deletetd

This release provides a minimal, self-contained reference implementation of an audit primitive for autonomous system accidents. The core contribution is not a model or a predictor, but a structural demonstration of how certain audit questions become unanswerable after failure, depending on internal update mechanisms. What this release contains Core figures (Fig.1–4) visualizing: Outcome, expectation, and event dynamics Irreversible internal updates ("scar") Resulting audit completeness Narrative-level summary metrics A formal minimal specification (spec/audit_minimal_spec.md) A single-file reference implementation reproducing the figures What this release does NOT claim This is not a production system This is not a full world model or controller This does not explain why an accident happened Instead, it demonstrates which questions cannot be answered after the fact, even with identical inputs and logs, due to internal update structure. Intended use Conceptual reference for accident audit and accountability Reproducible minimal example for research and regulatory discussion If you use or reference this work, please cite the repository URL.