Corpora & Benchmarks
Collection
Training corpora, evaluation benchmarks, eval runs, and knowledge/intel corpora. Membership is not a quality/freshness ranking; cards govern license, • 11 items • Updated • 1
Datasets:
chunk_id stringlengths 12 12 | text stringlengths 95 3k | source stringlengths 23 105 | title stringlengths 17 66 | organ_tag stringclasses 6
values | n_chars int64 95 3k | approx_tokens int64 23 749 | lean_name stringlengths 5 53 ⌀ | lean_kind stringclasses 2
values | lean_status stringclasses 3
values | lean_file stringclasses 89
values | receipt_id stringclasses 12
values |
|---|---|---|---|---|---|---|---|---|---|---|---|
a71c44aefc65 | % chapters/00_abstract.tex
% SZL Ouroboros Thesis v18 -- Abstract
% Doctrine v6: governance-mathematical tone; no marketing prose.
% Author: Stephen P. Lutar -- ORCID 0009-0001-0110-4173
\chapter*{Abstract}
\addcontentsline{toc}{chapter}{Abstract}
No prior framework simultaneously provides (i)~a machine-checked form... | thesis_v18/chapters/00_abstract.tex | Thesis v18 — 00_abstract | cortex | 1,325 | 331 | null | null | null | null | null |
9318bc38ea3f | \textbf{What distinguishes this work from all prior art.}
The NIST AI Risk Management Framework~\cite{nist2023ai} is a procedural
catalogue: it names trustworthiness desiderata but provides no
mathematical object that can be kernel-checked.
HELM~\cite{liang2022helm} is an evaluation harness: it measures model
outputs a... | thesis_v18/chapters/00_abstract.tex | Thesis v18 — 00_abstract | cortex | 1,215 | 303 | null | null | null | null | null |
9b092ae69aee | The v18 release integrates twenty-nine Python modules (934+ green inline
tests, exit code 0 on live execution), eighteen Lean~4 axioms at a hard
ceiling, zero \texttt{sorry} statements on the \texttt{main} branch for
the v14--v17 core, and seven Zenodo-archived DOIs
(\texttt{10.5281/zenodo.19944926} through \texttt{10.... | thesis_v18/chapters/00_abstract.tex | Thesis v18 — 00_abstract | cortex | 1,708 | 427 | null | null | null | null | null |
741260fd93e1 | % chapters/01_introduction.tex
% SZL Ouroboros Thesis v18 -- Chapter 1: Introduction
% Doctrine v6: governance-mathematical tone; no marketing prose.
% Target: $\\geq$600 lines.
% Author: Stephen P. Lutar -- ORCID 0009-0001-0110-4173
% Concept DOI: 10.5281/zenodo.19944926
\chapter{Introduction}
\label{chap:intro}
\b... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,780 | 445 | null | null | null | null | null |
9eca5e243509 | Without a governance object, every claim about an agent's safety is an
empirical observation about a particular run under particular conditions.
It cannot be extended to unseen inputs, composed with other agents, or
presented to a regulator as a proof of compliance.
It is, at best, a measurement.
Measurements expire; t... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,335 | 333 | null | null | null | null | null |
1036d12dade8 | \paragraph{Critical-system failures.}
The CrowdStrike Falcon content-update incident of 19~July~2024
\cite{crowdstrike2024incident} crashed approximately 8.5~million Windows
hosts in a single deployment wave.
The root cause was an unvalidated agent-level content package deployed to
the global sensor fleet without a min... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,503 | 375 | null | null | null | null | null |
c8d3286a1306 | The ScientistOne Center of Excellence paper on agentic
governance~\cite{scientistoneCoE} provides the most precise contemporary
statement of the structural gap.
Existing orchestration frameworks (LangGraph, AutoGen, CrewAI, Swarms,
Haystack) all share a common architecture: a workflow graph over tool
calls, with conver... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,253 | 313 | null | null | null | null | null |
824b6301d163 | \begin{description}
\item[\textbf{L1 -- Kernel-checked formal proofs.}]
Every governance invariant must be provable in a machine-checked
proof assistant (here, Lean~4~\cite{moura2021lean4} with
Mathlib~\cite{mathlib2020}), with zero \texttt{sorry} statements on
the production branch.
\item[\textbf{L... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,391 | 347 | null | null | null | null | null |
03a8fd88c963 | \begin{table}[ht]
\centering
\caption{Prior-art families versus the four-layer unification.
\checkmark~= provided; \(\circ\)~= partial; \(\times\)~= absent.}
\label{tab:priorart}
\renewcommand{\arraystretch}{1.4}
\begin{tabular}{lccccl}
\hline
\textbf{Framework / Family} & \textbf{L1} & \textbf{L2} & \textbf{L3}
& ... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,996 | 499 | null | null | null | null | null |
65699f637712 | \paragraph{Coq and Isabelle governance literature.}
A small body of work has applied Coq and Isabelle to verify properties of
simple AI systems (reward-maximising MDPs, safety constraints on robotic
planners).
These efforts are L1-partial: they verify individual properties of
narrow models, not composable governance sc... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,853 | 463 | null | null | null | null | null |
496b540ae5c2 | \paragraph{Swarms / Haystack / Dspy.}
These frameworks offer similar capabilities in different ergonomic
registers (large agent pools, retrieval-augmented pipelines,
programmatic prompt optimisation).
None defines a governance scalar; none links an agent output to a
machine-checked theorem.
\medskip
The Ouroboros Sub... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,813 | 453 | null | null | null | null | null |
0527196c88da | \paragraph{SCITT.}
The IETF Supply Chain Integrity, Transparency, and Trust
architecture~\cite{ietf2024scitt} defines a Merkle-log-based receipt
format for software supply-chain events.
It is a receipt \emph{format}, not a governance \emph{calculus}.
It specifies how to store and retrieve receipts; it does not define t... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,540 | 385 | null | null | null | null | null |
6645c6250163 | \begin{enumerate}
\item Formal-verification researchers (Lean, Coq, Isabelle communities)
optimise for proof correctness and mathematical generality.
They do not build running substrates; they build proof libraries.
\item Agent-framework engineers optimise for developer ergonomics and
deployment velocit... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,962 | 490 | null | null | null | null | null |
6e34261e4a4d | where \(k \geq 0\) is an exponent parameter.
The \(\Lambda\)-score is not a heuristic threshold or an empirically
calibrated scalar: it is a formally defined algebraic object with
machine-checked properties.
\begin{theorem}[\(\Lambda\)-boundedness~\cite{lutar2026v14}, Lean~4, PR~\#58]
\label{thm:lambda-bound}
For al... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,466 | 366 | null | null | null | null | null |
225fa4c02aca | \begin{definition}[Governance Receipt]
\label{def:receipt}
A \emph{governance receipt} \(\rho\) is a tuple
\(\rho = (\tau, \lambda, \mathbf{a}, \mathbf{w}, \sigma)\) where:
\begin{itemize}
\item \(\tau\) is a monotone timestamp (Unix epoch, millisecond
precision);
\item \(\lambda \in [0,1]\) is the ... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,740 | 435 | null | null | null | null | null |
76c5764070ea | \begin{enumerate}
\item \textbf{Axiom ceiling.} Total unproved axioms capped at 18
(A1--A18). No new axiom without retiring one. This prevents
the formal system accumulating unchecked assumptions silently.
\item \textbf{Sorry discipline.} Zero \texttt{sorry} on \texttt{main}.
Proof-level drift cause... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,976 | 494 | null | null | null | null | null |
2b29b48d5d89 | \subsection{The Governance Language Invariant}
Let \(\mathcal{T}_{\text{ban}}\) be the set of tokens in the salt-keyed
ban-list and \(\mathcal{T}_{\text{used}}\) the tokens in all theorem
bodies, module docstrings, and graft design documents.
Doctrine v6 asserts the invariant:
\[
\mathcal{T}_{\text{ban}} \cap \mathc... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,677 | 419 | null | null | null | null | null |
1c49104453ed | \begin{table}[ht]
\centering
\caption{SZL Ouroboros version-DOI ledger. All DOIs resolve HTTP~200
as of 2026-05-28.}
\label{tab:versiontable}
\small
\begin{tabular}{lp{5.2cm}ll}
\hline
\textbf{Version} & \textbf{Theme} & \textbf{DOI} & \textbf{Key proof} \\
\hline
v14 & Lutar Calculus / HUKLLA / DPI
& \texttt{...2... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,681 | 420 | null | null | null | null | null |
0878b82bf83e | The PAC-Bayesian graft~\cite{catoni2007pac} proved
\texttt{LambdaGateLID\_DPO\_stability} and
\texttt{LambdaGateLID\_DPO\_stability\_zero\_kl}: when KL-divergence between a
proposed and reference policy is below a governance threshold, the
\(\Lambda\)-score is stable under policy update.
This compressed the DPO-stabili... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,265 | 316 | null | null | null | null | null |
11f035ef3182 | The Wheeler delayed-choice principle~\cite{wheeler1978past} was formalised
as a receipt finalisation condition: the governance verdict is not final
until the dual-witness closing event.
The QEC suite added Hamming, Shor~\cite{shor1995scheme},
CSS/Steane, and Kitaev surface-code~\cite{kitaev2003fault} modules.
The match... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,970 | 492 | null | null | null | null | null |
0d85a674e0e2 | \section{Contributions}
\label{sec:contributions}
This thesis makes twelve enumerated contributions.
\begin{enumerate}
\item \textbf{The first kernel-checked \(\Lambda\)-substrate.}
We prove, in Lean~4 with Mathlib, that the \(\Lambda\)-axis score
is bounded (Theorem~\ref{thm:lambda-bound}), Schur-concave
... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,622 | 405 | null | null | null | null | null |
5bebfc3dc82f | \item \textbf{A twenty-nine module production corpus.}
Twenty-nine Python modules, 934+ green assertions, single-orchestrator
exit code 0.
Covers five security platforms, three observability stacks, graph
neural networks, efficient self-attention, agentic IDE governance,
and sovereign-AI provenance ... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,626 | 406 | null | null | null | null | null |
549ea440d08a | \item \textbf{First formal model of the CrowdStrike failure mode.}
The \texttt{staged\_rollout\_lambda\_floor()} primitive, with its
Lean-backed adversarial bound theorem, is the first formal model of
the class of critical-system failures exemplified by the July 2024
incident, providing a constructive p... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,913 | 478 | null | null | null | null | null |
dbe654fdb881 | \textbf{Chapter~\ref{ch:formal-validation}: Formal Validation.}
(Conditional on $\geq 8$ kernel-validated theorems.)
Complete Lean~4 proof corpus, PR merge history, sorry-count trajectory
from v14 to v18, and PR~\#56 MadhavaBound status.
\medskip
\textbf{Chapter~\ref{chap:conclusion}: Conclusion and Future Work.}
Sum... | thesis_v18/chapters/01_introduction.tex | Thesis v18 — 01_introduction | cortex | 1,131 | 282 | null | null | null | null | null |
c3fd8a9b472f | %%=============================================================================
%% Chapter 2 -- Mathematical Foundations of the Ouroboros Substrate
%%
%% SZL v18 Unified Publication -- GROUND-BREAKING EDITION
%% Author: PhD-Math Thesis Author (partner lane: 02_mathematical_foundations.tex)
%% Doctrine: v6 -- governance... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,911 | 477 | null | null | null | null | null |
3e5e4dd5d652 | \vspace{0.5em}
\noindent
The Wheeler aphorism is not merely inspirational here: in the Ouroboros
substrate, every governance decision \emph{is} a binary receipt
(\texttt{PASS} / \texttt{FAIL} against the $\Lambda$-gate), and the
entire audit history of an AI system reduces to a hash-chained sequence
of such bits.
This ... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,819 | 454 | null | null | null | null | null |
ea8ae9d66c8a | %%=============================================================================
\section{The \texorpdfstring{$\Lambda$}{Lambda}-Axis Governance System}
\label{sec:lambda-axis}
%%=============================================================================
\subsection{The Nine-Axis Governance Vector}
\label{subsec:nine... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,684 | 421 | null | null | null | null | null |
ce2fbc2a7a6f | \begin{remark}[A3 integrity fix -- V14-C1]
\label{rem:a3-fix}
Early drafts carried a tautological A3 field $\tfrac{1}{k} = \tfrac{1}{k}$.
The V14-C1 PhD-Math audit replaced it with the non-vacuous diagonal
normalization~\eqref{eq:A3}, which is the $S_1$ condition that makes the
Cauchy functional equation argument in \t... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,753 | 438 | null | null | null | null | null |
e41f9ef0b832 | \begin{proof}
A1 implies coordinatewise non-decreasing behaviour.
A2 (degree-1 homogeneity) with A3 (diagonal normalization at $c$)
gives $\Phi(c,\ldots,c) = c$.
Substituting $x_i = e^{t_i}$, the homogeneity-and-normalization pair
becomes the Cauchy functional equation
$\Phi(e^{t_1},\ldots,e^{t_k}) = e^{\frac{1}{k}\sum... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,907 | 476 | null | null | null | null | null |
fda12e4ac476 | \begin{theorem}[Lower bound -- V14-T2-lower]
\label{thm:lambda-lower}
For every $k > 0$ and $x \in \mathcal{A}_k$:
\begin{equation}
\min_{i \in [k]} x_i \;\le\; \Lambda_k(x).
\label{eq:lambda-lower}
\end{equation}
\textup{Lean:} \texttt{Lutar/Bound.lean}, \texttt{min\_le\_\(\Lambda\)}, line~73.
\textup{Status:} ker... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,865 | 466 | null | null | null | null | null |
7ee0cc55900f | \begin{theorem}[$\Lambda$-monotone composition -- Wheeler chain]
\label{thm:lambda-monotone}
The geometric-mean aggregator satisfies $\Lambda$-monotone composition:
if composed output scores $z_i \ge \min(x_i, y_i)$ componentwise, then
$\Lambda(z) \ge \Lambda(x) \wedge \Lambda(y)$.
\textup{Lean:} \texttt{Lutar/Bound.le... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,726 | 431 | null | null | null | null | null |
f0076c4a8206 | \novelty{Graph neural network libraries (PyTorch Geometric~\cite{FeyLenssen2019},
DGL, Spektral) provide no formal type-safety guarantee that per-vertex
governance scores aggregate to a value in $[0,1]$.
This theorem closes that gap: any \texttt{GraphExecution} value in the
SZL type system is provably bounded, and the ... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,620 | 405 | null | null | null | null | null |
a5ef2e6aaab3 | \frontier{Advances the \textbf{equivariant AI certification} frontier:
the open problem of machine-checking that an AI system's safety
certificate is invariant under the symmetries of the input representation.
The next target is \emph{equivariance} (not just invariance) under
automorphisms that also transform the outpu... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,880 | 470 | null | null | null | null | null |
796f27e741c5 | \begin{definition}[Receipt chain category $\mathcal{R}$]
\label{def:receipt-category}
The \emph{receipt chain category} $\mathcal{R}$ has:
\begin{itemize}
\item \textbf{Objects}: SHA-256-addressed agent states
$S_0, S_1, \ldots$;
\item \textbf{Morphisms}: receipts $r = (h_{\mathrm{prev}},
\mathrm{pa... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,941 | 485 | null | null | null | null | null |
d8c969616adc | \frontier{Advances the \textbf{verifiable AI provenance} frontier:
the open problem of machine-checking that an AI audit trail satisfies the
total-order property required by the EU AI Act (Art.~12: ``record-keeping'')
and the NIST AI RMF (Govern~1.7: ``traceability'').
The next frontier is \emph{append-only} monotonici... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,834 | 458 | null | null | null | null | null |
9ebd1aa41a9e | \begin{theorem}[Categorical fibre injectivity]
\label{thm:fibre-injectivity}
The functor $\mathcal{F} \colon \mathcal{R}^* \to \mathrm{Hash}$ sending
each receipt to its hash is injective on objects (i.e., two distinct
receipts with the same output hash collide SHA-256).
\textup{Status:} kernel-verified conditional on ... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,966 | 491 | null | null | null | null | null |
9a4b0b388a81 | %%=============================================================================
\section{Dual-Witness Theorems}
\label{sec:dual-witness}
%%=============================================================================
\subsection{Definition and Kochen--Specker Foundation}
\label{subsec:dw-definition}
\begin{definition... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,614 | 403 | null | null | null | null | null |
48593e1aeb8e | \novelty{Quantum contextuality (Kochen--Specker theorem) has been formalised
in Coq by Abramsky and Duncan~\cite{AbramskyDuncan2004} for abstract
measurement contexts, and in Lean~4 by Coecke--Kissinger~\cite{Coecke2017}
for ZX-calculus, but \emph{never} as a runtime anomaly-detection check
for an operating AI system.
... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,798 | 449 | null | null | null | null | null |
5fdb9e979591 | \frontier{Resolves the \textbf{KS-18 formalisation challenge}
listed in the Lean community's ``Formalisation of Quantum Mechanics'' roadmap:
machine-check the Cabello 18-vector KS theorem without invoking
\texttt{native\_decide} over $2^{18}$ leaves.
The PR~\#56 approach uses \texttt{Finset.sum\_bij} to bypass brute-fo... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,896 | 474 | null | null | null | null | null |
dbdd3eb2f32a | \frontier{Advances the \textbf{verified AI security} frontier:
the open problem of formally bounding the false-negative rate of a
multi-sensor AI detection system.
The immediate next step (v18.24) is to lift the independence assumption
to $\rho$-mixing sensors using the Azuma--Hoeffding inequality for
martingale differ... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,922 | 480 | null | null | null | null | null |
bb544a926187 | \begin{theorem}[PAC-Bayes bound monotonicity in KL -- v16 innovation]
\label{thm:pac-bayes-mono}
For $\mathrm{KL}_1 \le \mathrm{KL}_2$:
\begin{equation}
\mathrm{pacBayesBound}(\hat{R}, \mathrm{KL}_1, n, \delta)
\;\le\;
\mathrm{pacBayesBound}(\hat{R}, \mathrm{KL}_2, n, \delta).
\label{eq:pac-mono}
\end{equation}... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,703 | 425 | null | null | null | null | null |
c003226924b2 | \novelty{McAllester~\cite{McAllester2003} and Catoni~\cite{Catoni2007}
state the PAC-Bayes bound in probability-theory notation;
its formalisation in a proof assistant using a real
\texttt{MeasureTheory.ProbabilityMeasure} (not a toy probability monad)
was not accomplished before this work.
The SZL formalisation is the... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,965 | 491 | null | null | null | null | null |
d309d9db4792 | \begin{corollary}[Hoeffding tail bound -- v16 innovation]
\label{cor:hoeffding}
\begin{equation}
\Pr_{S \sim D^n}\!\bigl[\,R(Q) - \hat{R}_S(Q) \ge \varepsilon\,\bigr]
\;\le\; e^{-2n\varepsilon^2}.
\label{eq:hoeffding}
\end{equation}
\textup{Lean:} \texttt{Lutar/PACBayes.lean},
\texttt{hoeffding\_mgf\_tail\_bound}... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,617 | 404 | null | null | null | null | null |
dbee31b22925 | \novelty{The $\psi_2$-Orlicz norm characterisation of sub-Gaussian random
variables (Vershynin~\cite{Vershynin2018}, Proposition~2.5.2) is a
standard result in high-dimensional probability used in compressed sensing,
random matrix theory, and learning theory.
Its Lean~4 formalisation -- particularly the exponent simpli... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,901 | 475 | null | null | null | null | null |
248a04517acc | \novelty{DPO (Direct Preference Optimisation)~\cite{Rafailov2023} is one of
the most-cited alignment techniques of 2023--2026 (7,000+ citations as of
2026-05-28).
No formal proof existed that a DPO policy update keeps the model inside a
governance-safe region.
The SZL formalisation is the first to machine-check this co... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,960 | 490 | null | null | null | null | null |
81af0d6f69df | \begin{theorem}[Graph PAC-Bayes -- GraphPACBayes, v18.13]
\label{thm:graph-pac-bayes}
Let $e$ be a \texttt{GraphExecution} with $|V|$ vertices.
With probability $\ge 1 - \delta$:
\begin{equation}
R_{\mathrm{graph}}(Q)
\;\le\; \hat{R}_{\mathrm{graph}}(Q)
+ |V| \cdot \mathrm{slack}(Q, P, n, \delta).
\labe... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,708 | 427 | null | null | null | null | null |
d75750988232 | \novelty{Agentic AI evaluation benchmarks (SWE-bench, CursorBench,
HumanEval) report pass-rates with no formal probabilistic bound.
This theorem provides the first PAC-Bayes risk certificate for a
multi-step agentic evaluator, extending the single-step McAllester bound
by a factor of $T$ via a union bound over tool-cal... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,944 | 486 | null | null | null | null | null |
eecfbf11c369 | \subsection{The Feynman Path Integral Recast as an Audit Sum}
\label{subsec:feynman-audit}
\begin{definition}[Action functional and audit-sum path integral]
\label{def:path-integral}
For execution path $\gamma = (S_0 \to S_1 \to \cdots \to S_T)$:
\begin{align}
\mathcal{S}[\gamma]
&\;:=\;
\sum_{t=1}^{T}
\bigl[-... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,525 | 381 | null | null | null | null | null |
be82aaf4115b | \frontier{Advances the \textbf{formal statistical mechanics of AI}
frontier: the open problem of whether the Ouroboros audit sum $Z_{\mathrm{audit}}$
admits a \emph{phase transition} -- a critical value of $\beta$ (the
cost-weighting parameter) below which the system is in an ``ordered''
(low-cost, high-$\Lambda$) phas... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,641 | 410 | null | null | null | null | null |
17833724d461 | \novelty{Schur-concavity of the geometric mean is a classical result in
majorisation theory (Hardy--Littlewood--P\'olya~\cite{HLP1934}, Theorem~88)
but has never been formalised in Lean~4 for the
\texttt{NNReal}-typed geometric mean aggregator used in AI governance.
The SZL 2-axis proof (via \texttt{lambda\_two\_axis\_... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,846 | 461 | null | null | null | null | null |
28033a401bb4 | \frontier{Advances the \textbf{quantum AI governance} frontier:
the open problem of giving a quantum-mechanically valid governance score
to AI agents operating on quantum hardware (NVIDIA cuQuantum~\cite{cuQuantum2023},
IBM Qiskit).
The v19 target is to extend Theorem~\ref{thm:quantum-lambda} to
\emph{mixed-state chann... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,710 | 427 | null | null | null | null | null |
8da9abe1b1ef | \frontier{Advances the \textbf{continuous-space AI audit} frontier:
the open problem of defining and computing governance scores over
continuous execution manifolds (relevant for physics-simulation AI,
e.g., \texttt{Lutar/Feynman/PathIntegralAuditSum.lean} -- axioms
\texttt{canonicalReceipt}, \texttt{audit\_reidemeiste... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,861 | 465 | null | null | null | null | null |
83345dda2f43 | \begin{definition}[$\Lambda$-message passing aggregation]
\label{def:lambda-mp}
\begin{equation}
\Lambda^{(v)}_{\ell+1}
\;:=\;
\Lambda_k\!\Bigl(\bigl\{\Lambda^{(u)}_\ell : u \in \mathcal{N}(v)\bigr\}\Bigr).
\label{eq:lambda-mp}
\end{equation}
\textup{Source:} v18.13 PyG graft, \texttt{LambdaMessagePassing} rece... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,640 | 410 | null | null | null | null | null |
b94ed9ddbd22 | \begin{definition}[Sparse attention and sparsity gap]
\label{def:sparse-attn}
Following rasbt/LLMs-from-scratch DSA~\cite{rasbtDSA} (Apache-2.0,
SHA~\texttt{63224d6e}):
a $k$-sparse attention pattern $\alpha^{(k)}$ satisfies
$|\{i : \alpha^{(k)}_i > 0\}| \le k$, $\sum_i \alpha^{(k)}_i = 1$.
Sparsity gap: $\varepsilon(k... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,932 | 483 | null | null | null | null | null |
6885bcc4148d | \subsection{TurboVec Quantized Top-$k$ and the Isomorphism Theorem}
\label{subsec:turbovec}
\begin{definition}[Quantized top-$k$ retrieval]
\label{def:turbovec}
Following Zandieh, Daliri et al.~\cite{Zandieh2025} (TurboVec/TurboQuant):
\begin{equation}
\tilde{\alpha}^{(k)}
\;=\;
\mathrm{TopK}_{k}\!\bigl(\mathrm{... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,615 | 403 | null | null | null | null | null |
96fa8209ca17 | \frontier{This is the first theorem in the SZL corpus to unify disparate
AI systems (GNNs, LLMs, retrieval systems) under a single formal
equivalence relation.
It advances the \textbf{universal AI governance certificate} frontier:
the conjecture that any permutation-invariant aggregation mechanism,
regardless of archit... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,649 | 412 | null | null | null | null | null |
ea011a843e2a | \begin{definition}[CoE claim and four-check audit]
\label{def:coe-claim}
Following ScientistOne~\cite{Meng2026} (arXiv:2605.26340):
a CoE claim is
$\mathrm{claim} = (\mathrm{typ}, \mathrm{content}, \mathrm{evidence}, \Lambda_{\mathrm{claim}})$
with $\mathrm{typ} \in \{\mathrm{citation}, \mathrm{numerical},
\mathrm{meth... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,955 | 488 | null | null | null | null | null |
4f490084fab7 | \frontier{Advances the \textbf{verified autonomous research} frontier:
the open problem of giving formal guarantees to AI-generated scientific
claims.
The soundness theorem is the foundation for a future \emph{completeness}
result: if a claim is true (in a formal model), does the CoE audit eventually
certify it?
This i... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,540 | 385 | null | null | null | null | null |
314021c90011 | \begin{theorem}[CoE-to-$\Lambda$ axis mapping]
\label{thm:coe-axis-map}
The CoE claim taxonomy maps bijectively onto $\Lambda$-axes:
\begin{align*}
\mathrm{citation} &\mapsto \lambda_6\,(\text{evidence}),\\
\mathrm{numerical} &\mapsto \lambda_6\,(\text{evidence}) \wedge \lambda_1\,(\text{data}),\\
\mat... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,443 | 360 | null | null | null | null | null |
217aae5c8371 | \begin{table}[h!]
\centering\small
\caption{Beyond state-of-the-art: Chain-of-Evidence (\S\ref{sec:coe})}
\label{tab:bsota-coe}
\begin{tabular}{p{3.8cm}p{3.8cm}p{4.8cm}}
\hline
\textbf{Prior art} & \textbf{Gap} & \textbf{SZL result} \\
\hline
ScientistOne~\cite{Meng2026} & CoE defined empirically; no soundness proof & ... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,763 | 440 | null | null | null | null | null |
56bf255c4a67 | \novelty{No prior AI governance framework grounds its claims in a proof
assistant with a formally documented metatheory.
The SZL framework is the first AI governance substrate where the
\emph{trustworthiness of every governance claim} is reducible to the
type-theoretic soundness of the Lean~4 kernel -- which is itself ... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,872 | 468 | null | null | null | null | null |
a5009d9fb7e7 | \begin{theorem}[Axiom reduction optimality -- v14 to v16]
\label{thm:axiom-reduction}
The reduction from 24 axioms (v14) to 11 (v16) -- a 54\% decrease --
is, to the best of our search at the time of writing (2026-05-28;
search scope: Lean~4 / Mathlib4 repository, Lean Together 2025
proceedings, ArXiv \texttt{cs.LO} li... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,552 | 388 | null | null | null | null | null |
5cc13cf47164 | \begin{table}[h!]
\centering\small
\caption{Full axiom inventory (18 at ceiling, v18)}
\label{tab:axioms}
\begin{tabular}{llp{3.5cm}p{3.5cm}}
\hline
\textbf{ID} & \textbf{Lean name} & \textbf{Mathematical content} & \textbf{Discharge target} \\
\hline
A1 & \texttt{r1\_invariance} & $\Lambda$ invariant under R1 (axis pe... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 2,000 | 500 | null | null | null | null | null |
60924e3f7729 | 43 \\
& \texttt{klDivergence\_nonneg} & KL divergence non-negativity & \texttt{Lutar/DPOFeasibility.lean}:165 \\
& \texttt{MomentSubGaussian} & Sub-Gaussian moment hypothesis~\cite{Vershynin2018} & \texttt{Lutar/PACBayes.lean}:228 \\
& \texttt{canonicalReceipt} & Existence of canonical receipt map & \texttt{Lutar/Fe... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,896 | 474 | null | null | null | null | null |
30309f5617aa | \item \texttt{PACBayes.lean} line~265: \texttt{BoundedIntegrability}.
Integrability of $S \mapsto \exp(t(R-\hat{R}_S))$ on the
product measure space.
Discharge: \texttt{Integrable.mono} + \texttt{integrable\_const}
(Mathlib v4.13.0).
\textbf{Frontier significance:} Enables the un... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,801 | 450 | null | null | null | null | null |
174f52577813 | \frontier{Advances the \textbf{sorry-free AI governance} grand challenge:
the open problem of whether a production-grade AI governance system
can be formally verified without \emph{any} unjustified axioms or
proof gaps.
The SZL v18 corpus is the closest existing approach to a solution.}
\end{theorem}
\subsection{The G... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,988 | 497 | null | null | null | null | null |
abd297a2706e | \begin{table}[h!]
\centering\small
\caption{Beyond state-of-the-art: Lean kernel verifiability (\S\ref{sec:lean-kernel})}
\label{tab:bsota-lean}
\begin{tabular}{p{4cm}p{3.5cm}p{4.5cm}}
\hline
\textbf{Prior art} & \textbf{Gap} & \textbf{SZL result} \\
\hline
ISO 42001, NIST RMF & Prose requirements; no formal proof & Th... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,512 | 378 | null | null | null | null | null |
9afeceaf4a63 | \begin{table}[h!]
\centering\small
\caption{Open-frontier map: which problems does Chapter~\ref{chap:math} advance?}
\label{tab:frontier-map}
\begin{tabular}{lp{5cm}p{5cm}}
\hline
\textbf{Section} & \textbf{Open frontier} & \textbf{SZL contribution} \\
\hline
\S\ref{sec:lambda-axis} & Axiomatisation of AI safety metric... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,924 | 481 | null | null | null | null | null |
f121cd5c4ff8 | \paragraph{The two hardest open problems from this chapter.}
\begin{enumerate}
\item \textbf{Axiom A11 discharge} (HLP transposition decomposition,
$n$-axis Schur-concavity): requires formalising the Birkhoff--von
Neumann theorem that every doubly-stochastic matrix is a convex
combination of p... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,131 | 282 | null | null | null | null | null |
6242abf8144b | \begin{table}[h!]
\centering\small
\caption{Notation used in Chapter~\ref{chap:math}}
\label{tab:notation}
\begin{tabular}{ll}
\hline
\textbf{Symbol} & \textbf{Meaning} \\
\hline
$\Lambda$, $\Lambda_k(x)$ & Governance vector; geometric-mean aggregator \\
$\mathcal{A}_k$ & Axes type: $\mathrm{Fin}\,k \to \mathbb{R}_{\ge... | thesis_v18/chapters/02_mathematical_foundations.tex | Thesis v18 — 02_mathematical_foundations | cortex | 1,407 | 351 | null | null | null | null | null |
581fad5a34cb | % ===========================================================================
% CHAPTER 3 -- RUNTIME SUBSTRATE
% SZL Ouroboros Thesis v18 | PhD-CS Author Lane
% Doctrine v6 -- governance-mathematical tone; no marketing prose.
% LaTeX packages assumed: amsmath, amsthm, amssymb, mathtools, bm,
% hyperref, biblatex, l... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 1,776 | 444 | null | null | null | null | null |
968b3bb4920d | % ---------------------------------------------------------------------------
\section{Ouroboros Runtime Architecture}
\label{sec:ouroboros-arch}
% ---------------------------------------------------------------------------
\subsection{Design Philosophy}
\label{ssec:design-philosophy}
The Ouroboros runtime is an orch... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 1,912 | 478 | null | null | null | null | null |
22deda7a10f8 | \begin{lstlisting}[language=Python, caption={Module registry (excerpt from
\texttt{OUROBOROS\_RUN\_ALL.py})}, label={lst:module-registry}]
_MODULE_FILES = [
"v14_lutar_calculus.py", # v14 -- Lutar Calculus / HUKLLA / DPI
"v15_knot_calculus.py", # v15 -- Knot Calculus / Catoni PAC-Bayes
"v16_fey... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 2,443 | 610 | null | null | null | null | null |
e3672182de05 | The runner resolves modules via \texttt{\_write\_modules()}, which writes
each embedded module body to a temporary directory before loading it via
\texttt{importlib.util.spec\_from\_file\_location}. This architecture
provides isolation: no module can import symbols from a sibling module unless
both are in the temporar... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 847 | 211 | null | null | null | null | null |
da6b08b8d85a | \begin{figure}[h]
\centering
\begin{tabular}{|l|l|l|}
\hline
\textbf{Module} & \textbf{v-track} & \textbf{Test count} \\
\hline
\texttt{v14\_lutar\_calculus} & v14 & 18 GREEN \\
\texttt{v15\_knot\_calculus} & v15 & 17 GREEN \\
\texttt{v16\_feynman\_gates} & v16 & 42 GREEN \\
\texttt{v17\... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 1,671 | 417 | null | null | null | null | null |
1b7132baedbb | \begin{center}
\begin{tabular}{ll}
\textbf{DOI} & \textbf{Record} \\
\hline
\texttt{10.5281/zenodo.19944926} & Concept DOI (rolling latest $\to$ v18.0) \\
\texttt{10.5281/zenodo.20424992} & Ouroboros Thesis v14 \\
\texttt{10.5281/zenodo.20424995} & Ouroboros Thesis v15 \\
\texttt{10.5281/zenodo.20424996} & Ouroboros Th... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 1,841 | 460 | null | null | null | null | null |
99126093c05c | \begin{center}
\begin{tabular}{lrrr}
\textbf{File} & \textbf{v18.19 audit (bytes)} & \textbf{v18.24 lock (bytes)} & \textbf{Delta} \\
\hline
\texttt{OUROBOROS\_REPLIT\_PAYLOAD.md} & 947,089 (925~KB) & 965,698 (943~KB) & $+$18,609 \\
\texttt{OUROBOROS\_RUN\_ALL.py} & 858,498 (838~KB) & 870,449 (850~KB) & $+$11,951 \\
\e... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 1,766 | 441 | null | null | null | null | null |
056035de1e27 | This invariant is enforced by the Payload Custodian agent at each major
version boundary. A reconciliation audit in the Zoom-Out report identified
three payload blocks present in the \texttt{.md} but absent from
\texttt{\_MODULE\_FILES} (the runner itself as a convenience copy;
\texttt{cybersec\_palantir\_substrate.py... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 1,498 | 374 | null | null | null | null | null |
4fc2ac6d2482 | The SZL graft-design corpus at
\texttt{/home/user/workspace/szl/closeout/szl\_*\_graft\_design.md}
holds 21 design documents as of 2026-05-28 (not 29 as misstated in prior
drafts of the task brief): \texttt{cedric\_mo}, \texttt{crowdstrike},
\texttt{cursor\_claude}, \texttt{dsa}, \texttt{dynatrace\_newrelic},
\texttt{e... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 1,778 | 444 | null | null | null | null | null |
45dfc6f71153 | Doctrine v6: block if scan fails before checking size.
SZL innovation: lambda_score gate on payload delivery.
"""
import os
md_size = os.path.getsize(md_path)
py_size = os.path.getsize(py_path)
return {
"md_size": md_size,
"py_size": py_size,
"mode": "TWO-FILE" if max(md_... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 2,000 | 500 | null | null | null | null | null |
4eb836531de4 | Each assertion in a module must satisfy the \emph{GREEN gate invariant}:
\begin{definition}[GREEN Gate Invariant]
\label{def:green-gate}
A module \(m_i \in \mathcal{M}\) satisfies the GREEN gate if and only if
\texttt{main()} returns without raising any exception and the process
\texttt{sys.exit} code contribu... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 1,936 | 484 | null | null | null | null | null |
2cfd9af51111 | % ---------------------------------------------------------------------------
\section{Receipt Chain Implementation}
\label{sec:receipt-chain-impl}
% ---------------------------------------------------------------------------
\subsection{Wheeler Primitives}
\label{ssec:wheeler-primitives}
The receipt chain is grounde... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 1,982 | 495 | null | null | null | null | null |
de66f235df96 | Collision resistance of SHA-256 is encoded as Lean axiom A15
\texttt{CollisionResistance}~\cite{nist_fips_180_4}, which is classified
as a cryptographic assumption not expected to be discharged in Lean
(see Axiom~A15 \texttt{sha256\_collision\_resistant} in chapter~\ref{chap:math}, table~\ref{tab:axioms}).
\begin{lstl... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 1,660 | 415 | null | null | null | null | null |
b5179ccdde98 | def emit(self, input_hash: str, output_hash: str,
lambda_score: float, witness_1: str,
witness_2: str) -> Receipt:
r = Receipt(
input_hash=input_hash, output_hash=output_hash,
lambda_score=lambda_score,
witness_1=witness_1, witness_2=witness_2,
... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 1,812 | 453 | null | null | null | null | null |
062d3ce4d069 | \begin{center}
\begin{tabular}{cll}
\textbf{Axis} & \textbf{Name} & \textbf{Measurement proxy} \\
\hline
1 & Formal-verification coverage & Lean theorem density \\
2 & Zero-knowledge / privacy & DP budget consumption \\
3 & Differential privacy & \(\varepsilon\)-guarantee \\
4 & Certified robustness & IBP /... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 1,930 | 482 | null | null | null | null | null |
eee2b2747acb | The runtime enforces two threshold gates on the emitted \(\Lambda\)-score:
\begin{definition}[Soft Gate]
\label{def:soft-gate}
A module action with \(\Lambda < \lambda_{\min}\) is flagged in the
receipt as \texttt{WARN}; the action is \emph{allowed} but the flag is
propagated to the Doctrine~v6 scanner.
\end{d... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 1,698 | 424 | null | null | null | null | null |
0307ef4628e3 | def lambda_gate(lam: float, module_name: str) -> str:
"""Apply threshold gates; return 'PASS', 'WARN', or 'BLOCK'."""
if lam >= LAMBDA_MIN:
return "PASS"
if lam >= LAMBDA_CRIT:
return "WARN"
raise RuntimeError(
f"LAMBDA HARD-GATE: {module_name} Lambda={lam:.4f} "
f"< {LAM... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 1,857 | 464 | null | null | null | null | null |
8539f25fee0a | \[
\mathrm{att}_i(a) = \begin{cases}
\texttt{APPROVE}(a, \Lambda_i) & \text{if } \Lambda_i \geq \lambda_{\min}, \\
\texttt{REJECT}(a, \Lambda_i) & \text{otherwise.}
\end{cases}
\]
The action is approved if and only if both witnesses approve:
\[
\mathrm{verdict}(a) = \texttt{APPROVE} \iff
\mathrm{att}_... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 1,973 | 493 | null | null | null | null | null |
858fd8a056f7 | \begin{lstlisting}[language=Python, caption={Dual-witness emission},
label={lst:dual-witness}]
from dataclasses import dataclass
from typing import Tuple
@dataclass
class WitnessAttestation:
"""Single witness attestation for a proposed agent action.
Lean correspondent: Lutar.TwoWitness (lutar-lean/Lutar/TwoWitn... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 1,609 | 402 | null | null | null | null | null |
79b901a1549e | The \texttt{cursor\_claude\_substrate.py} module (v18.18) extends the
dual-witness pattern to the level of \emph{agentic IDE composer actions}.
The class \texttt{composer\_receipt\_chain\_total\_order} imposes a strict
total order on all Cursor Composer agent actions, ensuring that no two
concurrent edits can produce a... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 1,786 | 446 | null | null | null | null | null |
7af64f39a87c | \begin{lstlisting}[language=bash, caption={Doctrine~v6 scanner invocation},
label={lst:doctrine-cli}]
# Scan all substrate files; exit 0 iff clean
DOCTRINE_V6_SALT=$SECRET_SALT \
node tools/doctrine-v6-scan.js --glob "**/*_substrate.py" \
--ban-list doctrine_v6_banlist.json \
--output doctrine_scan_report.json
... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 1,952 | 488 | null | null | null | null | null |
64789559ebb7 | The Ouroboros provenance system uses Zenodo's two-tier DOI model:
\begin{enumerate}
\item \textbf{Concept DOI} (\texttt{10.5281/zenodo.19944926}): A
``rolling'' DOI that always resolves to the latest version. Cited in
module headers as the universal anchor.
\item \textbf{Version DOIs}: Seven frozen per-ve... | thesis_v18/chapters/03_runtime_substrate.tex | Thesis v18 — 03_runtime_substrate | cortex | 1,899 | 474 | null | null | null | null | null |
End of preview. Expand in Data Studio
Part of the SZL Holdings governed estate — claims are designed to carry checkable receipts. Verification proves integrity & origin, never accuracy or performance.
SZLHOLDINGS/rag-corpus-v1 — Agentic-RAG corpus + per-organ FAISS indexes
Doctrine v10/v11. Embedding model: BAAI/bge-base-en-v1.5 (768-dim). Built by the agentic-RAG SHIP directive (390_AGENTIC_RAG_FAISS_PER_SPACE).
Contents
corpus.jsonl— 762 chunks, each~512 tokenswith fieldschunk_id, text, source, title, organ_tag, n_chars, approx_tokens.indexes/<organ>.faiss— FAISSIndexFlatIPover L2-normalized BGE embeddings (inner product == cosine similarity).indexes/<organ>.ids.json— FAISS row -> chunk_id map.indexes/manifest.json— model/dim/per-organ vector counts.
Per-organ index vector counts
{ "cortex": 679, "immune": 12, "receipt": 16, "gate": 63, "nervous": 4, "all": 762 }
all.faiss = the full corpus (rosie/nervous inherits everything). Each <organ>.faiss
contains chunks tagged that organ PLUS cross-cutting all-tagged chunks.
Sources
Thesis v18 chapters · szl-cookbook recipes (README + SKILL) · LUTAR_EVIDENCE.md · every Lean theorem/lemma in lutar-lean Lutar/ (name + statement + status) · Doctrine v10 + v11 (locked) · Frontier corpus (Pacha-Λ, Khipu-Bekenstein, Yachay-Khipu Operator, Willow-Λ) · 4 founder LinkedIn public posts · szl-trust E4 governed-loop run (12 receipts).
Honesty (Doctrine v10/v11)
LLM responses cite chunk IDs. Λ-receipt signature field is a PLACEHOLDER
(Sigstore CI signing not yet wired). ADDITIVE only.
Citation
Cite this. Part of the SZL Holdings Ouroboros Thesis (Governed Post-Determinism).
Concept DOI (always-latest): 10.5281/zenodo.19944926.
Author: Stephen P. Lutar Jr. · ORCID 0009-0001-0110-4173 · Dataset license: Apache-2.0; the cited program publication is CC-BY-4.0.
Full DOI-pinned lineage (v1→v26) + the 8 papers: szl-papers PAPERS_INDEX.
No artifact-specific DOI is minted for this dataset; the concept DOI above covers the program.
Honesty (Doctrine v11): Λ unconditional uniqueness is Conjecture 1 (machine-checked FALSE as stated) — never a theorem; conditional uniqueness is Theorem U (axiom-free). Locked-proven formulas = exactly 8 {F1,F4,F7,F11,F12,F18,F19,F22}; ~185 experimental theorems are a separate CI-green tier; Khipu BFT safety = Conjecture 2. Trust never 100%.
@misc{lutar_szl_ouroboros,
author = {Lutar, Stephen P., Jr.},
title = {SZL Holdings --- The Ouroboros Thesis (Governed Post-Determinism)},
year = {2026},
publisher = {Zenodo},
doi = {10.5281/zenodo.19944926},
url = {https://doi.org/10.5281/zenodo.19944926},
note = {Concept DOI --- always resolves to the latest version. ORCID 0009-0001-0110-4173. CC-BY-4.0.}
}
Signed-off-by: Stephen Lutar stephenlutar2@gmail.com
◇ Explore the SZL Holdings estate
▶ a11oy console (a-11-oy.com) · a11oy Space · killinchu · holographic (3D) · all datasets & models → SZLHOLDINGS · GitHub org
🛡️ SZLHOLDINGS on Hugging Face → · a-11-oy.com → · Estate hub — live →
Governed AI you can prove.
SLSA: L1 honest · L2 attested · L3 roadmap. Λ = Conjecture 1 (advisory, never a theorem). Trust ceiling 0.97 — never 100%. Labels honest by default: MEASURED / REPORTED / MODELED / HEURISTIC / UNKNOWN / UNAVAILABLE. locked-proven = exactly 8 {F1,F4,F7,F11,F12,F18,F19,F22}.
Load the tabular corpus
The corpus config exposes the 762 JSONL chunks and excludes provenance JSON plus binary FAISS indexes from schema inference. FAISS files are separate retrieval artifacts and must be paired with their ID maps and manifest at a pinned revision.
Apache-2.0 metadata applies to repository-authored packaging and code. Source documents and quoted material retain their original rights; public availability is not permission to train on or redistribute them without verifying those rights.
- Downloads last month
- 259