Encoding classical proportions into AI prompts
Journal note · 18 March 2026
Architects have always trusted proportions to do quiet, powerful work: a 1:1 courtyard that feels still, a 2:3 façade rhythm that calms the street, a 3:4 room that invites conversation. As AI tools start sketching with us, we want those same instincts embedded in the prompts we write and the concepts we share.
Our lab distilled centuries of observation on how light grazes a surface, how bodies move through space, how a sequence of rooms builds anticipation. When we ask AI to invent, we want to anchor it with those clues so the output is aligned with these principles.
What the ratios reveal
Looking across our knowledge base, patterns emerge: compact courtyards sit near golden rectangles, riverfront homes lean on thirds to choreograph views, Japanese interiors pulse with ken modules. These aren’t rules to copy blindly—they’re signals that tell us what kinds of harmony people respond to.
Listening to precedents at scale
When we feed those signals into our models, we’re essentially inviting Palladio, Nakamura, and Anupama Kundoo to whisper in the same ear. Instead of handing AI a blank page, we hand it a choir of voices and ask it to learn the common language. The result isn’t a formula; it’s a taste level.
Where this work is headed
We keep sketching the future, reminding ourselves that tomorrow’s buildings can be as generous, calm, and surprising as the ones we grew up admiring—and that AI can help us multiply those moments across the city.