The Principle
AI performs at full capability when given a complete, typed, declared brief.
Not guardrails. Not constraints. A clean input that matches the capability brought to it. Semantic Intent is a methodology practice built on that principle — designed for organisations that want to use AI seriously, not just to write code faster.
Project Phoenix is how we apply it to legacy systems. RECALL is how we build for what comes next.
Built from Production, Not Theory
Project Phoenix wasn't designed in a lab. It was forged in production — from years of building, breaking, and rebuilding enterprise systems across dozens of organizations.
The methodology emerged from a simple observation that the legacy modernization industry has spent decades avoiding: the code is not the asset. Every failed migration, every stalled transformation, every $200M consulting engagement that delivered a modern wrapper around old complexity — they all failed for the same reason. They tried to preserve the implementation when they should have been preserving the intent.
AI changed the equation. Not by translating old code faster — but by making it possible to read an entire legacy codebase and extract what it actually does. The business rules. The workflows. The decision trees. The edge cases. The things that matter.
Once you can extract the intent, you don't need the old code anymore.
That's Phoenix. Extract. Rebuild. Validate. Move on.
Semantic Intent
Phoenix is built on the principle of Semantic Intent — the idea that observable state and behavioral purpose must be unified, not separated. When implementation drifts from intent, systems fail. When you anchor on intent and let implementation follow, systems succeed.
This principle governs every stage of the Phoenix pipeline. Agents don't translate code. They extract intent. The rebuilt system isn't a copy of the old one. It's a faithful implementation of what the old one was supposed to be.
“One is archaeology of structure. The other is archaeology of intent. That difference is everything.”
The Evidence
Production Validated
Applied to a 9,160-line Flutter mobile app — rebuilt as a typed React 18 codebase with 51 passing tests, A-06 certified, live in production at w2t.semanticintent.ai.
Industry Analyzed
UC-024: The Obsolescence Cascade maps the market forces driving the need for Phoenix across five dimensions of enterprise impact.
Methodology Documented
The complete six-agent pipeline is published and peer-reviewable at semanticintent.dev/phoenix.