AiUnifiedProcess
AI Unified Process (AIUP) is a requirements-driven software development methodology that treats a specification, rather than the code, as the permanent source of truth for a system. Its primary artifact is a set of use cases describing what a system must accomplish, independent of implementation detail; AI is then used to generate code, tests, and documentation from those use cases, and to regenerate them as the specification evolves.
This inverts the pattern used by most other spec-driven AI coding approaches, where a specification is effectively a detailed one-time prompt: useful for guiding a single generation pass, but discarded once the resulting code exists, at which point the code becomes authoritative again and the spec decays. Under AIUP, use cases, tests, and generated code are meant to improve together across short iterative cycles for the lifetime of the system, with tests specifically responsible for ensuring behavior stays consistent as code is regenerated.
AIUP defines four agile-style phases and supports both greenfield and brownfield workflows. Its prescribed tooling is Claude Code together with a set of open-source plugins tying the methodology to specific frameworks -- Vaadin, Angular, and JPA -- plus an IntelliJ plugin that links specifications to tests.
AIUP is placed in Assess. It occupies the same problem space as spec-kit, OpenSpec, and Verified Spec-Driven Development, all of which are also Assess-ring specification-first methodologies on this radar. AIUP's distinguishing claim -- that the spec, not the code, should remain the long-term source of truth -- is a genuinely different position from tools that treat a spec as a disposable prompt, but its concrete tooling is narrowly scoped to a Java/Vaadin/Angular stack, which limits how broadly it generalizes compared to the other entries in this space. No first-person use has been confirmed.
