We pointed our decomposition pipeline at an old trip-planner app as a test. It produced a full 3-layer platform — 11 domains, 28 tables, 4 packages, Storybook, catalog, tests. Six hours of AI work. He...
Our original roadmap had Personal Finance as the flagship proof. Peer feedback and a clear-eyed look at scope pushed us to a smaller, faster vehicle for proving the full generative cycle.
We claimed an AI could turn natural language into valid page compositions and that an orchestration engine could drive the loop. Then we built the test harness and ran it. Here are the results.
The Layers of Resolution architecture defines the what. Sinew is the how — an event-driven AI layer that watches for changes at any layer and propagates them to the layers above and below.
Testing a simple toggle revealed that components alone aren't enough. We need page recipes, a generative cycle, and three levels of Admin/User abstraction.
Our three-phase roadmap for JE-PFM — finishing the platform integration, building a cashflow engine, and creating a system where users speak new features into existence.
How we built a numbered SDLC pipeline in Obsidian, created a Claude Code skill to manage it, and turned Excalidraw drawings into actionable development plans.
The story of how we used domain-driven design, 135 Storybook stories, and AI-generated proof of concepts to build a personal finance platform.
A step-by-step guide to downloading OFX files from your bank and importing them into JE-PFM.
Learn how JE-PFM's budget tracking replaces complex spreadsheets with simple category-based budgets.