Verification-Gated Human–AI Co-Development: A Methodology and Cost Analysis

Verification-Gated Human–AI Co-Development (VGC) is a methodology for high-assurance Human–AI software engineering. This research preprint introduces Verification-Gated Co-Development, formalizes six reusable collaboration primitives, presents a cross-representation parity-verification architecture, and reports an empirical case study together with quantitative software cost analysis. The work argues that the primary contribution is a reusable engineering process in which every AI-generated artifact is accepted only after passing an independent machine-verifiable gate. The methodology includes: • Cross-representation parity verification • Exact-anchor patching • Guarded irreversibility • Source-of-truth reconciliation • Read-before-write grounding • Honest backlog deferral The accompanying case study documents reproducible software engineering, Git provenance, reproducible publication, DOI minting, OpenTimestamps integration and scholarly artifact publication. Intended audience: Software Engineering Artificial Intelligence Human-AI Collaboration Empirical Software Engineering Programming Languages Formal Methods Knowledge Engineering