threat-modeling.sgit.ai — a threat model is a claim you can check
The industry's own diagnosis is that threat modeling is a fairly subjective process producing static, siloed, context-poor documents. This site publishes two things instead: models built as machine-readable graph data, and — its actual differentiator — the audit that says which parts of a threat model turned out to be wrong.
Eleven linked threat models, live at ThreatModCon 2025 · one threat model validated line-by-line against its code the next day · seven white papers · why this site is not a neutral vantage point ↗.
Source: https://threat-modeling.sgit.ai/index.html↗ · site v0.1.0 · markdown twin of the front page.
The two strongest things this site has
Everyone publishes threat models. Almost nobody publishes the audit that says which parts of theirs turned out to be wrong.
- Eleven linked layers, live ↗ — "One model answers what could go wrong here. Eleven linked models answer what does this line of code put at risk." A published vault — Customer through Compute, 51 nodes, 179 threats — not a diagram of the idea.
- A claim, checked the next day ↗ — 16 March 2026: a threat model of a real flow. 17 March: every finding checked against the code. Two confirmed real, one confirmed already safe, two confirmed missing, two marked as proposals with nothing built yet — and all four kinds published.
The move, in two parts
The industry's diagnosis, in the founder's words: threat modeling is "a fairly subjective process" producing "static documents" that are "siloed", "fragmented" and "context-poor". The answer here is not another methodology.
- Make the model machine-readable — a semantic knowledge graph, not a Word document. Threat models as graph data → ↗
- Make the model falsifiable — then falsify it. A threat model was written 16 March 2026; a validation pass checked every finding against the code the next day. The validation pair → ↗
The policy position that follows: threat models as mandatory disclosures ↗ — security is a market for lemons, and a published, dated, checkable threat model is the correction.
Seven white papers, one argument
Written in a burst — five of the seven inside four days at the end of May 2025 — building from a diagnosis to a mechanism to a scaling case to a policy position. All seven, in order → ↗
The honest constraints
Most of the practice material is a security review of the founder's own live product. It names unmitigated weaknesses. The rule this site publishes under: publish the method always, publish findings only once they are closed or harmless by design, hold anything still open. The disclosure boundary, in full → ↗
The graph tooling is real but young. The ThreatModCon vault is live; the continuous-modelling pipeline the papers describe is a vision with partial implementation. What is real versus argued → ↗
This site is published by the estate whose own product it threat-models — a conflict of interest worth naming rather than discovered later. The participant disclosure, in full → ↗
Related
- https://sgit.ai↗ — the parent project and the network index
- https://sgit.ai/demos/vaults/threatmodcon-2025/↗ — the eleven-layer vault, live
- https://wardley-maps.sgit.ai↗ · https://graphs.sgit.ai ↗ · https://standards.sgit.ai↗
- https://risks.sgit.ai↗ · https://pki.sgit.ai↗