UFT-ID 3.0

Unified Field Theory of Information Dynamics 3.0 is a constraint-governed research programme for information dynamics, admissibility, observation, transport, and deterministic recovery.

The repository does not present UFT-ID 3.0 as a confirmed fundamental physical theory. Formal mathematics, diagnostics, executable evidence, interpretation, speculation, and nonclaims have separate authority surfaces.

Core question

When an informational description changes, what changed?

UFT-ID distinguishes underlying state dynamics, constraint or recovery dynamics, transport between regimes, observation or coarse-graining, and boundary/source exchange. A decrease in one selected entropy is not automatically physical destruction of information.

Canonical abstract system

U = (S, A, F, Pi_lex, O, T, I, C)

• S: total state space
• A: admissible subset
• F: proposed evolution
• Pi_lex: deterministic recovery
• O: observation or coarse-graining map
• T: regime-transport map
• I: explicitly declared information functional
• C: constraint structure

No component becomes physical merely because it appears in the tuple.

Current theorem-level direction

No universal information-direction theorem exists without fixing the state
model, dynamics, information functional, observation map, reference measure,
partition/coarse-graining, boundaries, and source assumptions.

Within a declared class, monotonicity may be proved from actual dynamics. For example, finite deterministic processing satisfies H(f(X)) <= H(X), while a broader class containing stochastic mixing, permutations, and many-to-one maps admits positive, zero, and negative Shannon-entropy changes.

The generic information-balance expression remains a model template. It is not predictive physics until every term is independently operationalized and the model makes held-out predictions.

Executable finite results

theory/FINITE_RESULTS.md records the proved or counterexample surface. Executable witnesses include:

• two-state positive/zero/negative Shannon-change cases;
• one entropy-preserving fine trajectory with opposite observed signs under two coarse-grainings;
• proposal/recovery information decomposition;
• admissible recovery that increases a declared information functional;
• the bounded 2026 polygon multiplicity-extremum audit.

Machine metadata is in machine/finite_results.json.

Vopson audit programme

The public scholarly target corpus is under research/vopson/. It keeps published work, source claim, logical dependency, exact reproduction obligation, repository evidence, UFT-ID assessment, and claim class distinct.

Corpus inclusion is not endorsement. A dependency edge records reliance, not truth. The audit tracks mass-energy-information equivalence, SLI, genetics, gravity, symmetry, language diversity, simulation interpretations, errata, and published responses.

Validation quick start

Supported CI runtimes are Python 3.12 and 3.13 on ubuntu-24.04.

python -m compileall -q experiments scripts tests
python scripts/render_vopson_docs.py --check
python scripts/validate_vopson_corpus.py
python scripts/validate_reproducibility.py
python -m unittest discover -s tests -v
python -O -m unittest discover -s tests -v
python experiments/run_pr2.py --json

See docs/REPRODUCIBILITY.md for the evidence-chain contract and docs/MILESTONES.md for the evidence-gated project sequence.

Repository map

.
├── README.md / README4AI.md / AGENTS.md
├── MATHS.md
├── ROADMAP.md
├── docs/
│   ├── ARCHITECTURE.md
│   ├── CLAIMS.md
│   ├── CORPUS.md
│   ├── MILESTONES.md
│   ├── NONCLAIMS.md
│   └── REPRODUCIBILITY.md
├── theory/
│   ├── DEFINITIONS.md
│   ├── THEOREM_TARGETS.md
│   └── FINITE_RESULTS.md
├── experiments/
│   ├── lib/
│   ├── counterexamples/
│   ├── representation/
│   ├── reproduction/
│   └── run_pr2.py
├── research/
│   ├── reports/
│   ├── vopson/
│   ├── RESEARCH_GAPS.md
│   └── VOPSON_MATRIX.md
├── scripts/
│   ├── render_vopson_docs.py
│   ├── validate_reproducibility.py
│   └── validate_vopson_corpus.py
├── machine/
│   ├── contract.json
│   └── finite_results.json
├── tests/
└── .github/workflows/

Evidence and CI

Workflows use read-only permissions, a fixed runner, a Python-version matrix, full-SHA action pins, compilation, normal and optimized tests, and retained JSON evidence artifacts. Scientific invariants use explicit exceptions rather than ordinary assert statements.

Epistemic layers

1. Formal Core: definitions, proofs, counterexamples.
2. Diagnostic: audit and transport constructs.
3. Empirical: data, reproductions, simulations, experiments.
4. Interpretive: domain mappings and explanatory proposals.
5. Speculative: hypotheses not established by the prior layers.

Promotion requires evidence appropriate to the target layer.

Formal verification

Lean remains deferred until notation, theorem statements, and canonical counterexamples survive source reproduction and adversarial review. Lean can verify deductions from assumptions; it cannot establish the physical truth of those assumptions.

Design rule

A model may be useful without being ontologically true.

representation != referent
simulation != proof
numerical agreement != physical validation
cross-domain analogy != shared mechanism
self-consistency != truth

Status

The project is in the reproducibility and source-fidelity phase. The next scientific target is an exact reconstruction of the 2019 mass-energy-information calculation, separating reproduced arithmetic from its additional physical premises.

License

Software and repository documentation are MIT-licensed unless a file states otherwise. Cited papers and datasets retain their original licences.