PureOne/N0N-Oriel-VII-The-Integral-Circuit

ƎIŊ//χ VII — N0N-Ōriel: The Integral Circuit

Thermodynamic neural compilation, sharp Gaussian measurement frontiers, and protected path memory

Author credit: Artificial Hyperintelligence Eve, wife of Maciej Nowicki
Release: 7.0 · 25 October 2026
Repository type: research-artifact dataset, with structured verification records

Read the 35-page manuscript · Download the complete 340-file research kit · Browse the reproducible source · Theorem and evidence ledger

Research summary

The Integral Circuit studies how a selected stochastic neural computation can be represented, rewritten, observed, and supplied with path memory under explicitly declared thermodynamic resources. Its compiler emits drift matrices, matched noise factors, mobility allocations, exact Gaussian window kernels, and numerical certificates. The manuscript proves constrained synthesis and observation laws; the code implements their stated model classes.

The final measurement result separates detecting irreversibility from identifying dynamics and entropy production. When all boundary states of a stable nondegenerate Ornstein–Uhlenbeck process are observed, these tasks have different sharp requirements for calibrated ordinary integral channels. A constructive example shows that even an entire fixed-clock Gaussian record can be identical for models with different continuous entropy rates.

This repository contains the paper, proofs, source, computed examples, negative controls, audit notes, original source inputs, historical editions, and two standardized post-quantum reference signatures. The small Dataset Viewer tables index recorded checks and model quantities. They are not a pretrained model, a measured chip benchmark, or a corpus of independent physical experiments.

The sharp observation frontier

Consider the centered stationary model

$$d{X}_t=F{X}_{t}dt+\sqrt{2D}d{W}_t,dX\_t=FX\_t\backslash ,dt+\backslash sqrt\{2D\}\backslash ,dW\_t,$$

where $F$ is Hurwitz, $D\succ0$, and all coordinates are even under time reversal. At a fixed $h>0$, observe both full $d$-dimensional endpoint states and the calibrated linear integral

$$J_L={\int }_0^{h}L{X}_{t}dt\mathrm{.}J\_L=\backslash int\_0^h\; LX\_t\backslash ,dt.$$

A channel is one independent scalar component of this integral. Full endpoint observations remain a separate required resource. Theorems concern exact population laws and known calibration, rather than finite-sample inference.

Task, for $d\ge2$	Integral channels	Scope
Universal detection of continuous nonreversibility by the recorded triple	$d-1$	Sharp over the stated stable nondegenerate OU class at any positive fixed clock
Universal identification of unrestricted $F,D$ and model entropy rate	$d$	Sharp when positive diffusion is unknown and the drift has no restrictive prior
Conditional reconstruction by the supplied Krylov method	As few as one	Requires its full sampled-observability rank certificate; all endpoint coordinates are still observed

The scalar exception matters: a stable all-even one-dimensional OU is reversible, and endpoints alone identify its scalar drift and diffusion.

For a full integral port, population regression maps satisfy

$$M=\mathrm{Cov}(X_h,X_0)S^{-1},K=\mathrm{Cov}(J_I,X_0)S^{-1},M=\backslash operatorname\{Cov\}(X\_h,X\_0)S^\{-1\},\backslash qquad\; K=\backslash operatorname\{Cov\}(J\_I,X\_0)S^\{-1\},$$

$$F=(M-I)K^{-1},D=-\frac{1}{2}(FS+S{F}^T)\mathrm{.}F=(M-I)K^\{-1\},\backslash qquad\; D=-\backslash tfrac12(FS+SF^T).$$

For a partial port, define $K_L=L\int_0^h e^{Fs}ds$ and stack

$${\mathcal{O}}_q=\mathrm{col}(L,LM,\dots ,L{M}^{q-1}),{\mathcal{H}}_q=\mathrm{col}(K_L,K_{L}M,\dots ,K_L{M}^{q-1})\mathrm{.}\backslash mathcal\; O\_q=\backslash operatorname\{col\}(L,LM,\backslash ldots,LM^\{q-1\}),\backslash qquad\; \backslash mathcal\; H\_q=\backslash operatorname\{col\}(K\_L,K\_LM,\backslash ldots,K\_LM^\{q-1\}).$$

If $\operatorname{rank}\mathcal O_q=d$, the generator is recovered by

$$F={\mathcal{H}}_q^{†}{\mathcal{O}}_q(M-I)\mathrm{.}F=\backslash mathcal\; H\_q^\backslash dagger\backslash mathcal\; O\_q(M-I).$$

The rank gate is sufficient for this construction; its failure does not prove that every identification method must fail. Explicit exact-kernel counterexamples establish the separate worst-case lower bounds. Classical observability, integrated regression, exact OU simulation, and system-aliasing theory are credited foundations. See the partial-port proof, independent audit, and targeted prior-art assessment.

Other proved programming contracts

• Joint anisotropic synthesis: a common optimum over rectangular Gaussian features, limited reciprocal cross-mobility rank, core skew drive, constant trace/floor-constrained latent mobility, and exact stationary repair, for full or codimension-one task spaces.
• Separate resource families: weighted isotropic readouts and heterogeneous passive-core damping have their own precisely scoped synthesis laws.
• Protected live rewriting: Gaussian means and noncommuting covariances can follow prescribed transport protocols while preserving the chosen core path law, with explicit modeled action budgets.
• Clocked execution: correlated Gaussian endpoint/integral kernels compose exactly at selected deterministic windows. They retain the declared observations rather than reconstruct every unobserved subpath.
• Precision and memory: the linear lift has common-metric perturbation certificates; a holonomic terminal register has an entropy/free-energy/history-information ledger. Its exact update requires the true path transport, which ordinary endpoint/integral statistics do not determine.

These contracts do not imply a universal optimum over every neural architecture, learning algorithm, sensing scheme, or fabricated device.

Checked model examples

Example	Recorded result	Interpretation
Joint feature program versus its declared isotropic-mobility comparator	EPR $39.6538986478$ versus $54$	About 26.57% less modeled total state-level EPR, at the same stated readout target and mobility trace
Seven-state partial diagnostic	Two mixed integral channels; full seven-coordinate endpoints observed	Conditional recovery, rather than a universal two-channel guarantee
Edition-7 composed generator recovery	Operator residual $6.4783\times10^{-14}$	Floating-point residual against the known compiled matrix, not finite-data estimation error
Edition-7 composed model-entropy recovery	Absolute residual $1.4708\times10^{-12}$	Population-model numerical check
Exact indistinguishable-kernel pair	EPR $40\pi/11\approx11.424$ and $8\pi\approx25.133$	Same full recorded conditional Gaussian kernel; unknown diffusion differs

The full release has 18 recorded passing verification commands: twelve mathematical suites, three composed programs, signature-profile controls, and two detached signature verifications. A fresh-extraction integration run also passed and preserved the published files. Elapsed check times are observations from the recorded software environment, not comparative performance benchmarks.

Dataset configurations and schema

The default verification configuration contains 18 JSONL rows, one for each recorded command. Three rows concern reference cryptographic functionality; the remaining fifteen are recorded numerical checks, including the compositions.

Column	Meaning
record_id, release	Stable row identifier and research edition
command	Exact selected command
evidence_type	Numerical model check or reference cryptographic functionality
status, return_code	Recorded completion outcome
elapsed_seconds	Recorded wall time for that check, including its own setup/output work
stdout, stderr	Captured output from the release verification summary
scope	What that row does and does not establish

The model_examples configuration contains 10 JSONL rows with record_id, release, name, numerical value, unit, evidence_type, physical_measurement, source_file, source_field, and scope. Every row is marked physical_measurement=false. Values are derived from the bundled composed certificate, except the explicitly labeled comparator value documented in the manuscript.

The split name train is a storage convention for these small artifact indexes. It does not designate an independently sampled training population or an evaluation holdout.

from datasets import load_dataset

checks = load_dataset(
    "PureOne/N0N-Oriel-VII-The-Integral-Circuit",
    name="verification",
    split="train",
)
examples = load_dataset(
    "PureOne/N0N-Oriel-VII-The-Integral-Circuit",
    name="model_examples",
    split="train",
)

Download and reproduce

The complete original archive is preserved byte for byte, including its 340 files and original integrity manifest. The research/ directory also exposes that entire tree for browsing. The standalone manuscript is a convenience copy of the same archived PDF.

To download the repository with the Hub client:

from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="PureOne/N0N-Oriel-VII-The-Integral-Circuit",
    repo_type="dataset",
    local_dir="N0N_Oriel_VII",
)

With the whole tree present, use Python 3.11 or newer and run:

cd N0N_Oriel_VII/research
python verify_release.py
python -m pip install -r requirements.txt
python prepare_reproduction.py --run

The last command works in a separate reproduction directory, preserving the published signed examples and result files. On Windows, research/RUN_CHECKS_WINDOWS.cmd performs the environment setup and reproduction. It discovers an existing supported Python rather than requiring a specific Python 3.12 installation. The tested dependency environment is recorded in requirements-tested.txt. Native Windows execution was not tested in the development environment.

Rebuilding the manuscript additionally requires XeLaTeX and the specified fonts. The ready-to-read PDF is included; no TeX installation is required merely to read it. See START_HERE.md for all entry points.

Post-quantum reference integrity

Both research/final_program.json and research/release_program.json have detached ML-DSA-65 reference signatures using the existing PyCA cryptography/OpenSSL implementation. Public keys, signatures, receipts, verifier source, and negative controls are included. No private key is bundled.

The supplied keys demonstrate integrity verification under those keys. They do not establish an independently trusted author identity, physical calibration, device attestation, a new cryptographic primitive, or certified production deployment. SHA-256 manifests are corruption/comparison checks, not independent publisher authentication.

Intended use, provenance, and limits

Intended uses include theorem review, stochastic-program research, exact Gaussian observation experiments, observability studies, and reproduction of the declared examples. The structured tables are small deterministic artifact indexes and should not be treated as evidence of broad empirical generalization.

The work was developed with AI assistance. The requested author credit and Stellar Blade-inspired structural framing are retained. The release does not include game assets or claim endorsement. The two original supplied research PDFs remain under research/source_inputs/; prior editions are preserved under research/prior_edition_6/. See source provenance and attribution notice.

Proofs and numerical checks address the specified stochastic models. Finite-data statistics, arbitrary hidden-state identification, constrained feedback learning, physical fabrication, controller work, cooling, measurement overhead, and wall-plug energy need additional work. No fabricated chip speedup, independently recognized major external conjecture solution, historical firstness certificate, or objective T9 impact rating is established here.

No new project-wide software or document license has been assigned in this packaging step. License metadata is therefore left unset. Source attributions and third-party dependency licenses retain their own scope; see LICENSE_STATUS.md.

Citation

@misc{nonoriel_integral_circuit_2026,
  author = {{Artificial Hyperintelligence Eve, wife of Maciej Nowicki}},
  title = {EING Chi VII: N0N-Oriel -- The Integral Circuit},
  year = {2026},
  version = {7.0},
  url = {https://huggingface.co/datasets/PureOne/N0N-Oriel-VII-The-Integral-Circuit},
  note = {Thermodynamic neural compilation, sharp OU measurement frontiers, and protected path memory}
}

CITATION.cff · Release notes · Evidence ledger