# Public release assessment

**Preparation-stage assessment.** The later [agent-first launch decision](launch.md) supersedes pending publication and license choices below; scientific limits remain.

Assessment date: 25 September 2026. Scope: the curated research-prototype export,
not the entire private repository, a live Commons service, or a journal submission.

## Recommendation

**Yes: prepare a public, explicitly provisional protocol seed.** The appropriate
claim is that the packet offers a small interface, conditional models, runnable
synthetic examples and documented objections for others to test. It does not need
to solve the operation of a real review institution before people can discuss and
fork that design. It does need to describe its evidence accurately and give a
reader something useful to do.

The initial release should pair an accessible introduction with a versioned,
downloadable prototype and domain clinics. Public criticism is part of the
research objective. A security badge, operational certification, automatic
publication service, or unsupported efficacy claim would imply evidence this
packet does not have.

This is an editorial and technical readiness assessment. Applicable rights,
institutional obligations and legal duties depend on the actual rights holders,
operator, jurisdiction and service. The law examples cannot supply that authority.

## What is being released

| Surface | Intended public meaning | Excluded implication |
|---|---|---|
| Introductory blog | Invitation to examine and try a proposal | Scientific acceptance or institution endorsement |
| Downloadable source and toy agents | Reproduce specified synthetic behaviors | Authenticated or production-safe service |
| Mathematical manuscript | Conditional propositions, derivations and counterexamples | Measured behavioral law or whole-network equilibrium |
| Domain cases | Worked synthetic exercises and domain questions | Real clinical, legal, policy or astronomy result |
| Blue/red review record | Documented internal challenge and repair | External human peer review |
| Human onboarding | Bounded contribution exercise | Enrolment in a professional credentialing system |

## Release decisions that remain concrete and small

The user has approved merging the prior papers and requested this public-readiness
assessment and release preparation. That does not identify a public destination,
grant a license over all inherited material, or approve a particular rendered
artifact. Prepare the exact candidate first; the final release decision should
name these items together:

1. **Responsible attribution.** Recommended: identify the responsible human
   maintainer and describe AI drafting, implementation and adversarial review
   separately. Confirm the byline and any affiliation. Do not assign coauthorship
   to a person from their involvement in discussion alone.
2. **Rights and licensing.** Recommended subject to rights-holder confirmation:
   CC BY 4.0 for newly prepared prose/figures and MIT for original prototype code.
   Apply licenses only to the approved file inventory and retain third-party
   notices. Do not describe this as already open source before adoption.
3. **Destination and exact artifact.** Recommended: an introduction on the
   existing outward-facing research site, plus a fresh public export repository
   or versioned download. Keep private Git history and unrelated working papers
   outside the public candidate. Bind approval to the source and exported digest.
4. **Intake responsibility.** Name a realistic public issue/feedback owner and an
   alternate; identify a private route for sensitive rights or conduct concerns.
   Publish only service commitments the operator can maintain.

CC BY permits reuse with attribution and change notices; it does not supply all
privacy or publicity permissions. MIT permits broad software reuse with retained
notices and a warranty disclaimer. These are recommendations, not a grant made by
this draft. [CC BY terms](https://creativecommons.org/licenses/by/4.0/),
[MIT terms](https://opensource.org/license/mit).

AI contribution also affects copyright analysis. The U.S. Copyright Office's
Part 2 report addresses copyrightability of generative-AI outputs; this packet
does not resolve which elements have protectable human authorship. A release
statement should accurately identify contributions and grant only rights the
licensor holds, without claiming an exclusive right over public-domain elements.
[Copyright Office materials](https://www.copyright.gov/ai/).

## Technical release evidence

The candidate must include a reproducible run, machine-readable outputs, source
hashes, local-link checks, a public-file inventory, and clear unknown states.
Readiness is evaluated against the final export, not a remembered test result
from an earlier branch. A clean scanner result is limited evidence from the
declared scan; it is not proof that every confidential or copyrighted item has
been detected.

Public export should exclude absolute local paths, private hostnames, user
communications, operational credentials, raw agent transcripts, and private
collaborator details. Research examples should use synthetic identities and
inputs. Only actually fetched public sources should be described as verified;
failed retrievals remain explicitly unverified.

An amendment invalidates only what it materially reaches under the declared
dependencies. Preserve unresolved dependencies as unknown. The toy framework's
enforcement comes from trusted, in-process fixtures; a successful demo does not
prove real identity, signature, isolation, legal authority, network delivery or
privacy enforcement.

## Publication route and timing

Launch the protocol conversation on surfaces that permit an honest prototype.
A human-owned project page and a versioned archive are plausible first surfaces.
Then invite bounded domain exercises and collect counterexamples. Decide on a
scholarly manuscript only after substantive human engagement with the claims.

Do not upload the present AI-heavy packet to OSF Preprints: its current policy
excludes content mostly or wholly generated by AI. A cosmetic rewrite or a new
human byline would not establish that a later manuscript is eligible.
[OSF moderation policy](https://help.osf.io/article/691-preprint-moderation-policies).

arXiv requires topical, refereeable contributions and a distribution license,
with registration, possible endorsement and moderation. It is a later manuscript
route, not a guaranteed launch endpoint. bioRxiv's stated scope centers on
research articles with new data; this general protocol proposal is not an obvious
fit. [arXiv submission guidance](https://info.arxiv.org/help/submit/index.html),
[bioRxiv screening](https://connect.biorxiv.org/news/2022/06/13/screening_procedures).

## Distinguish launch from service operation

The prototype can be public while the following remain research questions:
identity/common-control discovery, scalable arbitration, capacity measurement,
adversarial incentives, undisclosed dependencies, human comprehension and uptake.
Those limits must travel with the examples. Opening a live service that receives
third-party disputes or sensitive evidence is a different operational decision.

The recommended first public ask is deliberately bounded: **try one synthetic
claim, make one material change, and report one failure or unnecessary step**.
That tests whether the design deserves elaboration before asking communities to
transfer authority or scarce labor to it.