Research-Stack/0-Core-Formalism/lean/Semantics/ExtensionScaffold

ExtensionScaffold

Pattern: Research extensions that extend the core Semantics library.

Rule: Core (Semantics/) remains tight. Extensions live here until proven essential.

Derivation Pipeline

Core (Semantics/)
├── Invariants          (surfaceInvariant, localExpansionInvariant)
├── Admissibility       (CausalStatus.admissible, SensorDetection.admissible)
└── Preservation/Collapse  (collapseProne, .collapsed variants across domains)

    ↓ (extends)

Extension Kernel
└── Temporal/TemporalVariantIndex.lean  (generic TVI framework — testing branch)

    ↓ (adapts)

Extension Adapters
└── Temporal/SpikeSync.lean             (neural spike network)
└── (planned: history/decoherence)

    ↗ (may promote if passes all 3 criteria)

Core (future — if irreducible, cross-domain, stable)

Purpose

This scaffold holds:

• Special regimes — Unusual physical conditions not yet canonical
• Experimental metaphors — Productive analogies being tested
• Domain-specific constructions — Vertical applications
• Candidate metrics — Measures not yet fully stabilized
• Rich examples — Patterns that may later be promoted into core

Structure

ExtensionScaffold/
├── Thermodynamics/         # Energy/entropy extensions
│   └── ThroatPhysics.lean  # Metastable wormhole maintenance
├── Topology/               # Geometric structure extensions
│   └── PlasmaTopology.lean # Regime classification for plasmas
├── Temporal/               # Time-variant systems
├── Decoherence/            # Branching histories alignment
├── Transmission/           # Horizon/boundary information flow
├── SemanticPrimes/         # Prime-based semantic encoding
├── Narrative/              # Graph OS ancestry structures
└── Dynamics/               # Special kernels, phonons, repair

Promotion Criteria

An extension moves to core only when it passes all three:

1. Irreducibility

You cannot explain the rest of the model without it.

2. Cross-domain reuse

It appears in multiple unrelated cases, not just one favorite example.

3. Stability under revision

It survives multiple rewrites without changing its meaning much.

---

Old criteria (deprecated): Required by bind, hardware target needs it, cost function depends on it, build fails without it.

Current Extensions

Extension	Category	Status	TVI Kernel
Temporal/TemporalVariantIndex.lean	Time-variant	Base kernel — generic TVI framework	—
Decoherence/HistoryTvi.lean	Branching histories	Active — decoherent history TVI adapter	✅ Uses kernel
Temporal/SpikeSync.lean	Time-variant	Active — neural spike network TVI adapter	✅ Uses kernel
Thermodynamics/ThroatPhysics.lean	Special regimes	Reference — metastable wormhole thermodynamics	⏸️ Standalone
Topology/PlasmaTopology.lean	Domain-specific	Reference — plasma regime classification	⏸️ Standalone

Architecture Boundaries

ENE (Endless Node Edges) — What It Is NOT

To keep the architecture clean, ENE does NOT:

Compute	Store
❌ Compute TVI	✅ Store TVI results
❌ Perform spike syncing	✅ Record sync outcomes
❌ Perform temporal alignment	✅ Document alignment
❌ Decide system actions (PTOS)	✅ Witness PTOS decisions
❌ Act as driver	✅ Maintain audit trail

ENE is the record, not the computation.

Stack Layering

┌─────────────────┐
│  Driver Layer   │  ExtensionScaffold/Temporal/SpikeSync.lean
│  (computes TVI, │  ExtensionScaffold/Decoherence/HistoryTvi.lean
│   sync, etc.)   │
└────────┬────────┘
         ↓
┌─────────────────┐
│      PTOS       │  Semantics/Prohibited.lean
│  (decides what  │  Semantics/Constitution.lean
│   to do)        │
└────────┬────────┘
         ↓
┌─────────────────┐
│      ENE        │  Semantics/Graph.lean
│  (records what  │  Semantics/Witness.lean
│   is true and   │  Semantics/Evolution.lean
│   how)          │
└─────────────────┘

One-Line Definition

ENE is a typed, witness-backed semantic graph with enforced admissibility, lawful traversal, and fully auditable evolution.

Verification Results

Core build: 43 jobs, zero warnings

TVI Kernel test: Two independent domains (history + spikes) successfully:

• Import TemporalVariantIndex kernel
• Map domain objects to TemporalProfile
• Compute TviVector via fromProfiles
• Check admissibility via admissible
• Expose same 4 failure axes: timing, rate, pattern, collapse

Kernel adapters:

• HistoryTvi.lean: 2 theorems + 4 #eval witnesses
• SpikeSync.lean: 1 theorem + 4 #eval witnesses

Planned Categories

Category	Scope	Candidate Content
Temporal/	Time-variant subclass systems	Temporal variants, clock domains
Decoherence/	Branching histories	Many-worlds alignment, collapse models
Transmission/	Horizon information flow	Boundary transmission, holographic principles
SemanticPrimes/	Prime-based encoding	Semantic primes, invariant roots
Narrative/	Graph OS structures	ENE layer, story graphs, ancestry
Dynamics/	Repair and phonons	Special kernels, phonon structures, repair dynamics

Rejected Approaches

File	Reason	Location
HumanManifold.lean	Fatally flawed approach	/data/archive/HumanManifold.lean.REJECTED_FATALLY_FLAWED

Usage

Extensions are not imported by Semantics.lean. They may:

• Import from core
• Be imported by other extensions
• Be promoted individually

Build

Extensions are not built by default:

# Build core only (default)
lake build

# Build with extensions (when needed)
lake build +ExtensionScaffold.Thermodynamics.ThroatPhysics

---

Note: Files here are preserved for reference. Promotion requires human sign-off per AGENTS.md.