AI Artifact Inflation and Claim-Boundary Control v0.1.3: A Reflexive Case Study of Non-Expert AI-Assisted Research

AI Artifact Inflation and Claim-Boundary Control v0.1.3 is a public repository refresh of a reflexive case study on non-expert AI-assisted research, artifact inflation, epistemic grounding failure, and claim-boundary governance. The archive examines a limited first-person methodological question: how can AI-assisted research workflows rapidly generate manuscripts, repositories, figures, audits, and apparent scholarly structure before the underlying claims are sufficiently grounded by domain expertise, prior-art control, external data, empirical validation, or expert-review-worthiness? This release should not be read as a general proof that AI-assisted research is invalid, nor as a statistical study of AI-generated science. It is a bounded, self-auditing case study based on the author’s own research-production workflow. Its purpose is to document how artifact completeness can diverge from epistemic warrant, and how claim-boundary discipline can be used to reduce overclaiming. The case study emphasizes that polished outputs—papers, repositories, figures, manifests, and release packages—do not by themselves establish scientific validity. It distinguishes artifact production from epistemic grounding, and argues for explicit gates such as prior-art checks, external data requirements, expert-review-worthiness tests, negative-result preservation, and claim-boundary statements. This v0.1.3 release does not introduce new empirical findings or technical claims. Its main purpose is to reorganize the repository for public readability, strengthen claim-boundary statements, add a GitHub Pages landing structure, refresh metadata, regenerate manifests, and improve discoverability. This archive does not claim to evaluate all AI-assisted research, all generative AI systems, or all non-expert scientific work. It does not provide a benchmark of AI research quality, a policy recommendation, or a universal theory of AI-generated scholarship. It is best read as a reflexive methodological record and failure-boundary case study for responsible AI-assisted research practice.