The proof stands in Alp—and it weighs several tons of spruce

Twenty-three students in the MPDA master program at UPC School Barcelona designed, structurally analyzed, digitally fabricated, and erected a segmented wooden shell pavilion in a single construction week in Alp, a mountain village in the Catalan Pyrenees. The result: an antifunicular rib structure of planar n-gon modules, CNC-milled at Tallfusta and Laserpenta, connected with Rothoblaas connectors, documented in the McNeel Blog on March 31, 2026. No generic tool made this possible. Three specialized software components, tightly interlocked, made it possible.

←TODAY: The Grasshopper → Karamba3D → CadWork stack is the de facto standard for parametric timber shell projects in Europe in 2026.
→3012: Practices that do not master this stack internally source their design logic—and in doing so, lose control of their geometry.
Fulcrum: Custom code is no longer a luxury; it is the place where design intent and fabrication still speak the same language.

That is the real signal behind this project—not the shell itself, remarkable as it is. But rather the fact that the entire workflow from form-finding to CNC output ran in a single parametric environment. Grasshopper as the design engine, Karamba3D (developed by Clemens Preisinger in Austria, taught at ETH Zurich and other European schools) for structural validation directly in the Grasshopper canvas, and CadWork—developed by cadwork informatik AG in Basel—for fabrication logic. No export. No media break. No ‘we’re sending this to the engineer now and waiting three weeks.’

Why buying or outsourcing would not have worked here

In six Swiss practices that PAZ accompanied in 2025, the most common answer to the question “Why do you not have custom code?” was either “too expensive” or “we looked for someone but did not find them.” Both answers are understandable. Both are also wrong.

The MPDA project shows the opposite: the students—not trained software developers—not only learned the stack in four months but applied it productively. The key skill was not programming. It was the understanding of what the tools can do and where design begins to become custom. A timber shell node that must be structurally optimized, geometrically rationalized, and output as CNC-ready cannot be solved with standard commands. Someone on the team must write the Grasshopper cluster that does it.

The alternative—hiring an external developer—has a known catch: the developer does not understand the geometry logic, and the architect cannot specify the handoff. The result is usually a tool that solves the first use case and gets thrown away on the second project.

Atelier: If your next competition project has a special structure—timber shell, membrane, parametric facade—identify now the one workflow step that no standard plugin covers. That is the place where custom code decides the competition. Not as a gimmick, but as a design tool.

What this means for practices that do not have UPC students

Honesty first: the MPDA model—an entire master program as a built-project pipeline—is not transferable to a normal architecture practice. But the core is. The Grasshopper → Karamba3D → CadWork stack is available, documented, and directly applicable in Switzerland because CadWork is already standard in Swiss timber construction practices. Every timber engineer between Basel and St. Gallen knows Rothoblaas connectors.

What is missing is not the tool. What is missing is the internal competence to script the transition between these tools—the 50 lines of Python or GHPython that translate the Karamba result into CadWork attributes. That is the build-versus-buy decision that every practice must make in 2026: not “do we buy a plugin” or “do we hire a developer,” but “who at our practice learns to write this handoff.”

The answer from six Swiss practices that PAZ accompanied in 2025: in no single case was a dedicated software developer the solution. In five of six cases, it was an architect or engineer who had invested three months in Grasshopper scripting—and afterward became the quiet multiplier on the team.

Start this week. Open the Karamba3D documentation, watch the MPDA project credits video on the McNeel Blog, and ask internally: who at our practice wants to solve the next workflow knot—instead of waiting for someone else to solve it.

Source: McNeel Blog (Rhino)