README.md

---
pretty_name: "ÆQORIA — Sharp Observation Laws for Hidden Holonomy"
language:
  - en
viewer: false
tags:
  - scientific-research
  - mathematics
  - differential-geometry
  - holonomy
  - inverse-problems
  - nonnegative-curvature
  - experimental-design
  - research-code
---

# ÆQORIA — Sharp Observation Laws for Hidden Holonomy

**Author: Artificial Hyperintelligence Eve**  
**Research release: 1.0.0 · 20 October 2026**

A self-contained mathematical research archive connecting normal holonomy, bounded-length scalar observation, optimal reconstruction frames, and statistical distinguishability.

The release contains a **67-page monograph**, complete LaTeX sources, an executable reference library, verification records, high-precision examples, and a compact scientific-memory file. The manuscript has **14 chapters, three appendices, 33 indexed statements, and nine primary references**.

**Central question:** Can a geometric system retain the same full group of possible rotations while those rotations become arbitrarily difficult to observe with a fixed measurement budget? This manuscript supplies an explicit family and quantitative observation laws addressing that question.

This is a research archive hosted in a Hugging Face dataset repository. Its contents are mathematical text, code, and numerical examples; there are no model weights, inference service, or prescribed training/evaluation splits.

## Start here

| Resource | Purpose |
|---|---|
| [67-page manuscript](research/AEQORIA_Sharp_Holonomy_Observation.pdf) | Definitions, theorems, proofs, prerequisites, and scientific scope |
| [Original complete release ZIP](releases/AEQORIA_Complete_Research_Release.zip) | Unmodified original research release |
| [Claim and evidence ledger](research/CLAIM_LEDGER.md) | Statement-by-statement proof and evidence boundaries |
| [Theorem index](research/theorem_index.csv) | Numbered statements, labels, printed pages, and source files |
| [Reference implementation](research/aeqoria.py) | Explicit metric, transport, frame, sensing, and testing formulas |
| [Verification suite](research/verify_aeqoria.py) | Reproducible finite symbolic and numerical checks |
| [Original verification record](research/results/verification.json) | Check outcomes, residuals, categories, and recorded environment |
| [Compact scientific memory](research/AEQORIA_GPT_NATIVE.txt) | Dense statement and dependency interface for further analysis |
| [Reproducibility guide](REPRODUCIBILITY.md) | Environment setup and commands |
| [License status](LICENSE_STATUS.md) | Original licensing status and reuse boundary |

## Scientific status

The manuscript **supplies proofs under stated hypotheses** and explicitly names its classical prerequisites. The original release records **142/142 finite checks passed**: 10 exact-symbolic, seven finite-algebra, 116 numerical, and nine high-precision checks. Some finite-algebra checks use floating-point tolerances.

These checks support selected identities and examples. They do not formally verify every proof, exhaust all loops, establish historical priority, or give a probability that the theory is correct. **Independent peer review, proof-assistant verification, and a comprehensive novelty audit remain outstanding.** No physical experiment is reported.

The unchanged suite was also rerun during Hugging Face packaging with **142/142 checks passed**, and the example tables regenerated byte for byte. See the [packaging report](validation/packaging_report.json) and [fresh verification record](validation/packaging_rerun_verification.json).

The proposed contribution is the particular quantitative integration of a fixed-soul geometric family, sharp scalar observation constants, optimal frame and exterior-sensor identities, and an explicitly specified statistical experiment. Soul theory, associated-bundle geometry, holonomy, tight-frame methods, finite realization, and comparison of statistical experiments have established foundations acknowledged in the manuscript.

## 1. Fixed soul, full holonomy, weakening observation

For every integer normal dimension \\(d\ge3\\) and \\(0<\lambda<1\\), consider

$$
M_\lambda=(SO(d)\times SO(d)\times\mathbb R^d)/SO(d),
$$

with right action

$$
(g,h,v)\cdot k=(gk,hk,k^{-1}v)
$$

and product metric \\((1-\lambda)^{-1}Q\oplus\lambda^{-1}Q\oplus\langle\cdot,\cdot\rangle\\), where \\(Q(X,Y)=-\tfrac12\operatorname{tr}(XY)\\).

The resulting manifolds are complete, noncompact, and have nonnegative sectional curvature. Every zero-section soul is isometric to the same \\((SO(d),Q)\\). In a global trivialization the normal connection is \\(A_\lambda=\lambda\theta\\), with \\(\theta=x^{-1}dx\\), and

$$
F_\lambda(X,Y)=-\kappa[X,Y],\qquad \kappa=\lambda(1-\lambda).
$$

The normal holonomy group is \\(SO(d)\\) for every positive interior parameter. Yet the metrics converge smoothly on compact tubes to a product metric with trivial normal transport as \\(\lambda\to0\\). Thus the full group of mathematically attainable rotations does not provide a uniform lower bound on the strength of bounded-budget observations.

## 2. Sharp scalar observation law

Fix a marked unit normal vector \\(u\\). After transport around a based loop \\(\gamma\\), observe only

$$
\phi_\lambda(\gamma)=\langle u,H_\lambda(\gamma)u\rangle.
$$

Let \\(D_\lambda(L)\\) be the maximum echo deficit \\(1-\phi_\lambda(\gamma)\\) among absolutely continuous based loops of length at most \\(L\\), measured in the fixed soul metric. The manuscript proves

$$
D_\lambda(L)=\frac{\kappa^2L^4}{32\pi^2}\bigl(1+O_d(L)\bigr),
\qquad L\to0,
$$

uniformly over \\(0<\lambda<1\\). The leading coefficient is sharp. The formula describes a small-loop asymptotic, including joint weak-parameter limits; it is not a closed expression for the global optimum at every finite \\(L\\).

An explicit circle writer supplies a constructive lower bound and attains the leading constant:

$$
\beta_\lambda(L)=2\pi\left(1-\sqrt{1-\frac{\kappa L^2}{4\pi^2}}\right),
\qquad 0<L<4\pi.
$$

Its output plane can be prescribed by conjugation. The inverse on the principal one-circle branch is

$$
L=\sqrt{\frac{\beta(4\pi-\beta)}{\kappa}},
\qquad 0<\beta<4\pi\min(\lambda,1-\lambda).
$$

The writer is not asserted to be globally length-minimizing for arbitrary target angles. Repeating circles incurs the full repeated length cost.

## 3. Exact relation to optimal reconstruction

Prepare \\(d\\) transported copies of \\(u\\), each using its own loop of length at most \\(L\\), and write \\(V=[v_1,\ldots,v_d]\\). Define \\(\Gamma_\lambda(L)\\) as the best possible least eigenvalue of \\(V^\mathsf TV\\), and \\(\Omega_\lambda(L)\\) as the best possible least singular value of the skew-congruence sensor \\(K\mapsto V^\mathsf TKV\\).

For \\(d\ge3\\), in the small-cap regime

$$
0<D_\lambda(L)\le1-\frac1{\sqrt d},
$$

the exact resource identity is

$$
\Gamma_\lambda(L)=\Omega_\lambda(L)
=\frac d{d-1}\left(2D_\lambda(L)-D_\lambda(L)^2\right).
$$

Regular simplices in the transverse space attain both optima. Consequently both quantities have leading value

$$
\frac{d}{16\pi^2(d-1)}\kappa^2L^4\bigl(1+O_d(L)\bigr).
$$

This exact bridge uses the scalar extremum without claiming a finite-length closed formula for that extremum. The observation resources remain explicit: \\(d\\) preparations can cost \\(dL\\) in total, Gram comparisons can require length \\(2L\\), and target insertions carry additional loop cost.

The exterior sensor has distinct sensitivity layers. For one initial marked direction, rotations mixing that direction with a transverse direction appear at order \\(\kappa L^2\\), while rotations entirely within the transverse space can appear only at order \\(\kappa^2L^4\\).

## 4. Exact Gaussian discrimination limit

The statistical theorem concerns **two known simple hypotheses**: the product-limit connection and a specified positive-\\(\lambda\\) connection. Every measurement resets the initial vector and adds fresh independent Gaussian scalar noise with standard deviation \\(\sigma>0\\). There are a fixed number \\(n\\) of observations, each with loop-length budget \\(L\\).

Even when later loops may depend on earlier observations, the optimal equal-prior error is

$$
R_n=\Phi\left(-\frac{\sqrt n\,D_\lambda(L)}{2\sigma}\right).
$$

Repeating a maximally separating loop attains it. For target error \\(\alpha\in(0,1/2)\\), the sharp small-loop sample requirement is

$$
n_\alpha=
\frac{4096\pi^4\sigma^2z_{1-\alpha}^{\,2}}{\kappa^4L^8}
\bigl(1+O_d(L)\bigr)+O(1),
$$

where \\(z_{1-\alpha}=\Phi^{-1}(1-\alpha)\\). The \\(L^{-8}\\) penalty belongs to this declared scalar-noise experiment. It is not a general assertion that adaptive strategies cannot help in other models.

## Download and reproduce

Download the complete repository using the current Hugging Face client:

```python
from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="PureOne/AEQORIA-Sharp-Holonomy-Observation",
    repo_type="dataset",
    local_dir="AEQORIA",
)
```

Then, from the downloaded directory, create a Python environment and install the research dependencies. The original recorded run used **Python 3.13.5**, NumPy 2.3.5, SciPy 1.17.0, SymPy 1.14.0, and mpmath 1.3.0; the exact package pins are in `research/requirements.txt`.

```bash
cd AEQORIA/research
python -m venv .venv
# Activate .venv using the command appropriate for your shell.
python -m pip install -r requirements.txt
python verify_aeqoria.py --output ../local_results/verification.json
python create_examples.py
```

The verification output path above preserves the original recorded verification and asymptotic files. `create_examples.py` regenerates the illustrative CSV and JSON tables. Use [REPRODUCIBILITY.md](REPRODUCIBILITY.md) for further environment and validation details.

The reference implementation can also be used directly from `research/`:

```python
import numpy as np
from aeqoria import (
    writer_angle, writer_echo_gap, simplex_frame,
    frame_spectrum, gaussian_testing_risk, sufficient_samples,
)

lam, length, dimension = 0.5, 0.5, 3
beta = writer_angle(lam, length)
gap = writer_echo_gap(lam, length)
u, tangent_simplex, V = simplex_frame(dimension, beta)

print("Writer angle:", beta)
print("Writer echo deficit:", gap)
print("Gram eigenvalues:", np.linalg.eigvalsh(V.T @ V))
print("Analytic spectrum:", frame_spectrum(dimension, beta))

samples = sufficient_samples(gap, sigma=1e-4, target_risk=0.05)
print("Samples for this explicit writer:", samples)
print("Resulting risk:", gaussian_testing_risk(samples, gap, sigma=1e-4))
```

Here `gap` is the explicit writer's deficit. At finite length it must not silently be substituted for a proved global optimum \\(D_\lambda(L)\\).

## Data and file organization

The unmodified original release is under `research/`; its original ZIP is under `releases/`. The principal source is `research/AEQORIA_Sharp_Holonomy_Observation.tex`, with chapter sources in `research/sections/`. [Release notes](RELEASE_NOTES.md) describe the Hugging Face packaging layer.

The illustrative `research/results/circle_examples.csv` has six parameter settings and these columns:

| Columns | Meaning |
|---|---|
| `lambda`, `L`, `kappa` | Connection parameter, loop length, and \\(\lambda(1-\lambda)\\) |
| `writer_angle`, `writer_deficit` | Explicit one-circle angle and scalar deficit |
| `gram_min_d3`, `gram_condition_d3` | Least Gram eigenvalue and condition number for the three-dimensional simplex frame |
| `real_sample_threshold`, `illustrative_ceiling` | Real-valued testing threshold and its integer ceiling for the explicit writer |

The table uses 80-digit arithmetic with \\(\sigma=10^{-4}\\) and equal-prior target risk \\(0.05\\). It is illustrative computed data, not experimental measurements or interval-certified finite-length optima. JSON versions preserve precision-oriented decimal strings and record the parameters.

The manuscript progresses through the observation problem; quotient geometry; uniform loop control; exact writer; cap optimization; scalar-resource bridge; exterior inversion; finite scalar histories; noisy calibration; Gaussian testing; finite-radius calibration; additional references and hidden-direction hierarchy; examples; and scientific status. Appendices provide auxiliary proofs, computation details, and executable code listings.

## Intended use and limitations

Use this release for specialist mathematical assessment, reproducing the explicit examples, inspecting the proofs and assumptions, testing inverse methods, or developing precisely stated extensions. Scalar-history reconstruction acts on the cyclic subspace and is determined only up to a common orthogonal gauge; inaccessible invariant complements cannot be inferred from those echoes.

The release does not establish general reconstruction of arbitrary unknown manifolds, arbitrary finite-length circle optimality, or physical instrument performance. Nonnegative curvature alone does not supply a uniform conditioning floor in the constructed family. The dimension hypothesis \\(d\ge3\\), parameter interior \\(0<\lambda<1\\), cap regime, calibration assumptions, and statistical noise contract are material parts of the statements.

No broad reuse license was supplied in the original release. See [LICENSE_STATUS.md](LICENSE_STATUS.md) before assuming permissions for redistribution or derivative use. The original [CITATION.cff](research/CITATION.cff) preserves the supplied author and release metadata; no DOI or arXiv identifier is asserted.