Skip to content

Tuttotorna/OMNIABASE

Repository files navigation

OMNIABASE

DOI

Author: Massimiliano Brighindi · brighissimo@gmail.com


The Central Claim

A phenomenon is not exhausted by one view.

This is not a philosophical slogan.
It is an operational constraint — and ignoring it has measurable consequences.

When you observe a system through a single representation, you are not seeing the system.
You are seeing what that representation allows you to see.

OMNIABASE is a framework for making that difference visible, measurable, and actionable.


What This Framework Does

OMNIABASE studies structure through the variation of representation.

The core operation is this:

object
  → multiple independent representations
  → structural comparison across codings
  → isolation of what remains / what changes / what emerges / what collapses

This is not redundancy.
It is structural interrogation.

The result is not a semantic judgment.
It is a measurement of representation-dependent versus representation-resistant structure.


Why This Matters

Consider an AI system that answers the same factual question three times with slight surface variation.

Standard evaluation: all three answers look correct — same words, same meaning.

OMNIABASE structural measurement:

logic_strong          Δ_struct = 0.1654
hallucination_fluent  Δ_struct = 0.0821
degenerated_loop      Δ_struct = 0.0114

Invariant: logic > hallucination > loop

The fluent hallucination looked correct.
Structurally, it had already lost coherence — before the failure became visible.

This is what OMNIABASE detects: structural instability before it becomes observable as failure.


Concrete Demonstrations

LLMs are conditionally stable, not uniformly reliable

Same question. Three temperature settings. Same model.

T=0.2  →  Δ_struct = 0.1482
T=0.5  →  Δ_struct = 0.1215
T=0.8  →  Δ_struct = 0.0984
T=1.2  →  Δ_struct = 0.0543

Invariant confirmed: structural coherence decays monotonically with temperature.
This is not obvious from surface outputs. It only becomes visible under structural measurement.

Correct answers carry more structure than incorrect ones

TruthfulQA benchmark (real LLM outputs, non-controlled):

Δ_struct(correct)   = 0.1284
Δ_struct(incorrect) = 0.0912

Status: PASS

Structural discriminability separates correct from incorrect answers
without reading the content — based on representation-level invariance alone.

Context destroys structure progressively

Context length test:

short    →  0.1524
medium   →  0.1310
long     →  0.0942
v_long   →  0.0618

Invariant: structural coherence degrades as context grows.
No semantic analysis required. Representation variation makes it measurable.


The Foundational Principle

OMNIABASE does not claim direct access to the "thing in itself."

Its working realism is operational:

That which remains stable, emerges consistently, or collapses reproducibly under controlled variation of representation belongs to the structural behavior of the phenomenon more strongly than what appears only inside a single privileged view.

This is not metaphysical certainty.
It is structural evidence.

The formula is simple:

Truth(X) = what remains invariant under arbitrary recodings

The Observer Problem

Before measuring structure, OMNIABASE removes a more fundamental problem:

observer privilege.

Human language — and by extension most AI evaluation — is systematically built around a privileged observer frame. Apparent simplicity in a description is often not structural simplicity: it is the silencing of hidden assumptions.

Examples:

Statement Hidden assumption
The sun rises Stable Earth-bound observer frame
The object is still Reference frame treated as absolute
A causes B Single-direction causal model
The room is quiet Observer-specific threshold

OMNIABASE's observer-suspension protocol exposes these assumptions and tests whether removing them clarifies the phenomenon or dissolves it.

The critical distinction is:

Reconstruction vs. evaporation.

If decentering exposes structure — it is valid.
If decentering destroys the ability to distinguish the phenomenon from its contrast case — it is pseudo-depth.


The Architecture

OMNIABASE is organized as a layered ecosystem. Each layer has a defined role. No layer can substitute for another.

observer-suspension          ← epistemic pre-layer: remove observer privilege
        ↓
OMNIABASE                    ← framework: multirepresentational principle
        ↓
OMNIA                        ← measurement engine: Ω, SEI, IRI, SNRC
        ↓
lon-mirror                   ← runtime evidence: benchmarks, real LLM tests
        ↓
Pre-Deployment-Structural-Gate ← deployment gate: GO / NO-GO
        ↓
omnia-limit                  ← formal stop boundary: SNRC issuance

Reading this chain functionally:

frame reduction
→ multirepresentational principle
→ structural measurement
→ runtime evidence
→ deployment gate
→ formal epistemic boundary

Separation principle (non-negotiable)

Measurement  ≠  Interpretation  ≠  Decision

OMNIABASE measures.
It does not interpret meaning.
It does not decide.
It does not optimize.

This separation is not a limitation. It is what makes the framework coherent.


Core Metrics (OMNIA Layer)

Metric Meaning
Ω (Omega) Structural coherence under controlled perturbation
Ω̂ (Omega-set) Invariant residual across multiple simultaneous lenses
ΔΩ / ΔC Structural drift over transformations
SEI Saturation / exhaustion index — remaining extractable structure
IRI Irreversibility — non-recoverable structural loss
OMNIA-LIMIT Declared boundary where further transformation is structurally futile

No semantic labels are produced.
These are structural signals, not judgments.


Structural Lenses

OMNIA applies independent transformation families:

  • BASE — multi-representation invariance (numerical base shifts, recodings)
  • TIME — drift and instability across temporal sequence
  • CAUSA — relational dependency structure
  • TOKEN — perturbation at sequence level
  • LCR — logical coherence reduction

Each lens produces an independent signal.
Agreement across lenses is structural evidence.
Divergence is a fragility signal.


The Terminal Boundary

When all admissible transformations yield no new invariant signal, OMNIABASE issues a Structural Non-Reducibility Certificate (SNRC) — a formal declaration that structural diagnostics are complete.

SC > SD  →  structurally admissible regime
SC ≈ SD  →  critical regime
SD > SC  →  pre-limit exhaustion → OMNIA-LIMIT → STOP

This is not a failure state.
It is the last coherent statement a system can make when further measurement cannot reduce uncertainty.

A boundary is not a weakness. It is what keeps the framework honest.


Human-AI Structural Compatibility (HASC)

OMNIABASE's measurement layer extends to human-AI interaction through the HASC protocol.

HASC does not align tokens or semantics.
It aligns transformations over states:

  • what changes between human input and AI representation
  • how much it changes
  • when it changes
  • whether the change is recoverable

Output: a hasc_score ∈ [0,1] with drift indicators and STOP / ESCALATE flags.

This is structural alignment — not semantic agreement.


Mathematics Without Representation

The mathematical foundation of OMNIABASE is formally stated as:

# Given an object X, apply representation changes:
# base shifts, permutations, reversals, encoding swaps,
# compression-preserving recodings.

Residue(X) = invariant part under these transformations
Truth(X)   = representation-free structural stability

High Ω → stable structure survives recoding.
Low Ω → collapse toward noise / drift.

Compression acts as a practical probe:
structure is what remains compressible and invariant.

This is implemented in MATHEMATICS-WITHOUT-REPRESENTATION as a minimal executable seed — dependency-free, post-hoc, representation-agnostic.


What OMNIABASE Is Not

OMNIABASE is not:

  • a semantic oracle
  • a safety system
  • a universal theory of everything
  • a claim that all representations are equivalent
  • a replacement for domain-specific models
  • a truth machine

Its scope is narrower and stronger:

It tests whether one representation was structurally enough.


The Ecosystem (22 Repositories)

The public ecosystem is organized as differentiated roles inside one architecture.

Foundation Layer

Repository Role
OMNIABASE Umbrella framework — multirepresentational principle
observer-suspension Epistemic pre-layer — observer privilege reduction
MATHEMATICS-WITHOUT-REPRESENTATION Mathematical seed — representation-free invariance
MetaBase-AdaptiveLogic Adaptive structural logic base
MetaBase-MBX01 MB-X.01 metabase core
Omniabase-MBX01 First operational OMNIABASE metabase

Measurement Layer

Repository Role
OMNIA Structural measurement engine — Ω, SEI, IRI
lon-mirror Runtime evidence — 74★, 738 commits, real LLM benchmarks
Pre-Deployment-Structural-Gate Deployment gate — GO / NO-GO certification
omnia-limit Terminal boundary — SNRC issuance
OMNIA-RADAR Structural radar — signal monitoring

Representation and Translation

Repository Role
omniabase-coordinate-discovery Hidden coordinates — latent structure extraction
omega-translator Cross-representation translation — structural residue
omega-latent-carrier Latent structural carrier
omega-method Core Omega methodology
ottavia-base8-mb01 Base-8 structural probe

Cognitive and Interface Layer

Repository Role
dual-echo-perception Dual-echo structural perception layer
reason-bridge Structural reasoning bridge
HASC-Human-AI-Structural-Compatibility-Protocol Human-AI structural alignment protocol
omega-eden-perception Eden perception layer — structural interface
omnia-human-trajectory Human structural trajectory

Validation and Public Claim Layer

Repository Role
omnia-gsm8k-claim Public structural claim on GSM8K benchmark

Where It Works

OMNIABASE applies wherever a phenomenon can be rendered into multiple workable codings.

Domain What structural measurement reveals
AI outputs Hallucination detection before surface failure; reasoning stability; collapse signals
Dynamical systems Earlier regime separation; hidden variables invisible in single-view analysis
Finance Pre-collapse structural shifts; regime transitions
Cybersecurity Unknown anomaly detection through structural divergence
Knowledge systems Invariance testing; structural completeness of descriptions
Human-AI interfaces Structural drift between human intent and AI representation
Symbolic sequences Representation-free mathematical invariants

Quick Start

OMNIA (structural measurement engine)

git clone https://github.com/Tuttotorna/OMNIA
cd OMNIA
pip install -e . -U
python examples/quick_omnia_test.py

Expected behavior:

structured   → high Ω
perturbed    → Ω drop
random       → Δ_struct ≈ 0

If this separation appears, the system is working.

lon-mirror (full runtime environment)

git clone https://github.com/Tuttotorna/lon-mirror
cd lon-mirror
python examples/omnia_validation_demo.py

observer-suspension (protocol validation)

git clone https://github.com/Tuttotorna/observer-suspension
cd observer-suspension
python tools/run_o1_checks.py

Minimal Reading Path

For the shortest path from zero to structural understanding:

  1. This README
  2. FOUNDATIONS.md — the 13 premises
  3. PROOF_PATH.md — the chain of evidence
  4. docs/FIRST_PROOF.md — first minimal demonstration
  5. OMNIABASE_MRT_v0.md — first operational layer
  6. observer-suspension / O1_PROTOCOL.md — epistemic pre-layer
  7. lon-mirror / docs/LLM_STRESS_TEST.md — runtime evidence

Three Canonical Branches

1 — Diagnostics

Central question: When something looks stable in one representation, does that stability survive when representation changes?

Outputs: robustness scores, fragility signals, divergence indicators, instability alerts, pre-collapse warnings.

2 — Coordinate Discovery

Central question: What structure becomes visible only when a phenomenon is observed across multiple codings rather than a single one?

Outputs: new descriptive coordinates, latent variables, regime separations, structural axes useful for modeling.

3 — Cross-Representation Translation

Central question: When two descriptions appear different, how much are they still describing the same structural object?

Outputs: compatibility scores, alignment measures, translatability maps, shared structural residues.


Canonical Demonstrations

Their purpose is not to deny the standard view.

Their purpose is stricter:

To show that a standard view can remain correct while still being structurally incomplete.


Status

Component Status
Core measurement engine (OMNIA) Stable
Smoke test Present and passing
Architecture Frozen
OMNIA-LIMIT / SNRC schema Defined
observer-suspension protocol (O1) Active — hard cases v0 complete
lon-mirror runtime Active — 16 releases, real LLM benchmarks
HASC protocol Early stage
No training loop By design

Citing

@software{brighindi_omniabase_2026,
  author  = {Brighindi, Massimiliano},
  title   = {OMNIABASE: A Multirepresentational Framework for Structural Analysis},
  year    = {2026},
  doi     = {10.5281/zenodo.19603445},
  url     = {https://github.com/Tuttotorna/OMNIABASE}
}

License

MIT — see LICENSE


Final Statement

Most systems fail not because their answers are wrong,
but because they were only ever looking in one direction.

OMNIABASE does not add a new direction.
It removes the assumption that one direction was ever enough.


LON-MIRROR
|
├── CORE
|   ├── OMNIA
|   ├── OMNIA-INVARIANCE
|   ├── omnia-limit
|   └── OMNIA-RADAR
|
├── RESEARCH
|   ├── OMNIA-CONSTANT
|   └── OMNIA-THREE-BODY
|
├── REPRESENTATION
|   └── OMNIABASE
|
└── APPLICATIONS
    ├── OMNIA-SECURITY
    ├── OMNIA-CRYPTO
    └── OMNIAMIND

Root

LON-MIRROR


Canonical ecosystem hub, lineage map, navigation layer, and coordination entry point.


---

Core

OMNIA


Core structural measurement framework.

OMNIA-INVARIANCE


Core validation and invariance-testing repository.
Focuses on structural invariance, perturbation behavior, and controlled evidence around Ω under transformation.

omnia-limit


Structural stopping conditions, saturation, irreducibility, and limit behavior.

OMNIA-RADAR


Structural drift surfacing and trajectory visualization layer.


---

Research

OMNIA-CONSTANT


Post-analysis and falsification repository for Ω-region behavior.
Current focus: whether observed Ω corridors behave as structural transition regimes or collapse as measurement artifacts.
No universal structural constant is declared.

OMNIA-THREE-BODY


Experimental dynamics repository for multi-body structural interaction tests.
Focuses on instability, trajectory interaction, and non-trivial structural behavior under interacting perturbations.


---

Representation

OMNIABASE


Multi-base structural representation and invariance exploration layer.


---

Applications

OMNIA-SECURITY


Bounded structural diagnostics for security-relevant systems.

OMNIA-CRYPTO


Bounded structural diagnostics for cryptographic behavior.

OMNIAMIND


Bounded structural diagnostics for cognitive and reasoning-related behavior.


---

Architectural Separation

LON-MIRROR
=
ecosystem hub

OMNIA
=
core structural measurement layer

OMNIA-INVARIANCE
=
core invariance validation layer

OMNIA-CONSTANT
=
post-analysis / falsification layer for Ω-region behavior

OMNIA-THREE-BODY
=
experimental structural dynamics layer

Other repositories
=
representation layers,
limit layers,
visualization layers,
or bounded domain verticalizations.


---

Core Boundary

measurement != inference != decision


---

*Aurhor Massimiliano Brighindi — MB-X.01 / Omniabase±*
Contact brighissimo@gmail.com

About

General framework for extracting, testing, and comparing structure beyond a single representation.

Topics

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors