The Prime Radiant: a golden-path navigator over Γ

What this document is

UHM already fixes a state space ($\Gamma \in \mathcal{D}(\mathbb{C}^7)$), a law ($\mathcal{L}_\Omega$), cheap observables ($P$, $R$, $\Phi$, $C$, $\sigma_k$, Gap), thresholds ($2/7$, $1/3$, $1$, $2$) and a transfer engine that reads persons, organizations, machines and markets into the same $\Gamma$ (Domain Transfer). This document assembles those parts into a solver: a machine that (i) encodes a situation, (ii) diagnoses it with an address, (iii) searches paths to a goal region, and (iv) reports what fraction of paths is golden — with every claim printed as a measured number by a reference implementation (§3, the machine). The name honors Asimov's device — and §6 states precisely which half of the fantasy survives contact with the theorems. The method throughout is categorical calibration (§5): a symbolic system is exactly as good as its ability to describe the computational model of reality, so every symbol here earns — or loses — its number in a re-runnable table.

§1. The door and the claim

The claim to be examined is the strongest available: on the UHM basis one can build a universal solver. Made naively, the claim is false — «universal» collides with the freedom theorem (the kernel $\dim\ker H_\Gamma + 1 > 1$ makes point prophecy structurally impossible, Freedom), with chaos, and with the honesty registry itself. Made precisely, it survives and becomes buildable. The precise form has three parts:

1. One state space for everything that holds together. Whatever can be read as seven interacting aspects — a mind, a team, a model, a market — is encoded as $\Gamma$ by the Enc functor; the seven dimensions apply fractally across levels (holarchic ladder [Т/И], transfer levels I–V). This is why «the world is full of analogies»: the analogies are one structure instantiated at many scales, and most of them have simply not been noticed yet.
2. A solver solves toward viability, not toward arbitrary goals. The law itself carries every pattern toward the viable window (the T-124 attractor, Conscious window); goals aligned with that drift are assisted by the dynamics; goals fighting it pay $\sigma$ continuously. «The universe removes obstacles» is a theorem-shaped experience: past the basin boundary, the vector field points your way (§4).
3. The output is a navigation, never a verdict. For ensembles the machine predicts; for persons it opens doors and prices paths. The boundary between the two is itself a theorem (§4.4), and it is the feature that keeps the Radiant a navigator instead of a cage.

Why should such a machine work at all? Because in UHM the levels of reality share one form. The diagonal of $\Gamma$ is a population — who is loud at this level; the off-diagonal coherences are the signal running between the voices and, through the Enc functor, down the holon hierarchy: a cosmos writes its levels, a society writes its members, a mind writes its habits — the same $7\times 7$ form each time (levels I–V of the transfer engine). That is why prediction works where it works. And it is why symbolic systems are not commentary about the world: the corpus's deepest floor holds that the Universe is a self-modelling holon — its self-description $\rho^* = \varphi(\Gamma)$ is inside the state (a Lawvere fixed point; observers are internal sections, T-221 — Universe as Holonom). Read against that floor, every calibrated symbolic system is an organ of the universe's self-cognition [I], and this machine is a small computational section of the self-model it models — which is exactly why the calibration table of §5, not any manifesto, is its licence to exist.

Everything below is built from three existing research programs — the Γ-canon (the symbolic layer and the П4 oracle protocol), the Σ-calculus (diagnosability rigidity T-224/T-225), and Coherence Cybernetics (the $\sigma$-panel and interventions) — plus two floors of mathematics that the solver runs on (§2).

§2. The two floors of residues

The word residue names the solver's core mathematics twice, and the coincidence is instructive: both floors extract a local address from a closed circuit.

2.1 The arithmetic floor: quadratic residues mod 7 [T]

The quadratic residues modulo 7 are $QR(7) = \{1, 2, 4\}$ — the squares $1^2, 3^2, 2^2 \bmod 7$. The corpus already rests on them once: $N_{\text{gen}} = |QR(7)| = 3$ [T]. The solver rests on them completely, because the entire Fano wiring is their arithmetic:

• $QR(7)$ is a $(7,3,1)$ perfect difference set: every nonzero residue arises exactly once as a difference of two elements. Hence the seven translates $\{1,2,4\}+k \bmod 7$ are seven triples in which every pair of points lies on exactly one triple — the Fano plane, wired by addition [T].
• In the corpus numbering ($A{=}1, S{=}2, D{=}3, L{=}4, E{=}5, U{=}6, O{=}7$ — note $U{=}6$, $O{=}7$: the arithmetic numbering swaps U and O relative to the display order $A,S,D,L,E,O,U$) the translates reproduce the seven named lines of Fano selection rules exactly: $\{A,S,L\}, \{S,D,E\}, \{D,L,U\}, \{L,E,O\}, \{A,E,U\}, \{S,U,O\}, \{A,D,O\}$ — machine-checked 7/7 (H03).
• The multiplier theorem in miniature: $2 \in QR(7)$, so $x \mapsto 2x$ maps the difference set to itself and therefore permutes the seven lines — this is the $\mathbb{Z}_3$ Frobenius symmetry of the selection rules, and its orbit structure on lines is $1+3+3$ (H04).
• The three lines through the clock voice $O$ pair the remaining six voices as $(A,D), (S,U), (L,E)$ [T] (H10) — the pairing that underlies the sectoral split of Spacetime.
• The same triples are the weight-3 codewords of Hamming $H(7,4)$: a single-axis fault is its own address (the syndrome equals the binary of the axis; H09, and in depth Σ-calculus §4). The corpus uses two coordinatizations of this one object — the XOR labeling of the Σ-calculus and the octonionic/QR labeling of the selection rules; the machine finds the dictionary between them, and it is the tidy involution $3 \leftrightarrow 4,\; 5 \leftrightarrow 6$ on labels (H07).

So the arithmetic floor gives the solver its diagnosis organ: where the pattern is broken, the broken lines spell out the address — no search required. That is the practical content of «a world woven on a self-checking alphabet complains with an address».

2.2 The analytic floor: Cauchy residues of the resolvent [T]

The second floor is the residue calculus of complex analysis, entering through the resolvent $(z - \Gamma)^{-1}$. For Hermitian $\Gamma$ with spectral decomposition $\Gamma = \sum_j \lambda_j P_j$:

$$\frac{1}{2\pi i}\oint_{\mathcal{C}} f(z)\,(z-\Gamma)^{-1}\,dz \;=\; \sum_{\lambda_j \in \mathrm{int}\,\mathcal{C}} f(\lambda_j)\,P_j ,$$

the sum of the residues of the resolvent inside the contour. Every observable of the theory is a functional of such objects: spectral projectors ($f \equiv 1$), entropy and $D_{\text{diff}}$ ($f = -z\log z$, on a cut-avoiding contour), Gibbs weights ($f = e^{-\beta z}$), the Gap spectrum on the $E$-sector, filtered subspaces for the reflection tower. The reference machine implements the contour engine and verifies it against direct diagonalization to machine precision: projector idempotence $|P^2 - P| \sim 10^{-17}$, completeness $|\sum_j P_j - I| \sim 10^{-16}$, $\operatorname{tr} e^{-2\Gamma}$ agreeing to $10^{-15}$ (H11–H15). At $7\times 7$ this engine is a luxury; at the composite scales of v1 ($49$-dimensional pair spaces, ecologies) contour extraction of one needed cluster without full diagonalization is the standard scalable method (FEAST-class), and the solver is born with it.

2.3 The residue principle [I]

Why do two unrelated mathematics share one name here? Both floors perform the same gesture: walk a closed circuit, and what survives the cancellation is a local invariant — the difference-set count that names a faulty axis; the contour integral that names an eigenspace. This is also the Gap-as-holonomy gesture of the corpus (go around a loop, return changed, read the change — Universe as Holonom). As an interpretation [I], the solver's motto: обойди контур — получи адрес; the circuit is the computation.

§3. The machine

The machine exists twice, deliberately. The reference is a numpy-only implementation — deterministic, self-verifying; a sibling of the TALOS v0 emulator whose canonical tick it reuses. The instrument is a Rust implementation with a live TUI (ratatui): eight tabs (Holon dashboard with the $|\Gamma|$ heatmap and σ-panel, Atlas, Oracle, Pair, Ensemble, Reading, Calibration, Help), spectrum-preserving dial controls, bold-mode, and domain presets; its --selftest runs a 28-hypothesis battery (28/28 VERIFIED) and reproduces the Python attractor to $P^* = 0.3208$ vs $0.321$ — a cross-language replication of the whole dynamical core. The kernel is generalized: one state space, one tick, one observable suite — what changes between application domains is only the gloss vocabulary and the parameter system (the Model presets: Universal, Mind, Team, LLM agent, Market), which is the precise sense in which «only the application models and the equation systems change». Organs of the reference machine:

Organ	Contents	Verification
[A] arithmetic floor	$QR(7)$ wiring, $\mathbb{Z}_3$ multiplier, Hamming decode, XOR↔QR dictionary	H01–H10, H46
[B] analytic floor	contour projectors, $f(\Gamma)$ via residues	H11–H15, $10^{-15}$ agreement
[C] observables	$P$, $R$, exact $\Phi$ (not the $7P{-}1$ proxy), $C = \Phi \times R$, $S$, $D_{\text{diff}}$, $\mathrm{Coh}_E$, the T-92 stress panel	H33–H38
[D] dynamics	the $\mathcal{L}_\Omega$ tick (Strang split, CPTP-safe), $\kappa(\mathrm{Coh}_E)$ feedback	H16–H32
[E] navigator	goal regions, SO(7) dials, bold moves, the solitaire oracle, ensembles	H39–H45
[F] calibration	the 66-hypothesis table, re-run on every change	§5
[G] pair space	$\mathcal{D}(\mathbb{C}^{49})$: bridges, the T-77 gain, the κ-budget contest	H47–H50
[H] towers	the centered-Gram meta-holon; the purity ladder	H51, H45
[I] geodesics	the m-chord identity of regeneration; path straightness	H52–H53
[J] phase atlas	the (dissipation, supply) basin map	H54–H55
[K] estimator	predict–measure–update over $\mathcal{D}(\mathbb{C}^7)$: full and partial noisy readouts, process noise	R26–R27, demo 6
[S] domain sensor	an external system → observation stream; git repository (v0.7 metadata-only; v0.8 coherence-bearing: per-commit feature correlations unlock $\sigma_O/\sigma_U$ — the κ₀ supply loop ignites exactly with the repo's inner-workshop era)	§3.10 item 7, held-out p = 0.014
[L] BKM floor	the Kubo–Mori metric, m/e-geodesic lengths, geodesic MPC	H60–H64

Two engineering facts deserve their own sentence. The wiring catch: the calibration discovered (H46) that the TALOS v0 emulator's translate wiring, read in display order $A..U$, realizes the corpus lines only up to the $U/O$ naming swap of §2.1 — functionally isomorphic (all spectral observables unaffected), but any line-named diagnostic must use the corrected wiring; the machine does. The proxy catch: the emulator's $\Phi = 7P - 1$ is a uniform-diagonal proxy; on a generic state it deviates from the canonical $\Phi$ by tens of percent (measured: $3.84$ vs $2.45$; H34) — the solver computes the exact one.

3.1 The state and the observables

The state is $\Gamma \in \mathcal{D}(\mathbb{C}^7)$: Hermitian, positive-semidefinite, $\operatorname{tr}\Gamma = 1$; seven populations $\gamma_{ii}$ and twenty-one coherences $\gamma_{ij}$. Everything the machine prints is one of:

$$P = \operatorname{tr}\Gamma^2, \qquad R = \frac{1}{7P}, \qquad \Phi = \frac{\sum_{i\neq j}|\gamma_{ij}|^2}{\sum_i \gamma_{ii}^2}, \qquad C = \Phi\,R,$$ $$S = -\operatorname{tr}(\Gamma\ln\Gamma), \qquad D_{\text{diff}} = e^{S}, \qquad \mathrm{Coh}_E = \sum_{j \neq E}|\gamma_{Ej}|^2, \qquad \kappa_0 = \omega_0\,\frac{|\gamma_{OE}|\,|\gamma_{OU}|}{\gamma_{OO}},\quad \kappa_{\text{boot}} = \frac{\omega_0}{7}.$$

Viability: $P > 2/7$; the window's upper wall $P \le 3/7$ is the inversion of $R \ge 1/3$ through $R = 1/(7P)$. The T-92 stress panel, componentwise:

$$\sigma_A = 1 - \frac{\gamma_{AA}}{P},\quad \sigma_S = 1 - \frac{\operatorname{rank}\Gamma_S}{3},\quad \sigma_D = 1 - 7\gamma_{DD},\quad \sigma_L = \frac{7(1-\gamma_{LL})}{6},$$ $$\sigma_E = 1 - \frac{D_{\text{diff}}}{7},\quad \sigma_O = 1 - \frac{\kappa_0}{\kappa_{\text{boot}}},\quad \sigma_U = 1 - \frac{\Phi}{\Phi_{\text{th}}},\qquad \Phi_{\text{th}} = 1.$$

At the grey state $\Gamma = I/7$: $\sigma_L = \sigma_U = 1$ exactly — the wall (H33).

3.2 The tick

One step of $\mathcal{L}_\Omega$, Strang-split and CPTP-safeguarded:

$$\Gamma_{n+1} = \Pi_{\mathrm{PSD}}\!\left[\, \mathcal{U}_{dt/2}\Big( \mathcal{U}_{dt/2}(\Gamma_n) + dt\big( g_D(\tfrac{I}{7} - \Gamma) + \kappa\,(\rho^* - \Gamma)\big) \Big)\right],$$ $$\mathcal{U}_{\tau}(\Gamma)_{jk} = e^{-i(\lambda_j-\lambda_k)\tau}\,\gamma_{jk}, \qquad \kappa = g\cdot 1.2\,(1 + 8\,\mathrm{Coh}_E),$$

with $H_{\text{eff}} = \mathrm{diag}(0,\,0.6,\,1.0,\,1.6,\,3.0,\,2.0,\,2.4)$ ($\lambda_E$ highest, the A5 order), $dt = 0.01$, $g_D = 0.2$, gain $g = 1$ ($3$ on a bold move). The three terms are the whole physics: the unitary preserves $P$ and $S$ exactly (H16); the dissipator alone flows to $I/7$, $S \to \ln 7$ (H18) — solve; the regeneration alone rides the mixture chord to $\rho^*$ (H52) — coagula. $\Pi_{\mathrm{PSD}}$ is the spectral clip $\lambda \mapsto \max(\lambda, 0)$ with renormalization.

3.3 Diagnosis

The fault profile $x \in \mathbb{F}_2^7$ meets the Hamming check matrix $H$ (columns $=$ binary of $1..7$): the syndrome $s = Hx$ is the address — for a single fault on axis $a$, $s = \mathrm{bin}(a)$ (H09). Continuously, the line-closure functional $c_\ell = |\gamma_{ij}| + |\gamma_{jk}| + |\gamma_{ik}|$ over the seven lines $\ell$ localizes degradation the same way (H36).

3.4 The contour engine

The trapezoid discretization of the residue formula on a circle $z_t = c + re^{i\theta_t}$, $\theta_t = 2\pi(t+\tfrac12)/T$:

$$f(\Gamma) \;\approx\; \frac{1}{2\pi i}\sum_{t=0}^{T-1} f(z_t)\,(z_t I - \Gamma)^{-1}\,\Delta z_t, \qquad \Delta z_t = \frac{2\pi i r e^{i\theta_t}}{T},$$

which for $f$ analytic inside the contour converges exponentially in $T$ (the trapezoid rule on a periodic analytic integrand): at $T = 256$ the machine reaches $10^{-15}$ (H13). One contour around one cluster extracts one projector without touching the rest of the spectrum — the property that scales.

3.5 The pair law — derivation

Let $\Gamma_p = \Gamma_1 \otimes \Gamma_2$ and the bridge $X = \gamma\,C + \bar\gamma\,C^\dagger$ with $C = (|i\rangle\langle j|)\otimes(|k\rangle\langle l|)$, $i\neq j$, $k \neq l$. Then

$$P(\Gamma_p + \varepsilon X) = \operatorname{tr}(\Gamma_p + \varepsilon X)^2 = P_1 P_2 \;+\; 2\varepsilon\operatorname{tr}(\Gamma_p X) \;+\; \varepsilon^2 \operatorname{tr}X^2, \qquad \operatorname{tr}X^2 = 2|\gamma|^2,$$

— three lines, exact, no approximation; the machine confirms the identity to $4\cdot10^{-17}$ (H47). And because $\langle i|j\rangle = \langle k|l\rangle = 0$, both partial traces of $X$ vanish identically: $\operatorname{tr}_2 X = \operatorname{tr}_1 X = 0$ — the reduced states of both members are untouched (H49). The gain has nowhere to live but the bond.

3.6 The ladder — derivation

$$P^{(m)}_{\text{crit}} = \frac{2}{7}\cdot\frac{3^{\,m-1}}{m+1}, \qquad \frac{P^{(m+1)}}{P^{(m)}} = \frac{3(m+1)}{m+2} > 1 \;\;\forall m \ge 1,$$

so the sequence is strictly increasing: $\tfrac27,\ \tfrac{9}{14},\ \tfrac{54}{35}$ at $m = 2, 3, 4$. Since $P = \operatorname{tr}\Gamma^2 \le \operatorname{tr}\Gamma = 1$ for any density matrix, $m = 4$ demands the impossible — the ceiling of three (H45, axiom Ω⁷).

3.7 Navigation as an optimization problem

Given $\Gamma_0$, the goal region $\mathcal{G} = \{\Gamma : P \in (2/7, 3/7] \wedge C \ge C^*\}$, and the control set $\mathcal{A} = \{R_{ij}(\pm\theta)\}_{i<j} \cup \{\kappa\text{-boost}\} \cup \{\text{drift}\}$:

$$\min_{u_{0:T}} \; T + \lambda \sum_{t} \mathrm{heat}(u_t) \quad \text{s.t.} \quad \Gamma_{t+1} = \mathcal{T}(\Gamma_t, u_t), \;\; \Gamma_T \in \mathcal{G},$$

where $\mathrm{heat}$ is the measured mean stress of a boosted step. v0 solves it greedily (one-step lookahead over the 43 actions); v1 replaces the search with $m$-geodesic shooting in the BKM metric (T-263) — the down-payment is H52/H53.

3.8 The oracle and the ensemble — error bars

The solitaire estimate is binomial: $\hat p = W/M$, $\mathrm{se}(\hat p) = \sqrt{\hat p(1-\hat p)/M}$ — at $M = 120$, a $0\% \to 25\%$ contrast is $\sim 6\,\mathrm{se}$ apart. The ensemble mean of $N$ holons carries $\mathrm{se} = \sigma_P/\sqrt{N}$; the measured WHERE-compression ($0.173 \to 0.0154$) is the attractor's doing, on top of which the $\sqrt N$ law prices any level-forecast — and the WHO-distance staying at $0.488$ is what no $N$ can average away.

3.9 A worked example, digit by digit

Take the explicit state (all numbers below are machine-computed, not narrated): populations $\mathrm{diag}(\Gamma) = (0.46,\, 0.09,\, 0.07,\, 0.07,\, 0.14,\, 0.10,\, 0.07)$ and three real coherences $\gamma_{AE} = 0.08$, $\gamma_{EO} = 0.07$, $\gamma_{OU} = 0.05$ (Apperception, Immanence, Completeness — an A-loud pattern fed through the E–O–U spine). It is physical: $\lambda_{\min} = 0.0128 > 0$, $\operatorname{tr}\Gamma = 1$.

$$\textstyle\sum_i \gamma_{ii}^2 = 0.2640,\qquad \sum_{i\neq j}|\gamma_{ij}|^2 = 2(0.08^2 + 0.07^2 + 0.05^2) = 0.0276,$$ $$P = 0.2916 \in (2/7, 3/7] \;\text{(viable)},\quad R = \frac{1}{7\cdot 0.2916} = 0.4899,\quad \Phi = \frac{0.0276}{0.2640} = 0.1045,\quad C = 0.0512,$$ $$S = 1.5271,\;\; D_{\text{diff}} = 4.60 \ge 2,\qquad \mathrm{Coh}_E = 0.08^2 + 0.07^2 = 0.0113,\qquad \kappa_0 = \frac{0.07 \cdot 0.05}{0.10} = 0.0350.$$

The panel: $\sigma = (-0.58,\; 0.00,\; 0.51,\; \mathbf{1.09},\; 0.34,\; 0.75,\; 0.90)$. The reading writes itself: a viable, A-dominant pattern whose Logic voice is starved past the wall ($\gamma_{LL} = 0.07$, $\sigma_L > 1$) and whose supply runs thin ($\kappa_0 \ll \kappa_{\text{boot}} = 0.143$, $\sigma_O = 0.75$) — prescription: restore the $L$-flow first, then feed $O$. Note the teaching moment of H38 live: the boolean $\max\sigma < 1$ criterion disagrees here with $P > 2/7$ — on the uncalibrated panel the ranking (L, then U, then O) is the trustworthy output, and this single example shows why calibration item 4 of §9 exists.

3.10 Honest demonstrations

«Real demonstration» has a definition here, enforced by construction, with four protocols shipped as subcommands of the Rust instrument (prime-radiant demo blind|calibration|bench, prime-radiant checkup):

1. Blind ground truth (demo blind): the RNG secretly damages one axis of a healthy state; the diagnoser observes only the seven line closures (aggregates, not individual coherences) and must name the axis — «the broken lines spell out the address» tested as an instrument, full confusion matrix printed, nothing curated. Measured: 200/200 = 100% under heavy damage; 193/200 = 96.5% under mild damage (crush to 70–95%) with 3% observation noise — the seven errors all confuse S↔D, the two voices sharing the line $\{S,D,E\}$, and the instrument's failure mode is thereby itself addressed.

2. Proper scoring (demo calibration): the oracle's $\hat p$ from 40 deals (seed group A) is tested against 40 fresh deals (disjoint seed group B) across unselected random starts. Measured: mean $|\hat p - \text{freq}| = 2.5\%$ against a binomial se of $7.9\%$; Brier $0.100$ (coin $= 0.25$). The oracle is calibrated — its probabilities mean what they say.

3. Baselines on identical starts (demo bench): four policies race from the same 30 random fogs. Measured: drift $121.4 \pm 11.3$ ticks; random dial-fiddling $148.8 \pm 25.2$ — worse than doing nothing (無為 measured: fussing beats stillness nowhere); bold start $30.5 \pm 3.0$; adaptive boost-below-the-wall $30.5 \pm 3.0$. All 30/30 succeed within budget.

4. Real input (checkup --scores 7,5,4,6,3,2,5 --coh AE=0.4 --model mind|team|llm|market): the user's own seven ratings (plus optional named coherences) are encoded by the printed protocol (populations $=$ scores$/\Sigma$; $\gamma_{ij} = c\sqrt{\gamma_{ii}\gamma_{jj}}$, PSD projection) and read by the interpretation organ — named populations, carrying coherences, the loudest stress with the domain prescription. Two honesty clauses print with every result: the self-report Enc is a protocol [О/И], falsifiable by retest, not an oracle pronouncement; and $P$ from self-report measures the contrast of the profile — near-uniform ratings land near the grey state by construction, so the verdict line reads profile shape, not a life sentence.

5. A maxim, decomposed and measured (demo maxim): «a person must have a goal of their own; without one, they become part of another's goal.» The reduction of meanings is printed with the run: goal $=$ the regeneration target $\rho^*$; own goal $=$ the share $\varphi_{\text{own}}$ of one's κ-budget aimed at one's own $\rho^*_B$; another's goal $=$ the social pull toward the other holon's live pattern $\Gamma_A$; becoming part $=$ the cosine of one's chord deviation with the other's ideal overtaking the cosine with one's own. Fairness by construction: the total κ-budget is identical in every condition; the whole curve prints. Measured (1200 ticks, fixed seeds): at $\varphi_{\text{own}} = 1.0$ — own-overlap $0.817$ vs capture $0.388$; at $0.5$ — $0.634$ vs $0.581$ (still holding); at $0.25$ — $0.412$ vs $\mathbf{0.613}$ (captured); at $0.0$ — $0.168$ vs $0.575$ (captured); with no goal at all — $P \to 0.144$, the grey dissolution (H23). So the maxim survives contact with the machine with a threshold attached: the crossover sits between $\varphi_{\text{own}} = 0.5$ and $0.25$ — roughly, keeping your chord costs at least a third to a half of your regeneration budget aimed at your own ideal; and serving a borrowed goal also runs measurably less viable ($P_B$ drops from $0.286$ to $\sim 0.22$). The mapping is [О/И] and printed; every number is measured, and the interpretation is emitted by the machine's own grammar, not narrated afterward.

6. The estimator through noise (demo estimator): a hidden holon carries process noise the filter cannot predict (a random dial each step — life jiggles the pattern) and is read through noisy observations; gains are tuned on seed-A trajectories, all results are fresh seed-B; two baselines run on the identical data. Measured: with full-matrix readout the filter beats both baselines at every noise level (at $\sigma_{\text{obs}} = 0.005$: filter $0.022$ vs observation-only $0.048$ vs dead-reckoning $0.080$). With magnitude-only partial readout (the questionnaire regime), full 21-cell coverage sharpens the observables ($|\hat P - P|$ down to $0.0016$, $|\hat\Phi - \Phi|$ to $0.019$) while the full chord floors at $\sim 0.08$ — phases are structurally unobservable from magnitudes, an honest limit of every scalar questionnaire, now measured. End-to-end: blind axis diagnosis through the estimator (the truth heals after damage — regeneration re-knits it — so diagnosis reads the peak line-drop in the acute 40-step window): 62% through the filter vs 22% on raw snapshots of the same data vs 14.3% chance — the filter triples the diagnostic yield.

7. The domain sensor on a real external system (sense-git --repo P --window N): the first sensor reads a git repository — 276 real commits of the holon repo itself, 13 windows of 20. Honesty by construction, printed with every run: the feature dictionary is fixed and versioned ([О/И]; A $=$ novelty intake, S $=$ held form, D $=$ activity, L $=$ first-pass coherence $= 1-$re-touch share, E $=$ the inner workshop share, O $=$ supply rhythm, U $=$ integration span) and consumes commit metadata only — messages are never sensed and are kept aside as the held-out human label; external dynamics are unknown, so the filter runs the stated weaker EWMA prior; populations-only sensing leaves $\sigma_O/\sigma_U$ degenerate (they need coherences), and the panel is read over the informed components. Two results on the real repo: the sensed trajectory reproduces known history the sensor was never told about — the April–July project pause appears as the O-rhythm drop of window 8, and the E-workshop share wakes from $0.00$ to $0.03$ exactly in the architecture/ era; and the pre-registered held-out test passes: fix-heavy windows (labeled from messages) show higher metadata-sensed $\sigma_L$, $+0.960$ vs $+0.899$, permutation $p = 0.014$ — the sensed signal agrees with an independent human label it never saw.

What is not demonstrable today, stated as sharply: any claim about an external system the machine has not encoded — the estimator organ [K] exists and is validated on known ground truth (demo 6); the first domain sensor exists and passes a pre-registered held-out test on a real repository (item 7). What remains open — and is said so — is coherence-bearing sensing (co-change correlations → $\gamma$ magnitudes, unlocking $\sigma_O/\sigma_U$), more sensor instances (LLM sessions, telemetry), and the Delphi consensus of many encoders.

§4. Golden paths

4.1 Definitions [D]

A path is a trajectory of $\Gamma$ under the tick plus controls. Controls come in two honest kinds: dials — SO(7) rotations, which preserve the spectrum (hence $P, S, D_{\text{diff}}$) exactly and change only the wiring of coherences (H29); and bold moves — temporary increases of the regeneration gain $\kappa$ (burning extra negentropy) or a temporarily purer ideal $\rho^*$, which do change the spectrum and are paid for in measured stress. A goal region is a subset of observable space, canonically «viable and $C \geq C^*$». A path is golden if it reaches the goal region within the horizon while remaining viable.

4.2 The drift does the road [Т-anchored]

The single most consequential fact for a solver: the largest part of every golden path is free. From fog ($P = 0.220$), pure drift reaches the conscious window in 131 ticks with zero control effort — the T-124 attractor does the road. Boldness buys time: a boosted start reaches it in 33 ticks (75% faster), running measurably hotter while it lasts. And what the drift can never do is choose the chord: at fixed viability, raising $C$ from 0.340 to 0.426 required six dials — the spectrum-preserving choices that are the operational content of freedom. The division of labor is exact and machine-printed: the field removes the obstacles on the way to being; which chord to be, it leaves to you.

This is the honest formalization of the question that motivated this document — why does the universe remove obstacles before the bold? In UHM the experience decomposes without residue [И over Т]:

1. The viable window is an attractor [Т]: within its basin, the vector field works for any goal aligned with viability. Obstacles «dissolve» past the basin boundary because from there the drift finishes the road.
2. A bold move is a basin crossing: a stress surge, honestly priced (the machine prints the heat), that relocates the state to where item 1 applies. Timid paths that never cross pay less per step and fight the field forever.
3. The solitaire measures it (§6.1): from the same fog, timid decks came together 0% of the time within the horizon; decks allowed 25% bold moves — 25%, at slightly higher mean heat. Fortune favours the bold because basins do.
4. And misalignment is punished the same way: a goal outside the viable cone means permanent $\sigma$-payment and eventual halt at the grey wall — the machine's starvation run (H23) is the minimal model of it.

The pair space sharpens the social half of this. In $\mathcal{D}(\mathbb{C}^{49})$ the machine now tests cooperation directly: an aligned cross-bridge obeys the exact purity law $\Delta P = \text{linear} + 2\varepsilon^2|\gamma_{\text{cross}}|^2$ (worst residual $4\cdot 10^{-17}$ over 60 random pairs, H47), never subtracts (H48) — and, the measured jewel: both reduced states are unchanged to $10^{-17}$ (H49). The gain of connection is stored in the bond itself, in neither member — the exact formal content of the ontology book's line «the increments were never kept inside the one who left». Rivalry, meanwhile, is not the bridge's negative (it has none): it is a supply contest — under a fixed κ-budget, an 80/20 split moves the pair from $(0.338, 0.338)$ to $(0.372, 0.242)$ (H50). Contest redistributes; the bridge adds. Ethics, stated in machine units.

4.3 Golden paths are geodesics [С/И, partially measured]

The corpus already knows what an optimal aligned path is: learning follows $m$-geodesics of the BKM information metric (T-263) — locally straightest lines of the same metric that gravity extremizes one level up (T-264). The machine now measures two shadows of this. First, an identity: the regeneration term is m-geodesic motion — under pure regeneration the state rides the mixture chord $\{(1-s)\Gamma_0 + s\rho^*\}$ to a distance of $3.7\cdot 10^{-16}$ (H52); the tick was a geodesic machine all along. Second, a measurement: the full drift road from fog to the window has length/chord $= 1.165$ (H53) — the field's road is nearly straight, and the $16\%$ excess is the unitary swirl around the chord. The v1 test is now done, and the conjecture receives its verdict — operationally confirmed [Г→С-measured]: (i) the true Kubo–Mori metric is implemented and validated (it reduces to the classical Fisher metric on commuting states to $10^{-19}$, H60); (ii) the m-chord is the BKM-shortest of the tested families — $L_m = 1.1435$ vs the e-geodesic $1.1479$ vs the drift road $1.2613$ (H61–H62) — and once inside the window it stays viable to the target (H63); (iii) a geodesic MPC that greedily descends the BKM distance to $\rho^*$ reaches the window in 33 ticks, reproducing the optimal bold strategy without being told about boldness (H64): the bold move is what the geodesic controller chooses — fortune favours the bold because the bold follow the geodesic. The global variational statement remains open as mathematics, honestly; every measurable consequence tested so far confirms it.

4.4 The prophecy boundary [Т-anchored]

What may a Prime Radiant predict? The machine measures the answer. An ensemble of 500 holons, each with its own identity (a personal $\rho^*_i$): after 300 ticks the spread of WHERE (purity) compresses from $0.173$ to $0.0154$ — everyone reaches the window, the mean trajectory is sharply predictable. The spread of WHO (the chord itself) stays order-one: mean pairwise distance $0.488$ — identities do not converge (H44). Psychohistory is real for ensembles and levels; for persons there is only navigation. The freedom kernel $\dim\ker H_\Gamma + 1 > 1$ [T] is what forbids the point oracle — not a deficiency of the machine but the theorem that keeps it on the right side of the door.

§5. Categorical calibration

The method of this whole program, stated as a discipline: a symbolic system is exactly as good as its ability to describe the computational model of reality. The Γ-canon gives the symbols; the machine gives the model; calibration is the re-runnable table that says, symbol by symbol, whether the number agrees. Three verdicts exist and all three are valuable: VERIFIED (the symbol earned its number), REFUTED (the naive symbolic reading fails — the most informative outcome), UNTESTABLE-in-v0 (the symbol names a structure the current machine cannot express — a roadmap item, not an excuse).

The v1 table: 66 hypotheses → 65 VERIFIED, 1 REFUTED (H46, the discovery), 0 UNTESTABLE. The strata:

Stratum	Hypotheses	Result
Arithmetic (QR, Fano, Hamming, dictionary)	H01–H10, H46	10 verified; H46 refuted by design — it documents the TALOS $U/O$ naming catch
Analytic (contour engine)	H11–H15	all verified to $10^{-15}$
Dynamics (attractor, entropy law, halting/reignition, metabolic rate, CPTP, $G_2$)	H16–H32	verified, incl. T-124 (6/6 random starts), T-271 signs at the attractor, T-273/T-276 frequency-independence (rate constant to 0.4% across two decades of $dt$), starvation-halt with hysteresis-free reignition, $\dim\mathfrak{g}_2 = 14$
Observables/symbols	H33–H38	the grey wall hits $\max\sigma = 1.000$ exactly; the $\Phi$ proxy deviation is real; $\mathrm{Coh}_E$–$\kappa_0$ correlation $+0.97$; the naive boolean $\sigma$-equivalence fails pre-calibration (22/30 mismatches, $\sigma_L$ dominates) — an honest open calibration (H37) rather than a decorated claim
Navigation	H39–H44	boldness shortens time-to-window 131→33; bolder decks win 0%→22–25%; dials raise $C$ beyond drift; an excursion above $3/7$ under a temporarily purer ideal returns cleanly (peak $0.585$ → settled $0.321$); determinism to the bit
Pairs (v0.5)	H24, H47–H50	the T-77 purity law exact to $10^{-17}$; aligned bridges never subtract; the gain lives in the bond (reduced states untouched); the κ-contest is a real trade-off
Towers (v0.5)	H45, H51, H56	the ladder $9/14 < 1 < 54/35$ [T] — the ceiling of three; a shared ideal makes a viable meta-holon (centered-Gram $P = 1.000$), personal ideals do not ($0.189 \le 2/7$); level-3 towers await v1
Geodesics (v0.5)	H52–H53	regeneration rides the m-chord to $10^{-16}$; the drift road is $1.165\times$ its chord
Atlas (v0.5)	H54–H55	the window is a proper band of the (dissipation, supply) plane (11/20 cells); the $\kappa = 0$ column is all grey — no supply, no being
Panel errata (v1)	H57–H59	the published T-92 table failed both implications (explicit counterexample: all $\sigma < 1$ at $P = 0.153$); the renormalized $\sigma_E, \sigma_U$ encode their thresholds exactly and the embedding is restored via $\sum_i \gamma_{ii}^2 \geq 1/7$ — 0/19,000 violations; the errata is applied corpus-wide (T-92 table, the viability proof, the registry: 7 files × 2 locales)
BKM floor (v1)	H60–H64	the metric validated ($10^{-19}$ vs classical Fisher); $L_m < L_e < L_{\text{drift}}$; the m-chord viable once in-window; geodesic MPC rediscovers boldness (33 ticks)
Towers-3 (v1)	H65	level 3 works on collective states ($P_3$ aligned $1.000$ vs distinct $0.561$); the first attempt on member-index Grams was ideal-blind — pass states up, not agreement matrices
Chord search (v1)	H66	rollout-guided search matches greedy at equal budget ($C = 0.4164$ both)

Two lessons the table already taught, worth the whole exercise: the H46 catch (two corpus artifacts wired the same plane under different names — exactly the class of silent error calibration exists to catch) and the H38 honesty, since closed as a corpus errata (what began as «the boolean σ-equivalence fails pre-calibration» ended with H57–H59: the published T-92 rows for $\sigma_E, \sigma_U$ renormalized, the viability-embedding proof repaired with the $\sum_i \gamma_{ii}^2 \geq 1/7$ lemma, $\Gamma_S$ canonized — the panel's ranking use was never wrong, and its boolean use is now exact). Dozens of hypotheses is not a ceiling but a cadence: every future organ lands with its stratum of the table, and the table re-runs on every change.

§6. The Dee parable and the Prime Radiant

6.1 The solitaire as an oracle [I]

Legend has it that John Dee — mathematician, navigator's advisor, keeper of the largest library in England — taught a queen to lay out patience, and that the game was played as a question: if it comes together, the path exists. The historical record supports no such episode (Dee advised Elizabeth I; the Medici queen and the patience oracle are apocrypha), so the corpus takes it for what it is — a parable with exact content [I]. Laying out the deck many times and counting how often it resolves is Monte-Carlo estimation of a reachability probability. That is precisely the machine's solitaire organ: $M$ random control decks over the horizon; $p_{\text{golden}}$ = the measured fraction that comes together; the deal decides nothing — it measures the basin. From the same starting fog: timid decks 0%, bold decks 25%, at a printed heat premium. Four centuries later the cards are density matrices, but the question — and the honesty of asking probability instead of certainty — is the same.

6.2 What survives of Asimov's device [Т/Г/И]

In Foundation, Hari Seldon's Prime Radiant holds psychohistory: the statistical prediction of civilizational trajectories, valid only for vast ensembles, blind to individuals, broken by one anomalous mind (the Mule), and kept secret lest knowledge of the prediction destroy it. Run against the theory, the fantasy splits cleanly:

• Survives, now with proofs: prediction is ensemble-and-level-bound (§4.4 measured it: WHERE compresses 11×, WHO does not); the «Mule» is not an anomaly but a theorem — the freedom kernel and the excursion mechanism guarantee individuals who leave the mean trajectory; crises are basin boundaries, and «Seldon crises» are saddle crossings where small pushes decide much.
• Repaired: secrecy. Psychohistory feared reflexivity; UHM builds on it — the self-model is inside the state ($\rho^* = \varphi(\Gamma)$), so the prediction knowing about itself is the normal regime of the theory, not its failure mode. A public registry with statuses is the anti-Seldon: the theory that numbers its postulates survives being known.
• Dropped: the point oracle over persons, for the reasons of §4.4 — and this is the moral line: a device that predicted persons would be a cage; a device that prices paths and opens doors is a navigator.

That the idea keeps being grasped — Seldon's Radiant, the Golden Path of Dune (a viability corridor bought at terrible $\sigma$-cost by one who sees the basins), the Tao's 無為 (act with the drift, and the ten thousand obstacles remove themselves), «fortune favours the bold» — is the fractality of §1 seen in culture: one structure, many instantiations, most not yet named. The solver names them.

6.3 Solve et coagula: the alchemical layer [I]

The oldest of the captures deserves its own subsection, because the correspondence is not decorative — it is organ-by-organ. The alchemists' twin imperative solve et coagula — dissolve and bind — names the two dissipative terms of the canonical tick literally: the dissipator $g_D(I/7 - \Gamma)$ dissolves every structure toward the prima materia (the grey state, their nigredo — and the machine confirms the dissolver alone ends exactly there, H18); the regeneration $\kappa(\rho^* - \Gamma)$ binds the dissolved material toward the self-ideal (coagulatio, the whitening — and it rides the m-chord, H52). The Great Work — nigredo → albedo → rubedo — is the golden path: grey → the window → the chosen chord. Transmutation is the dials: SO(7) rotations change the wiring while preserving the substance exactly (spectrum, purity, entropy invariant, H29) — lead into gold without creating or destroying a gram, which is what the honest alchemists always said they meant. As above, so below — the Emerald Tablet's axiom — is the holarchic Enc of §1, now with a measured floor: the same Γ-form at member level and at ecology level (the Gram meta-holon, H51). And the Stone itself: the tradition insisted lapis noster non est lapis — «our stone is not a stone». The machine agrees. What the Work produces is not a substance but an instrument — the emulator plus its calibration table, the thing that turns any state's lead into its gold by finding the path: the reading of the dream this program takes as its own, marked [I] and earned line by line in §5.

§7. The blueprint: the full spectrum of stalking, planning and forecasting systems

If the solver is real, humanity has been building fragments of it for three thousand years. So the strongest available test of the architecture (as opposed to the physics) is comparative: lay out the maximal spectrum of navigation systems — divinatory, strategic, cybernetic, statistical — and see whether their load-bearing parts converge. They do, and the convergence is the blueprint. Each row states the system's state space, its oracle/engine, the invariant it got permanently right, and what the Radiant absorbs from it; the reading of every row is [I] over the machine's [Т/О] organs.

System	State space	Oracle / engine	The invariant it got right	What the Radiant absorbs
I Ching (c. −1000)	64 hexagrams = 6-bit situations	stochastic cast → hexagram + moving lines → second hexagram	the reading is of the situation-dynamics, not the person; change has a typology	discrete shadow of state space; transition reading = basin-to-basin move (cf. the 16 archetypes)
Tarot / arcana	21+1 major arcana	spread = structured sampling	a small complete alphabet of situations, laid in relations	the Canon's 21 coherences + observer (П5 atlas)
Astrology	ephemeris = deterministic celestial clock	transits = scheduled perturbations of a natal state	time-indexed readout against a shared clock; the natal chart as an Enc of initial conditions	the autoephemeris П3: the holon's own trajectory as its ephemeris
Dee's patience (§6.1)	deck = path space	deal-until-it-comes-together	reachability asked as a probability	the solitaire organ, verbatim
Castaneda's stalking	one's own routines	systematic self-observation; controlled folly; impeccability	change the wiring of habits by small controlled moves; leaks of intent are the enemy	dials as routine-rewiring; impeccability = $\sigma$-hygiene; П8 stalking the bridges
Sun Tzu	terrain + dispositions	win before battle	choose the basin where the field fights for you («water flows downhill»)	§4.2 in one sentence, 25 centuries early
OODA loop (Boyd)	adversarial engagement	Observe–Orient–Decide–Act, faster than the environment	tempo: the loop's rate is itself a weapon; Orient = the encoder is the seat of victory	the tick cadence; Enc as the Orient organ
Scenario planning (Shell)	branching futures	ensemble narratives, no point forecast	institutionalized refusal of point prophecy	the §4.4 boundary as method, not modesty
Delphi method	expert beliefs	iterated anonymous consensus	many encoders beat one	multi-encoder $\hat\Gamma$ (v1)
Superforecasting (Tetlock)	probabilistic claims	Brier-scored calibration culture	calibration as a discipline: keep score or drift into myth	the [F] organ is a Brier board for symbols
Kalman filtering	hidden state + noisy readings	predict–measure–update	optimal estimation under noise	the missing organ this survey exposed: v0.5 has no estimator $\hat\Gamma(\text{data})$ — promoted to the v1 roadmap
MPC (control)	plant + horizon	receding-horizon optimization	plan far, commit one step, re-plan	navigate(), verbatim
VSM (Beer)	five recursive systems	variety engineering	viability as the criterion and recursion of the same form at every level	CC's direct ancestor; the [H] meta-holon organ
Active inference (Friston)	generative self-model	minimize expected free energy	perception and action are one functional; the self-model is load-bearing	$\rho^* = \varphi(\Gamma)$; $\sigma$ as the free-energy dial (cf. the Vanchurin bridge, T-258..262)
RL / MCTS (AlphaZero)	game states	tree search × learned value	search plus prior beats either alone	solitaire = flat MC today; MCTS-over-dials with a learned value = v1
TRIZ (Altshuller)	contradiction matrix	40 inventive principles; the ideal final result	resolve contradictions by rewiring, not compromise; the best system is one where «the function performs itself»	dials = rewiring without substance change; the ideal final result = maximize the drift's share of the road (measured: the field does most of it)

The synthesis. Strip each system to its load-bearing part and the same eight organs appear, in every mature tradition, under different robes:

1. The Encoder (natal chart, Orient, generative model) — read the situation into a state. Machine: Enc / [C].
2. The Estimator (Kalman, Delphi) — infer the state from noisy data, repeatedly. Machine: [K] shipped in v0.6; the first domain sensor [S] shipped in v0.7 (git, held-out p = 0.014); coherence-bearing sensing remains open.
3. The Panel (symptoms, base rates, σ) — addressed diagnosis. Machine: [C], Σ-decode.
4. The Atlas (scenarios, terrain maps, ephemerides) — the map of basins and schedules. Machine: [J].
5. The Oracle (patience, wargame, MCTS) — sampled reachability, answered as probability. Machine: solitaire.
6. The Navigator (OODA, MPC, stalking) — commit one step, re-plan, keep tempo. Machine: [E].
7. The Calibrator (Tetlock, backtesting) — the scoreboard that keeps the symbols honest. Machine: [F].
8. The Boundary (Shell's refusal, the Mule, free will) — the constitutional limit on point prophecy over persons. Machine: §4.4, enforced by theorem.

That convergent evolution found the same chart eight-for-eight is the strongest architectural evidence available: the blueprint is imposed by the structure of the task, not by any culture [I]. What the divinatory family got permanently right: state discretization, scheduled perturbations, stochastic reachability, and the practitioner's calibration diary. What it got permanently wrong — and what the Radiant refuses by theorem rather than by taste — is the point oracle over persons. UHM's contribution to the lineage is not a ninth organ but the floor under all eight: a state space with theorems, thresholds with proofs, and a calibration table that re-runs.

The same comparative move applied to the structural lineage — UML→C4, DDD, VSM, MAPE-K, the 2024–2026 agentic stack — yields the sister blueprint: HOLARCH, the meta-specification of architecture. The Radiant's eight organs are the navigation suite of a system whose standing structure HOLARCH types; its estimator gap (row «Kalman» above) and viability-engineering case №12 both land there as first-class machinery.

§8. Application cases

Extraction criteria: a live data loop (can $\Gamma$ be estimated repeatedly?), falsifiable feedback (does the prescription's outcome measure back?), cost of error, and payoff. Ranked:

#	Case	Horizon	Observables	Status
1	SYNARC introspection & safety — the σ-panel as the organism's own interoception; off-switch/halting per T-288/T-289; navigator as the deliberation organ over $\rho^*_\Omega$	now (spec'd)	full suite, exact	[D] engineering
2	LLM-agent context health — $\sigma_O$ (resource starvation) predicts context collapse before it happens; prescriptions = summarize/replenish (CC worked case)	now	$\sigma$-proxies from logs	[C]
3	Clinical consciousness monitoring — the PCI* index already grazes $2/7$ (0.31 vs 0.286); anesthesia emergence kink $\beta = 1/4$; the panel as an OR instrument	2–3 yr	$P$-proxy, $\Phi$-proxy	[C] falsifiable
4	Organizational & therapeutic audit — the П1 mapping + the panel + flow-restoration prescriptions; burnout reads as $O/U$ loss (the machine's demo)	now	28-cell audit	[О/И] protocolized
5	Education — the ontology book's map + a tutor-navigator that shows a learner their open doors (dials) instead of their rank	2–3 yr	desk/library model	[I]
6	Personal navigation (the Dee protocol) — daily Γ-mapping, solitaire runs over one's own basins; honest statuses on every readout	now, personal	П1–П4	[I] explicitly
7	Market/systemic risk — $\mathrm{SysRisk} = \max_k\|\sigma_k\|/P$ over ensemble Γ; psychohistory where it is legitimate (levels, not persons)	2–10 yr	index proxies	[H]
8	Research prioritization — the solver pointed at the theory itself: which open symbol, if calibrated, most enlarges the verified table (the calibration gradient)	now, meta	the H-table	[D]
9	Policy ensembles — basin maps for populations under interventions, with the §4.4 boundary as a constitutional constraint: navigate levels, never point-predict persons	10 yr	ensemble Γ	[H]
10	Merger & alliance design — the pair organ as an instrument: size the bridge (gain $2\varepsilon^2\|\gamma\|^2$, stored in the bond), separate it from the κ-budget contest, predict which splits starve whom	2–3 yr	pair Γ, H47–H50	[О/С]
11	Team alignment audit — the centered-Gram meta-P as a one-number answer to «is this group a subject?»: shared ideal 1.000 vs personal ideals 0.189 against the 2/7 wall	now	member Γ's	[О/И]
12	Viability engineering — the phase atlas as a design chart: place any system (reactor, org, model training run) inside the Goldilocks band; the $\kappa = 0$ column as the hard law no supply, no being	now	(gD, κ) proxies	[D]

The best cases (1–3) share one shape: the observables already exist, the feedback loop is measured in the same units as the prescription, and an error surfaces quickly. The 10-year ecosystem the author expects is, concretely, rows 4–9 growing the same shape.

§9. Roadmap to v1

1. The pair space — done in v0.5 (organ [G]; H24, H47–H50 closed).
2. Towers — done: level 2 in v0.5 (organ [H]; H45, H51); level 3 in v1 on collective states (H65, closing H56) — with the structural lesson that member-index Grams are ideal-blind one level up.
3. The Estimator organ [K] — shipped in v0.6 (demo 6, R26–R27): predict–measure–update over $\mathcal{D}(\mathbb{C}^7)$ with process noise, full and magnitude-only partial readouts; beats both baselines; end-to-end diagnosis through noise 62% vs 22% raw. The first domain sensor shipped in v0.7 (: a real repository, metadata-only dictionary, pre-registered held-out validation $p = 0.014$). Remaining: coherence-bearing sensing and multi-encoder Delphi consensus.
4. Panel calibration — done, and it became a corpus errata (H57–H59): the published T-92 rows for $\sigma_E, \sigma_U$ did not encode their thresholds and the embedding failed on an explicit counterexample; renormalized rows and the repaired viability proof applied across 7 corpus files × 2 locales; $\Gamma_S$ canonized as the $\{A,S,D\}$ sector block (closes H37).
5. Geodesic MPC — done in v1 (organ [L]; H60–H64): §4.3's [H] operationally confirmed; the MPC rediscovers the bold strategy from the metric alone.
6. MCTS over dials — done in v1 (H66): rollout-guided search matches greedy at equal budget; a learned value function stays optional until a scale demands it.
7. The Verum port — done at spec level: typed contracts carrying the calibrated invariants, with @test/@property blocks mirroring the H-numbers.
8. Calibration cadence: every organ lands with its hypothesis stratum; the table is the release gate.

§10. Status summary

Claim	Status
Fano wiring = translates of $QR(7)$; $\mathbb{Z}_3$ = multiplier 2; Hamming address; XOR↔QR dictionary	[T] machine-checked
Contour-residue spectral engine correct to $10^{-15}$	[T] standard math, machine-checked
Viable attractor; entropy-law signs; starvation-halt + hysteresis-free reignition; metabolic rate frequency-independent	[T] corpus theorems, reproduced
The drift does most of every golden path; boldness shortens time-to-window at printed heat cost; dials (not drift) choose the chord	measured in v0
T-77 in the pair space: exact purity law; the gain lives in the bond (reduced states untouched); rivalry = supply contest	measured in v0.5
The ceiling of three: ladder arithmetic [T]; shared ideal ⇒ viable meta-holon, personal ideals ⇒ none	[T] + measured
Regeneration = m-chord motion ($10^{-16}$); drift road $1.165\times$ chord	measured (T-263 shadow)
The window is a proper Goldilocks band of (dissipation, supply); $\kappa = 0$ ⇒ all grey	measured atlas
The eight-organ blueprint is convergent across the full spectrum of navigation systems	[I] comparative
Solve et coagula = dissipator + regeneration; transmutation = spectrum-preserving rewiring; the Stone is an instrument	[I] over [T] organs
«Obstacles removed before the bold» = basin crossing + attractor drift	[I] over [T] anchors
Golden path = viable BKM geodesic	[H] (v1 test)
Ensembles predictable, persons navigable-only (freedom kernel)	[T]-anchored, measured
σ-panel boolean use	closed by the 2026-07-22 corpus errata: renormalized $\sigma_E, \sigma_U$, embedding restored with proof, $\Gamma_S$ canonized (H37, H57–H59)
Golden path = viable BKM geodesic	[Г → С-measured]: $L_m < L_e < L_{\text{drift}}$, chord viable in-window, MPC rediscovers boldness (H60–H64); the global variational statement open
The solitaire oracle, the Dee parable	[I] with exact Monte-Carlo content
Psychohistory for levels, never persons	the §4.4 boundary

The table of §5 is the living contract of this program: 56 hypotheses today; every change re-earns it.