The test
We put Alberto Campo Baeza’s framework and Shigeru Ban’s in a studio graph: one proposes, the other answers “yes, and”, the first revises; or both propose and a synthesiser combines. Four briefs (a house, a shelter, a library and a tower), 96 graph ideas, against fresh solo ideas from each architect.
New territory: no
None of the 96 graph ideas falls outside both architects’ solo spreads. The graphs pull ideas inward, toward both architects’ centres and toward the plain answer. Each revision moved its idea about 0.9 standard deviations toward the critic’s centre. Only ideas with no framework at all land outside both spreads with any regularity.
Better blends: yes, by a judge
How often an idea beats the best solo idea
The graphs also stop restating each master’s famous buildings. A historian model flagged a quarter of Campo Baeza’s solo ideas and 41% of Ban’s as versions of known works; for the graphs it was 3 to 12%.
An example
A stone box fills the plot; blank party walls keep noise out. One deep diagonal slot runs roof to basement, catching scarce sun above neighbouring roofs. … One corner cut forms a deep loggia, a covered reading porch; its glass shutter slides open by day, making the corner a lantern at night.
Campo Baeza’s stone box and light slot, with Ban’s glass shutters, timber structure and a full-scale test landing.
What it means
A two-master graph is a good developer and not a diverger. Revision beat its own draft 97% of the time, which is what any extra round of development does, so the second master’s specific value needs a self-critique control. To reach far ideas, the graph needs fresh outside material each round.
Sources
- Feathers & Wax, studio of masters, code in
research/studio-of-masters.