Research paper 13
Publishability Control Plane: Separating Local Proof, Evidence Acceptance, and Public Release
AI-assisted development often treats "done" as a single state. A change was made, tests passed, a draft exists, a receipt landed, or a branch was pushed. Research and product workflows need finer control. This paper proposes a publishability control plane for agentic work. The control plane separates local closure, proof observation, evidence acceptance, claim promotion, public-safe packaging, release approval, and production deployment. It is grounded in local Switchboard, Spinoffs, Caliper, historical Fractal Governance, and WLG workflows where artifacts can be valid at one boundary while blocked at another. A paper draft can be final locally and still not publication-ready. A public-record claim can pass source-unit review and still need quote checks and editorial approval. A branch can pass local proof and still need remote ref alignment. A receipt can land and still be held. The contribution is a small state machine and evidence vocabulary for publishability: every artifact should say what state it has reached, what proof supports that state, what stronger claims are blocked, and who or what can move it forward.
- Paper
- 13
- Authors
- A.G. Mauro and C.A. Harris
- Date
- 2026-07-19
- Collection
- Standing Framework Research
Abstract
AI-assisted development often treats "done" as a single state. A change was made, tests passed, a draft exists, a receipt landed, or a branch was pushed. Research and product workflows need finer control. This paper proposes a publishability control plane for agentic work. The control plane separates local closure, proof observation, evidence acceptance, claim promotion, public-safe packaging, release approval, and production deployment. It is grounded in local Switchboard, Spinoffs, Caliper, historical Fractal Governance, and WLG workflows where artifacts can be valid at one boundary while blocked at another. A paper draft can be final locally and still not publication-ready. A public-record claim can pass source-unit review and still need quote checks and editorial approval. A branch can pass local proof and still need remote ref alignment. A receipt can land and still be held. The contribution is a small state machine and evidence vocabulary for publishability: every artifact should say what state it has reached, what proof supports that state, what stronger claims are blocked, and who or what can move it forward.
← Back to research papers