Skip to content

Index

Articles (17)

Field notes from the studio. AI tooling, AEC workflows, and what we learn shipping.

01

Claude Code Cheat Sheet for Architects and Designers

A practical Claude Code cheat sheet for architects: installation, essential commands, project setup, skills, and architecture workflow examples.

→
02

One Wall, Five Ways to Build It

We tested five ways for agents to build a brick wall in Rhino. Larger operations reduced exchanges; direct Python was cheapest with Sonnet in these runs.

→
03

Arch Studio 1.5 Improved Cold-Start Project Records

In 27 cold-start runs per arm, Arch Studio produced valid project records 55.6% of the time against vanilla Claude's 14.8%.

→
04

Arch Studio 1.5: The Four Layers a Firm Builds On

Arch Studio 1.5 is one release on two channels: the open-source plugin for Codex and Claude Code, and the hosted MCP inside Claude or ChatGPT. Four layers, 61 skills, Norma, and nothing to run but your assistant.

→
05

Architecture Studio Now Runs on Codex

Architecture Studio v1.4.2 brings all 46 open-source AEC skills to Codex, with native plugin packaging, portable workspaces, and the same project memory used in Claude Code.

→
06

AEC Magazine on the Judgement Stack

Martyn Day dedicated a feature in AEC Magazine to Skills for Architects, using it as the public reference architecture for what he calls the judgement stack — the layer a practice owns above the model.

→
07

Can Your Firm Train Its Own AI Model? Pre-Training, Fine-Tuning, and RAG, Explained

Pre-training and fine-tuning are both called training — one takes a frontier lab, the other runs on a Mac mini. Firms announcing they'll train their own model are usually describing the one they can't afford.

→
08

Architecture Studio 1.4: The Skills You Write Yourself

283 stars, 46 skills, and a breaking migration — v1.4 turns the plugin into a framework your firm extends, with a studio workspace, firm-owned skills, and project records that outlive every update.

→
09

Architecture Studio 1.3: A Course in the Terminal

255 stars, 40 skills, and a new kind of skill entirely — v1.3 ships /learn, a six-module hands-on course that teaches architects Claude Code itself, on a fictional Brooklyn art museum project.

→
10

Architecture Firms Need an Agent Harness

AI agents for architecture firms need more than a chat window. Vercel's eve gives teams a durable runtime for tools, subagents, approvals, channels, and memory, with Architecture Studio and Norma as the domain and material layers.

→
11

Introducing Norma

Norma is an AI material library and product-memory layer for architecture and design studios, turning messy schedules, quotes, specs, and product research into private, cited, reusable studio memory.

→
12

Architecture Studio at Three Months: Your Project Gets a Memory

179 stars, 39 skills, seven releases — and the biggest change isn't a skill at all. Architecture Studio v1.2 gives every project a dossier and a decision log, as plain files in the project folder.

→
13

How Architecture Firms Should Build Their AI Stack

A practical framework for architecture firms building an AI stack: model access, agent harnesses, domain agents, project memory, material libraries, governance, and review.

→
14

Agentic AI in AEC: What Actually Changes for Practice

A practical view of agentic AI in AEC: what changes inside architecture firms, what should stay human-led, and how agent harnesses, project memory, and material libraries fit together.

→
15

Architecture Studio's First Month: 100+ Stars on GitHub

Architecture Studio hit 100 GitHub stars in its first month. Here's what the numbers say about how architects are adopting AI tools — and what surprised us.

→
16

Introducing Architecture Studio

Architecture Studio is an open-source AI agent system for architects and interior designers — Claude workflows for site research, zoning, specifications, product research, FF&E schedules, and sustainability.

→
17

Most AEC AI Apps Are Just Claude Wrappers

A practical test for firms evaluating AI tools: what proprietary data does it hold, what does it actually compute, and what would it cost to build yourself?

→