OIA Application Matrix (v0) — feedback & refinement

[michaeloboyle]

What this is

A v0 OIA Application Matrix is now published on the committee dashboard:

https://agenticsorg.github.io/community-projects/oia-matrix.html

It maps each application to the OIA 10-layer stack (L0–L9) + 6 spans (security, sovereignty, auditability, energy, identity, provenance), with per-row evidence tiers, inline drill-downs, a gap-closing improvement-prompt generator, and a client-side repo "retest" scan. It is intended to structure the OIA intake discussion — a strawman for committee red-lining, not an official ranking.

This issue is the feedback channel

The page links here. Please use the comments to propose corrections and refinements. Specifically useful:

• Placement corrections — a layer/span assignment that's wrong or overstated (the page flags lean-agentic as one such case where code contradicted the description). Independent re-tagging by a second member is the cheapest way to move rows from desc to code ✓.
• The tagging rubric — a 1-page layer/span rubric with 1–2 anchor examples per layer, so classification is reproducible rather than one person's read.
• Weights — whether/how to add a composite score. Deliberately omitted in v0 because weighting the 12 dimensions is a committee decision, and a single number would launder that judgment.
• New submissions — repos that belong in the matrix. (Recent adds: proffesor-for-testing/agentic-qe, ohdearquant/lionagi.)
• Funnel wiring — how the matrix plugs into intake (metaharness score as the mechanical Stage-2a gate; layer/span footprint as Stage-2b).

Next steps (from the analysis)

1. Adopt the layer-footprint matrix as the standard submission artifact.
2. Wire metaharness score into intake Stage 2a.
3. Draft the 1-page layer/span tagging rubric (anchor examples).
4. Run the first live submission through both and compare the auto report card to a human read.