---
title: "DIASPORA internal adversarial review, 2026-09-30"
date: "2026-09-30"
language: "en"
document_role: "review"
visibility: "public"
status: "handler-review"
---

# Internal adversarial review

Reviewer: the same cold handler that built the slice. This is not an independent review. That limit is part of the finding.

1. **Is this more than a contact directory?** The model is. Skills, offers, needs, and explainable matches are first-class, and the tests reject a need-holder as a helper. The published population is still four organizations and two projects. It is a capability substrate with a very small graph, not yet a functioning network of people.

2. **Are non-working features being counted?** Search, match, validation, and contribution state are executed in `node scripts/validate-seed.js`. The map's tiles and Leaflet are optional and can fail. Metrics are computed from the seed, not typed as a slogan. communiti's counts are not shown as DIASPORA counts. No frozen edition is claimed.

3. **Are provenance claims inspectable?** Each public entity has source ids. The entity page lists them. The registry row for Corsica Diaspora points at the public API response read on 2026-09-30. The two Bastia addresses are both kept and labeled.

4. **Was personal data imported too aggressively?** No person entity exists. The Eysines municipal page's named contact, mobile number, personal email, and street were not copied. The association IBAN on the Corsica Diaspora website was not copied. The diffusible legal seat from the public registry was copied, because it is an organizational seat already published with `statut_diffusion: O`. Edmond Simeoni is named only as the founder named on that website, not as a reachable person. A legal review was not done.

5. **Does the benchmark compare unlike scopes?** It says so in the first lines and in the table. A same-day seed is not five years of network animation and not a ten-year professional network.

6. **Did speed create blocking debt?** The risky shortcut avoided was a graph database, accounts, and a port of the Suicide Corse renderer. The remaining shortcut is duplicated HTML navigation and a CDN map. Neither blocks a later static export or a later host. The seed and the pages can drift in wording; the counts cannot, because the pages compute them.

7. **Could this be simpler?** One page would be shorter and worse to link. The split used is the smaller one that still gives a stable entity URL, a match page, and a contribution page that does not submit.

8. **Was generic living-publication machinery rebuilt?** No renderer, freeze tool, or EPUB pipeline was copied. The triptych and the freeze rule are a short editorial note pointing at the existing documents.

9. **Is it useful on a phone?** The CSS is mobile-first, controls are at least 2.8rem, and the map has a list. No phone was held in this pass. Treat device use as unverified.

10. **Can another agent resume without private state?** Yes. Issue 96, this directory, and the reality-test journal are enough. No private agent memory is required. The handler's own review is marked as not independent, so a successor should not treat phase 1 as externally accepted.

## Judgment

The phase-1 repository slice meets the structural acceptance list in issue 96, with the coverage and publication gaps left open. It does not meet the stronger hypothesis of a useful worldwide directory. Result of the bootstrap: partial.