This is an independent research dissertation and accompanying software artefact. It has not been submitted for an academic degree or award.
Generative AI systems were used extensively during the work as a critical interlocutor, and that use included drafting and revising prose and code, critiquing arguments and formal claims, proposing tests and counterexamples, and analysing defects exposed during development and evaluation. The systems used were OpenAI Codex and Anthropic's Claude, in the Opus 4.8 and Sonnet 4.6 versions. The version control history records which contributed to each change.
The author directed the research, selected and verified the sources, determined the scope and methodology, made the substantive design decisions, reviewed and corrected generated material, and accepted or rejected every change. Responsibility for the published arguments, implementation, evaluation and conclusions remains with the author.
The executable specification, software tests, mutation operators, evaluation freezes, sealed corpora, amendment history and results provide the retained record against which the artefact's claims can be inspected.