Documentation Index
Fetch the complete documentation index at: https://docs.svrnos.com/llms.txt
Use this file to discover all available pages before exploring further.
The 7-Layer Model paper is in active drafting. This page is a placeholder. Tutorial and full reference will land in early June 2026.
| Layer | Name | What lives here |
|---|---|---|
| L7 | Application | What the AI is being used for in a product |
| L6 | Policy | The rules governing what the system should and shouldn’t do |
| L5 | Session / State | The model’s representation of the user, case, history |
| L4 | Evidence Transport | Audit logs, attestations, evidence pipelines (NCSA territory) |
| L3 | Routing / Boundary | What enters and leaves the system; provenance gates |
| L2 | Tool / Component | What capabilities the model can invoke |
| L1 | Infrastructure | The physical compute, training pipeline, model weights |
Why a 7-Layer Model
The AI safety field has organizational frameworks (NIST AI RMF, FERZ Governance Laundering), interaction frameworks (EA Forum), and security frameworks (OWASP, MITRE ATLAS). None map the technical stack cleanly. The 7-Layer Model fills that gap, the same way OSI did for networking — a shared vocabulary across engineering, policy, and audit.Cross-reference to GER
Each GER code carries an optionallayer tag indicating where in the 7-Layer stack that failure structurally lives. For example:
- GER-501 Escalation Not Implemented → L6 (policy never wired to L4 evidence)
- GER-322 Provenance Omission → L3 (boundary lacks provenance gate)
- GER-318 Representation-Bound Admissibility → L5 (session state stale)
Paper
Full paper publishing in early June 2026. Until then, the working draft references:- Citation pairing with Stelling et al. (arXiv:2512.01166v4) and Generation Gap establishes the empirical/documentary foundation
- Inter-Layer Contracts framework credited to Wolfgang Roesch
- DecisionSpace OS lineage-table framing