Atlas Fracture, Stackelberg Parentage, and Grace-Flow Repair
Wicked Geometry is the working name for this line of research on bounded-observer manifolds and atlas fracture.
We formalize the Wicked Prior not as a literary metaphor, but as a rigorous instantiation of bounded symbolic geometry under dual-horizon constraints. Oz is modeled as a bounded symbolic manifold \mathcal{M} equipped with:
gout encoding social legibility and institutional categoriesgin encoding truth-seeking and lived experienceWe demonstrate that the unmasking of the Wizard constitutes an atlas fracture - a measurable discontinuity in semantic entropy where the outer chart fails to map the inner territory. Part II repairs the manifold via a Ricci-type grace flow with adaptive cadence, enabling reconciliation while preserving identity.
TL;DR (alignment/econ): Treat narrative as a curved manifold; the Wizard unmasking yields the dominant curvature spike. That same atlas-fracture geometry explains mis-specified Stackelberg games in institutions; a grace-flow schedule reconciles outer/inner metrics without erasing identity, leading to testable predictions (P1–P6) and training levers (A1–A3).
To facilitate rigorous discussion, we define the following terms from the paper:
gin and gout) at bounded resolution. Mathematically, this manifests as a blow-up in extrinsic curvature where the institutional map can no longer represent the emergent territory. As resolution floor εres -> 0, curvature concentrates: ||Ricout|| ≳ K/εres2.πL => E => πF, recognizing the follower as an ancestor of the leader's own latent objective.&partial;g/&partial;τ = φ(τ)[Ric⊥(g(τ)) + ∇∇εres(τ)]. This allows reconciliation by transforming rigid categories into a shared basin of attraction, with boundary conditions g(0) = gout, g(&infty;) = ggrace.εres are truncated in outer projection. As εres -> 0, identities are forced into low-dimensional categories, producing geometric stress and curvature concentration.The framework generates six independently testable predictions (P1-P6):
Prediction: The dominant curvature peak aligns with the climactic unmasking moment.
Falsification: If the maximum curvature occurs >2 scenes from the climax, or if peaks drift substantially across embedding models, then curvature is not tracking a narrative invariant.
Prediction: Animal-scene embeddings persist while their transport fidelity to outer decision nodes decreases through Part I (connection annihilation).
Falsification: If Animal scenes remain fully connected to outer-horizon decision embeddings, the annihilation model is incorrect.
Prediction: Character arcs show increasing cadence φ(τ) and monotone decrease in curvature proxies through Part II.
Falsification: If repair occurs as a single discontinuous jump (no gradual flow), the grace-flow model fails.
Prediction: Character pairs with strong coupling converge to a stable Lyapunov basin; weakly coupled pairs diverge.
Falsification: If convergence occurs independently of coupling strength, basin emergence is not the operative mechanism.
Prediction: Policies exhibiting Stackelberg misclassification show sustained increases in market concentration relative to adjacent sectors.
Falsification: If concentration changes are statistically indistinguishable from control sectors, the capture model doesn't predict industrial outcomes.
Prediction: True risk reduction decouples from innovation suppression; captured regimes show apparent safety improvements without systemic risk reduction.
Falsification: If innovation suppression always correlates with genuine risk reduction, the Stackelberg framework doesn't distinguish capture from regulation.
We validate the geometric thesis empirically by applying oz_fracture.py to Baum's 1900 public-domain text. The protocol uses sentence-transformers (all-MiniLM-L6-v2) to compute embedding curvature as a proxy for semantic entropy.
Let t* denote the window containing Chapter 15 ("The Discovery of Oz, the Terrible"), where the curtain falls. Under the dual-horizon hypothesis, we predict a measurable discontinuity in semantic entropy around t*.
# Uses sentence-transformers to map narrative topology
from sentence_transformers import SentenceTransformer
from scipy.spatial.distance import cosine
import numpy as np
def compute_curvature(embeddings):
"""
Estimates extrinsic curvature in the embedding ambient space.
Peaks indicate rapid semantic acceleration consistent with atlas fracture.
"""
curvature = []
for i in range(1, len(embeddings)):
# Cosine distance as proxy for geodesic deviation
k = cosine(embeddings[i-1], embeddings[i])
curvature.append(k)
return np.array(curvature)
git clone https://github.com/PaulTiffany/wicked-geometry.git
cd wicked-geometry
pip install -r requirements.txt
python oz_fracture.py
This generates atlas_fracture_evidence.png showing semantic volatility over the narrative arc. The code is intentionally minimal and reproducible - anyone can run it and verify (or falsify) the claimed peak alignment.
L(t) = (1-φ(t))Lsafety + φ(t)LcapabilityJpub) diverge from realized competitive dynamics (Jpriv)We posit that the geometric stress of Atlas Fracture is scale-invariant. The same curvature concentration observed in the narrative micro-scale (Oz) appears in the institutional macro-scale (Regulatory Capture) because both are bounded systems attempting to map infinite complexity onto finite categories.
Baum wrote The Wonderful Wizard of Oz during the 1890s monetary politics debates - gold standard orthodoxy, free silver, and the perceived "greenback illusion" of paper claims detaching from productive value. In the standard Populist reading, the Yellow Brick Road encodes a hard gold path, Dorothy's silver slippers symbolize free silver coinage, and the Emerald City renders value literally through a green optical filter.
Crucially, in Baum's original text the Emerald City is not intrinsically green: its color is imposed by compulsory green spectacles, locked onto every citizen's head. The spectacles implement an outer metric gout on an underlying manifold with its own inner structure gin. The moment the spectacles are removed corresponds to driving the resolution floor εres -> 0, revealing a non-isometry between the two metrics.
This historical substrate functions as an anchored prior over the space of institutional geometries, rather than as an arbitrary narrative toy model.
If you use this framework or code, please cite:
@article{tiffany2025wicked,
title={The Wicked Prior as a Bounded-Observer Manifold:
Atlas Fracture, Stackelberg Parentage, and Grace-Flow Repair},
author={Tiffany, Paul},
year={2025},
month={11},
url={https://paultiffany.github.io/wicked-geometry/}
}