Continuity, Successor Readiness, and Documentary Repair in a Longitudinal Human–AI Project

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

What does a successful handover or local test establish about the continuity of a human–AI project, and what remains to be established before subsequent action is warranted? This article examines selected episodes from Dufour 24, a project combining the technical reconstruction of a sailboat with the development of a documentary architecture for sustained human–AI work. It is a retrospective, within-case analysis informed by design science, not an independent replication of the project's earlier architectural preprint. Three evidence sets are examined: early continuity and the C03→C04 handover; successor preparation and corrective execution in C07→C08; and an operational glossary pilot with subsequent versioning and documentary repair. The records distinguish preserved content, situated interpretation, authorization, execution, and the production of project-recognized effects. They document continuation after handover, source-based responses to hypothetical tasks, and localized recovery that preserved adverse results rather than retrospectively declaring the original attempts conforming. The glossary provides a complementary boundary case: technical completion and fulfilment of a test criterion do not necessarily resolve every evidential dependency, and evidence from one data–program pair does not automatically establish the operation of its successor. A later, unauthorized report also misdescribed the authority for its own production; its correction addressed both the action's record and the attribution embedded in the narrative. The contribution is a traceable account of these relationships and their limits. The evidence is internal, selectively recovered, and partly mediated by contemporaneous reports. It does not establish comparative superiority, net workload reduction, general transferability, or persistent behavioral compliance.