AEC Magazine on the Judgement Stack

The write-up

Martyn Day wrote a feature in AEC Magazine about Skills for Architects. It’s called “Dawn of the Judgement Stack,” and it runs the length of a cover essay. Read it — it’s the clearest outside account of what we’re building and why.

Day’s argument is one we’ve made here in narrower terms: the AI features arriving inside Revit, Rhino, Archicad and Bentley are real, useful, and coming to every customer at once. That’s exactly why they won’t be where advantage lives. He borrows Dustin Schafer’s phrase — the “judgement stack” — for the layer that does hold advantage: the accumulated reasoning that makes one practice design differently from another, using the same software and the same models.

Why the repo, specifically

Day could have written the thesis without us. What he needed was something to point at, and most firms talking about AI have nothing public to show. We do. Skills for Architects is an open, MIT-licensed reference architecture: 39 skills organized across the life of a project, plus the agents that orchestrate them, the shared rules that keep them consistent, and the hooks that check their output before it leaves.

His read of the structure is accurate, and it’s the part worth repeating: the individual skills matter less than the layering. Skills carry professional judgement. Agents coordinate. Rules enforce convention. Hooks govern. Each layer changes independently. That’s not an AI idea — it’s how software has been built for decades — and the contribution is showing it applies to architectural practice at all.

He also caught the thing we care about most: the intelligence stays readable. Most of it is plain language an architect can inspect, argue with, and correct — not source code only a developer can maintain. When Day describes our New York zoning skills, his point is that there’s nothing uniquely New York about the method. The same shape holds an occupancy calculation, an EPD comparison, London Green Belt policy, or Part L. Every practice already owns the knowledge. The work is making it reusable.

The part we’d underline

The essay’s spine is ownership, and it’s the right place to put the emphasis. If your firm’s judgement only runs inside someone else’s platform, you pay a toll every time you query intelligence you created. Day is blunt about it, and we agree: you don’t need to train a foundation model to escape that — most firms shouldn’t. You need to own the layer above it. The skills, the orchestration, the governance, and eventually the institutional memory the whole thing reasons over.

That’s the same case we’ve made in “How Architecture Firms Should Build Their AI Stack” and in “Architecture Firms Need an Agent Harness.” Seeing it laid out by someone who has watched this industry adopt technology for thirty years, and who arrived at it independently, is worth more than any release note we could write.

Where to start

The repo is public and free. Architecture Studio installs the whole catalog as one plugin, and if you’ve never opened a terminal, v1.3 ships /learn — a seven-module course that assumes you spent the last twenty years in Revit and none at a command line.

Thanks to Martyn Day and AEC Magazine for the read. The repo is here. The full feature is here.