Methodology · SSI v3.0 · verified 19 May 2026

Score the thesis, not the company.

Two rubrics. Two grammars. One hundred points each. The portrait is the readout.

Thesis I

Governed Agentic Ops

The deployment gateway for enterprise AI. Runtime governance, observability, evaluation, audit evidence, human oversight, policy enforcement. Eight dimensions. The visual language is geometric: orbital rings around an inner policy lattice.

Regulatory Embeddednessmax 20Orbital rings around the core. One ring per regulator, sandbox, or standards body the company touches. Ring radius scales with the score.

Runtime Governance Architecturemax 18The inner policy lattice. A high score draws a fine grid of cells; a low score draws a sparse one.

Regulatory-Technical Team Fitmax 15Node count on the central glyph. One node per credible institutional pedigree: ex-regulator, security lead, compliance ops.

Governance Build Velocitymax 12Lattice stroke weight. Heavier lines mean more governance features actually shipped.

Enterprise Buyer Tractionmax 12Vermillion anchor marks at lattice intersections. One anchor per named regulated buyer.

Technical Moatmax 10Lattice rotation angle. Zero degrees reads as a wrapper; up to 45 degrees reads as a differentiated runtime.

Capital Efficiencymax 8Negative-space ratio. A higher score leaves more whitespace around the structure, which reads as confidence.

Investor Signal Qualitymax 5Corner ticks at the canvas edges. One tick per mission-aligned investor on the cap table.

Total100

Thesis II

Vertical System-of-Record AI

Workflow gravity beats model novelty. AI that becomes, extends, or controls the regulated workflow's system of record. Eight dimensions. The visual language is organic: a faceted core, lateral arms, and root threads that descend into a system-of-record plate.

System-of-Record Integration Depthmax 20Root threads descending into a system-of-record plate at the canvas foot. More integration means more roots, reaching deeper.

Domain Data Advantagemax 18Spiral density around the core. A tight spiral reads as a compounding proprietary-data flywheel.

Team Domain Pedigreemax 15Core node faceting. One facet per founder or hire native to the regulated domain.

Workflow Lock-In Evidencemax 12Lateral arms reaching out from the core. One arm per integrated department or role.

Regulatory Alignmentmax 12Margin glyphs keyed to the catalyst regulation: EHDS, DORA, AMLA, MDR, or IVDR.

Switching Cost Architecturemax 10Anchor barbs on the root threads. The harder the company is to rip out, the more barbs.

Market Timingmax 8Asymmetric tilt of the core. A higher score leans the structure further forward.

Capital Efficiencymax 5Negative-space ratio, the same convention as the GAO grammar.

Total100

How a portrait moves.

The seed is sha256(slug + thesis + ssi_score), truncated to 32 bits. Same inputs always render the same portrait. When a new signal lands and the SSI score moves by even one point, the seed moves, the structure moves, and the portrait visibly shifts. This is the proof the methodology is operationally live.

When a portrait is muted.

A portrait renders in mute mode when the Falsifier Check returns ❌ Triggered or the Anti-thesis Filter returns Auto-pass. The vermillion is replaced with warm-grey, the pass reason appears in JetBrains Mono at the canvas foot. The portrait still exists. The methodology is honest about its no’s.

How does this stay live?

The Companies database in Notion is the only source of truth. When a row changes, Notion fires a signed webhook at /api/revalidate. The route verifies the signature against a shared secret, then marks the cached pages stale. The next page load refetches, the SSI score moves, the seed shifts, and the portrait visibly redraws.

Edit a row in Notion. The portrait moves. The data tells the story.