ArtiCAD: Articulated CAD Assembly Design via Multi-Agent Code Generation

ArtiCAD addresses a highly specific and difficult niche: moving from single-part CAD generation (which is already a crowded field with projects like Text2CAD) to multi-part articulated assemblies. The use of a multi-agent system (Design, Gen, Assembly, Review) is a logical extension of the current 'agentic' trend in software engineering applied to physical design. However, its defensibility is low (3/10) because it is 'training-free'—meaning its value lies entirely in the prompt engineering and the workflow orchestration rather than a proprietary model or dataset. While the 7 forks in 3 days suggest early academic/research interest, the lack of a deep technical moat makes it highly susceptible to displacement. The primary threat comes not just from frontier labs like OpenAI, but from CAD incumbents like Autodesk (Fusion 360) or PTC (Onshape), who are better positioned to integrate these agentic workflows directly into their proprietary kernels. Once a frontier lab improves the spatial reasoning of their base models, the 'Assembly' agent's logic becomes a commodity feature. The project is a valuable reference implementation for how agentic workflows can handle geometric constraints, but it lacks the data gravity or infrastructure lock-in required for long-term defensibility.