Agentic Agile-V framework aims to replace 'vibe coding' with verified engineering

A new arXiv paper by Christopher Koch, 'Agentic Agile-V: From Vibe Coding to Verified Engineering in Software and Hardware Development,' tackles the growing problem of unstructured AI-assisted coding. While agentic systems can inspect repos, run tests, and submit PRs, current evidence shows mixed results: productivity gains in controlled enterprise tasks but slowdowns in mature open-source projects, plus persistent failures in dependency handling and hardware verification. Koch argues the bottleneck is no longer prompt engineering but engineering process control.

To address this, Koch proposes Agentic Agile-V, a framework built on the Agile-V lifecycle and the SCOPE-V loop — Specify, Constrain, Orchestrate, Prove, Evolve, Verify. Key contributions include: (i) a taxonomy of minimum input artifacts for agentic software, firmware, and hardware work; (ii) a 'conversation-to-contract' gate that separates exploratory dialogue from formal implementation; (iii) risk-adaptive workflows for features, bug fixes, testing, and hardware; and (iv) an evidence-bundle acceptance model requiring independent verification and human approval. The paper concludes that agentic AI doesn't eliminate engineering discipline — it amplifies the need for requirements, traceability, and verification.