# Final content addendum: funded inspections and publication surfaces

25 September 2026. Reviewed `manuscripts/funded-audits.md`, `CLAIMS.md`, the
updated synthesis, `publication/site-integration.md` and
`publication/THIRD_PARTY_NOTICES.md`. No code or other packet source was changed
in this pass. The mathematical lane separately reviewed the finite paths; this
review concerns interpretation, authority, human claims and release readiness.

## Funded-audit interpretation

**The experiment closes the promised-versus-executed inspection gap within its
toy environment; it does not close the auditor-effort problem.** The manuscript
states that auditors perform scheduled work with stipulated detection behavior.
Its final boundary explicitly retains strategic auditor refusal/shirking, false
reports, dependence and appeal reversal as unmodeled. Neither the claim map nor
the synthesis upgrades funded execution into observed human diligence.

The eight output “actors” are now clearly eight task decision units, potentially
mapped to fewer execution people. Separate utilities, task-specific escrow and
pre-reserved capacity are premises, not evidence of eight independent reviewers.
The concealed marginal calculation needs the fixed population, information
timing and participating decision's independence from the subset; the prose
states those requirements. Disclosure and partial-refusal controls make their
importance visible. A future shared-person or coalition model cannot silently
inherit independent decision behavior from the labels.

Tokens, hours and utility effort are distinguished. The escrow transfer is
collectible only inside the fixture. It establishes no real asset, deposit policy,
consent, lawful fine or legitimate disciplinary process. The text explicitly
omits locked-collateral opportunity cost and nonlinear wealth effects. Therefore
it cannot show equitable participation across resource levels, and does not claim
to. Describing insufficient collateral as pre-offer unavailability is honest
within the model; it is not a recommendation to exclude actual poor contributors.

The shared-role reservation is deliberately conservative and its unused slack
is reported. Refusal/triage labor is zero only in this experiment and marked as
an omission. The coupled queue separately charges it; assembling both results
does not automatically create one integrated funded, deadline-aware system.
Rights, recusal, lawful recipient disclosure and remedies remain institutional
questions. No additional legal disclaimer or operating institution is needed
merely to publish this carefully limited synthetic experiment.

## Claims and synthesis

The claim map distinguishes prior work, design, conditional theory, implementation
and synthetic experiments, empirical evidence and open hypotheses. The relevant
mathematical manuscript now uses I for the simulated outcomes and D for the
proposed interface, consistent with that map. Human/field evidence remains none.
The funding claim is bounded to a fixed-population, fixed-information toy and
executing auditors. Internal review evidence is not presented as passed independent
human review or a submission-ready journal paper.

The synthesis preserves the explicit noncomposition table and calls this a
separate engine closing one narrow loop. It does not infer a whole-network
equilibrium, legally enforceable collateral or real inspector independence.
I find no new material overclaim in these revisions.

## Site integration and notices

The integration guide correctly separates an intentional Jekyll blog adaptation
from serving the already rendered static packet. The current blog source actually
contains the documented layout/disclosure fields and `published: false`. Relative
links require the selected public destination; copying front-matter-bearing
Markdown into a generator is not a guarantee of preserving bytes. `.nojekyll` is
limited to the proposed standalone export root, not the existing site's root.
The guide requires inspecting the actual rendered post and retrieving the delivered
archive, rather than equating a successful local build with publication.

The third-party notice is under `publication/`, not the packet root. It separates
the existing font license from the project's still-proposed original-material
licenses and does not claim an adopted open-source grant. I independently read
the current eight-page exported PDF's attachment inventory using pypdf: it contains
`bitstream-vera-license.txt`, and that attachment exactly matches the exported
notice file. This verifies that specific notice-preservation claim for the
inspected artifact, not ownership of every original contribution or completeness
of a legal rights audit. Optional libraries are described as build dependencies,
not vendored implementations.

## Final position

**Qualified assent remains unchanged: publish the curated, candid research seed.**
No new legal, economic-unit, human-efficacy or publication-framing blocker was
found. The funded-audit addition makes an earlier assumption testable without
pretending all institutions and incentives are solved. Responsible attribution,
rights disposition, destination and final exact-artifact approval remain concrete
publication decisions. The runtime lane's final frozen archive and clean
reproduction checks remain separate from this content assent.