exhaustive enumeration. An exact oracle (solver) exists. There is no hidden information during play, no randomness after setup, no chaos, no complexity barrier.
It is a children's game. It is also, on inspection, a minimal system where three themes that physics has discussed at length — prediction from a chosen “surface,” representational sufficiency, and reasoning before an irreversible choice — appear in discrete form without Lorentzian manifolds, complex amplitudes, or laboratory measurement. The point below is pedagogical and philosophical, not a reduction of physics to dice.
Ontic state vs model state (read this first)
Full (“ontic”) game state in PSDG — everything the rules care about at a moment — includes draft commitments (e.g. top and facing per twisted die) and everything else the v1.13 spec tracks. After setup, that information is in principle available to anyone looking at the table with the rules in hand: there is no poker-style hidden draw mid-game.
Board snapshot in what follows means a compressed model state: “what an agent, learner, or story in the prose chooses to treat as the state” — often visible tops and a few salient scores. That summary can collapse distinct true positions into one picture.
So when we say the snapshot “fails to propagate” or is “insufficient,” we mean: relative to a fixed coarse representation and a fixed decision task (optimal play), the summary does not support the same inferences the full specification does. We are not claiming that the universe of the game fails determinism, or that no Cauchy-style mathematical question exists in GR — only that a familiar epistemic shape (“I thought my data were enough”) appears when model ⊂ reality even under full physical visibility.
Lesson One: Propagation (rhyme, not identity)
The Cauchy surface problem (physics)
In general relativity, a Cauchy surface is a spacelike slice from which, in a globally hyperbolic spacetime, one can evolve the fields and recover a well-posed initial-value problem. Where global hyperbolicity fails, initial data on a slice may not suffice to determine the future in the same way — causal structure can place limits on predictability from a single surface.
That is a precise story about Lorentzian geometry, null geodesics, and extensions of spacetime. It is not defined by “what a human wrote on a notepad.”
What PSDG shows (compressed summary vs full spec)
PSDG has no metric. It has rules and commitments.
During the draft, each Twist locks both a Phase 1 top and a Phase 2-relevant facing. Those choices are irrevocable before Exchange, Phase 2 scoring, and tiebreakers.
If your model is only the board snapshot in the coarse sense above — say, visible tops — then two ontically different positions (different facings, same tops) can require different optimal play. The information needed to distinguish them is on the table in the full state; it is missing from the summary the model uses.
Rhyme with Cauchy-flavored unease: a surface you hoped would support reliable continuation (here: “I know enough to act”) fails because relevant degrees of freedom are not encoded on that surface. The mechanism in PSDG is lossy encoding / aliasing, not Einstein equations on a bad manifold.
We do not claim the mathematics of Cauchy horizons is the mathematics of PSDG. We claim a similar question can be asked in both settings: does this chosen “initial data” support the inferences I want to make about what comes next? In GR the obstacle can be global causal structure; in PSDG it can be using the wrong state vector for the task even when nothing is hidden.
Lesson Two: Representation (pedagogical parallel)
Matrix mechanics (physics)
Heisenberg’s move to observable transitions pushed the formalism toward two-index objects (matrices) and non-commuting structure — a deep shift in how “state” and “quantity” are represented.
What PSDG shows (two labels vs one)
The coarse snapshot behaves like a single-index story: “what the tops show now.” The oracle must track commitment pairs and phase coupling — effectively more than one scalar label per die if you want structural sufficiency for optimality.
Pedagogical parallel: insisting on a too-small state for the actual dynamics forces you to conflate situations the rules separate — the same flavor as learning that a too-naive configuration space cannot carry the physics you need. We do not claim PSDG reproduces Hilbert space, spectra, or the uncertainty principle as theorems.
Rule order: Twist-before-Tumble is not Tumble-before-Twist — the rules define a sequencing of commitment and evaluation. That is order dependence in a discrete update rule. Calling it “non-commutativity” is metaphor unless you exhibit an operator representation — useful for intuition, not a claim of quantum structure.
Lesson Three: Branching before commitment (decision-theoretic rhyme)
Superposition (physics)
Pre-measurement superposition in QM is a specific mathematical object: complex amplitudes, interference, Born rule, entanglement, and the measurement problem — none of which live in PSDG.
What PSDG shows (live branches in the extensive form)
Before a Twist, several legal commitments remain live; the oracle evaluates all continuations relevant to value. Choosing a Twist prunes the tree irreversibly.
Rhyme: reasoning well before commitment requires treating multiple futures as still possible for evaluation, not collapsing them to a guess that one was “already true.” That is close to how good backward induction behaves in a finite game — not a claim that the ontology of the die is “in superposition.”
Hidden-variable caution: equating “I’ll pretend I know which branch wins” with quantum hidden-variable debates overstates the parallel. The fair statement is epistemic / procedural: bad modeling of pre-choice uncertainty in a deterministic tree can look like classical ignorance mistaken for structural reasoning.
One structure, three windows (tentative)
The three rhymes are linked in PSDG: a coarse snapshot fails partly because commitment pairs and branching matter to value; representation and branching interact because the oracle sums over what is still live at each node.
Physics encountered related themes through different formalisms — geometry, spectra, interference. PSDG offers a toy where none of that machinery is necessary to see commitment-before-evaluation and summary-vs-full-state bite.
Burden (interpretive, not logical): if someone insists these themes are only “about” their native physics packaging, they still owe an account of why a six-dice rule set can exhibit separable, checkable instances of (i) “surface too small for the task,” (ii) “need richer labels than the obvious photo,” and (iii) “don’t collapse the tree before you compute value” — without importing GR or QM. That is philosophy of representation, not a proof that physics reduces to games.
What this does and does not claim
PSDG is a benchmark, not a theory of physics. Falsifiable claims live in ML, safety, and game theory — regret, win rates, protocols.
For physics, this page offers a lens only: no new predictions, no new equations, no claim that Cauchy horizons = coarse snapshots in any technical sense.
The fair claim is conditional: if you care about when data suffice for prediction, what state vector to use, and how to reason before irreversible acts, then PSDG is a small place where those questions are sharp and verifiable — orthogonal to whether you work in manifolds or Hilbert spaces.
The uncomfortable part is simplicity: six dice, published rules, a child can play, an oracle exists — yet coarse summaries still fail the task in ways that rhyme with much heavier science. That is an argument about pedagogy and humility, not a replacement for physics.
On this site: FAQ — physics implications · Home · Technical report (summary)
PSDG (Philosopher's Stone Dice Game) — psdg.pages.dev · GitHub — psdg