Operational and Scholarly AI Failure Mechanisms
Examines published AI and systems research to distinguish seventeen failure-mechanism groupings and consolidate them into thirteen operational families. Yacht Provenance-specific scenarios and counterfactual tests explain why the distinctions matter.
Read PDF →Why Single-Chat AI Research Fails to Produce Comprehensive Yacht Histories
Examines the unsuccessful comprehensive yacht-history commissions for Cangarda and Moonbeam IV and the subsequent Ticonderoga institutional-learning experiment. The manuscript analyzes failures of cumulative research and manuscript continuity, assesses the thirteen operational AI failure families against the observed records, and explains the case for a persistent, database-centered evidence architecture with independent review. It distinguishes demonstrated experimental findings from proposed capabilities still requiring end-to-end validation.
Read PDF →The Governing Principles and Specifications of Yacht Provenance
Explains how Yacht Provenance translates evidence-led historical research principles into engineering requirements, architectural constraints, alternative implementation mechanisms, verification evidence, and acknowledged limitations. It examines physical-vessel identity, documentary provenance, independent review, correction, and research continuity, distinguishing requirements adequacy from implementation conformance.
Read PDF →
An Eight-Node Information Architecture for Yacht History and Provenance Research
Explains the eight conceptual nodes—Entity, Case, Matter, Source, Evidence Item, Assertion, Assessment, and Conclusion—and how they preserve documentary genealogy, conflicting claims, historical identity, uncertainty, and accountable revisions. It distinguishes conceptual examination from implementation validation.
Read PDF →
Failure-Informed Architecture for Reversible AI-Assisted Research
Describes a reversible Discovery–Resolution–Completion Assessment cycle, continuing contextual challenge, isolated two-phase independent examination, and correction and recovery controls. It maps the thirteen failure families to design considerations and test scenarios.
Read PDF →USS America — The Making, Amendment and Preservation of a Constructed Historical Corpus
The USS America corpus is a wholly synthetic historical research collection developed specifically for testing and validating the Yacht Provenance research engine. Its vessel history, documentary records, illustrations, historical assertions, and associated materials were fabricated to create controlled challenges for evidence intake, source identification, investigative operations, contradictory records, historical interpretation, and research outputs. The constructed material is not presented as an authentic historical record of an actual yacht.
This 705-page publication documents the construction, amendment, preservation, and evaluation of the corpus. It brings together the constructed history, source descriptions, test questions, reference answers, evaluation criteria, and grading framework used to examine research-engine behavior. The complete PDF is available below, in the same manner as the other technical publications. A digital copy of the USS America Corpus, including its historical data, questions, answers, and evaluation materials, may also be requested at inquiries@yachtprovenance.com.
Read PDF →