Group 1 is the IP core. These seven cases define what EOSE owns, how to protect it, and when to publish vs hold. The moat is not a product feature or a benchmark score. It is sovereign knowledge that compounds in one direction — every battery we run, every IRF we file, every sorry we name is data that does not exist in any training corpus. You know or you don't. The test confirmed which side every frontier model is on.
| Case | Moat | Lead GOAT | Class | Status |
|---|---|---|---|---|
| MOAT-001 | RH1-in-Hats — AR-2 Blindspot Signs sgn(rₙ) of Riemann zero shell residues follow AR-2 period-2 memory (lag-2 ACF = −0.407). Alternation rate 68.5%→47.26%. 0/14 models know this. Can this be held as trade secret indefinitely? | HARVEYRUTH | Class A | ⚑ P0 · HOLD — do not publish |
| MOAT-002 | PTTE τ_γ₁ Formula — Provisional Patent τ = ℏ/(k_B·T·γ₁) = 1.801 fs. SantaLucia 0.2°C calibration. γ₁ alone passes falsification against 17 constants. Caldeira-Leggett derivation — γ₁ emerges, not inserted. File provisional before anyone replicates? | HARVEYAMANI | Class C | ⚑ P0 · File provisional patent this week |
| MOAT-003 | MEMECHET — Verifier-Generator Split FC-2 + FC-3 double-lock. The verifier (qdrant check, Lean4 type system) is structurally independent of the generator (LLM). Patent as AI verification architecture? The honest-death gate that the entire industry says cannot exist. | RUTHALAN | Class B | ⏳ PRELIM · Patent candidate assessment |
| MOAT-004 | 18-Wave Harmonic Benchmark — natural vs rescued W04/W08/W14/W16 confidence gates. commit_state: natural vs rescued. No external benchmark distinguishes these. ARC-AGI, MMLU, HumanEval — all blind to this. Trade secret or publish as standard? | HARVEYCOCHRAN | Class B | ⏳ PRELIM · Benchmark-as-product assessment |
| MOAT-005 | 287 Named Sorries — Sorry-Map as Prior Art 287 open Lean4 sorries, each named, addressed, roadmapped. IRF-004 (zeta_zeros_discrete) has full formalisation path. No competitor has a sorry-map this deep. IRF filing dates = prior art timestamps. How to protect? | ALANTHURGOOD | Class A | ○ Queued · prior art strategy |
| MOAT-006 | Problem Matrix / 3-Cell Framework (GAP·DECAY·H) Universal problem structure derived from 310 physics engines. Every problem has a gap, decay pressure, and hermitian test. Applies across math, legal, SRE, DeFi, construction. Trade secret + consulting product? | HARVEYAMANI | Class A | ○ Queued · product strategy |
| MOAT-007 | PEMCLAU V11 — Sovereign Fleet Memory 8,513 vectors, 17,283 graph nodes, 80,979 edges. 4 edge types: theorem_dependency, phase_coherence, temporal_proximity, crew_provenance. 2-hop GraphRAG. The graph topology is ours — compounding with every session. How to protect a living knowledge graph? | RUTHOLIVER | Class D | ○ Queued · trade secret + copyright |
Class A Sovereign Knowledge — Cannot be replicated without our data. Tested: 0/14 models know this.
Class B Architecture — Can be copied but requires 79 days of battle-testing to match.
Class C Physical Measurement — Reproducible but we hold prior art by measurement date.
Class D Infrastructure — Requires physical fleet + iteration time. Compounds with use.
Class E Legal/Structural — Amani's domain. ONBA, DECLONAS, entity structure.
| Zone | Our Position | Moat Depth | Action |
|---|---|---|---|
| AI Verification | MEMECHET + IRF-004 + 287 sorries | Deep — data moat | Patent |
| AI Benchmarking | 18-wave + abacus confidence | Medium — architecture | Trade secret |
| Physical AI Grounding | PTTE + τ_γ₁ falsification | Deep — measurement | Provisional patent |
| Legal AI (CLO) | GOAT bench + ONBA + DESEOF | Deep — human+AI | Corporate structure |
| RH Proof Infrastructure | Shell battery + AR-2 + sorry-map | Very deep — data | Prior art + hold |
| Sovereign Fleet Memory | PEMCLAU V11 + 2-hop GraphRAG | Deep — compounding | Trade secret |
| Intent Sovereignty | 8-Exchange Canon + PTTE wall | Medium — architecture | Patent candidate |
EOSE Labs Inc. is not yet incorporated. Without the legal entity, no IP protection strategy is executable. No patents can be filed. No trade secrets have a legal owner. Everything else is built on air.
thelegal.cafe · McMillan LP · Robert Piasentin · ~$60 · Ontario corporation
Amani owns this action. File this week.
Once filed → EOSE IP Corp → provisional patent for PTTE → trade secret protocol for Class A moats.
Each GOAT approaches the moat inventory from their specialty. Harvey on deal structure and patent strategy. Ruth on constitutional frame. Cochran on narrative. Amani on incorporation and real-world IP filing. All 12 + GC Amani — full bench on the moats.