A Process-and-Rules-Based Software Development Lifecycle
A Process-and-Rules-Based Software Development Lifecycle Engineering Software for Correctness, Human Effectiveness and Continuous Improvement This paper sets out a process-and-rules-based approach to software development. It comes from professional experience in systems design, and has been tested again through practical work with artificial intelligence. Business software is designed around the processes it performs, the rules that govern them and the outcomes it must produce. Languages, tools and methods deliver those requirements; they do not define them. Specification, implementation and independent verification are kept separate. Expected results are proven before the build. Each criterion is fixed together with the question it answers, and a checklist confirms before the build that each can be read only one way. Two testing regimes are used together: predetermined positive and negative testing from unit level to user acceptance, including efficiency, and testing against the complete operational population. One definition separates defects from changes. Quality is judged against eight measures, and finally by the business confirming that the system is fit for purpose. The paper also covers human interface engineering, the difference between education and training, and AI-assisted improvement under governed responsibility. The paper argues that the measure is the cost per accepted component, not the cost per line written, and that this cost is decided by the specification and the tests. It reports practice, one proposition and one case from current work. It does not claim that the approach’s effectiveness has been shown by controlled comparison. Google’s “The New SDLC With Vibe Coding” is recommended as further reading. Version 1.2: adds that the question must be as fixed as the answer, with a checklist run before the build to remove ambiguity; evidence from the METR trial and recent work on the verification burden; and cost per accepted component as the measure. The abstract is shortened. Author’s note: I am updating this not long after publishing the first version, simply because in practice I keep hitting road blocks, which I then invest time in addressing. If I hit another and define a solution, I will update again, as the process matters.
Authors
- michale alexander simpson (ORCID: https://orcid.org/0009-0008-0616-1637)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-20
- DOI
- https://doi.org/10.5281/zenodo.24385076
- Primary Topic
- Software Engineering Techniques and Practices
- Type
- article
- Field-Weighted Citation Impact
- 0.00