TALAMANA · THE AI LITERACY MAP FOR ARCHITECTURE AND DESIGN · The Brief · AGE 16—18 · POSITIONAL · HELD
Style words and substance words
Style words return a house, and substance words return something much closer to your house.
Our position
"Contemporary, warm, natural light" are style words. They name a preference and leave every decision to the model. "Single-storey courtyard house, board-marked concrete, north skylight, compressed entry through a low corridor" are substance words. They name spaces, materials, directions and sequences — decisions. Style words return a house. Substance words return something much closer to your house. We hold that most weak briefs fail here, and that the fix is more programme, not more adjectives.
Why we hold it
A model has no way to know what "warm" means to you. It has a learned tendency for what images described as "warm" tend to look like, and it follows that tendency wherever you leave the decision open. Substance words are the parts of an architectural programme — type, material logic, light with direction and time, sequence, moment. They leave less open. They do not guarantee geometry: a named courtyard can still come back wrong, and the spatial check on this map exists for that. But a brief made of substance can be checked line by line. A brief made of mood can only be liked.
The strongest objection
"Style" and "substance" are not stable categories, in architecture or in how a model is steered. "Kerala vernacular" looks like a style word and carries roof geometry, material system, climate response, spatial order and craft at once — and it can steer harder than a paragraph of geometry. Meanwhile a supposedly substantive instruction — "courtyard house", "compressed entry" — may be obeyed only at the surface. The split teaches specificity well. Taught as a rule about which words control appearance and which control architecture, it teaches a false map.
What would make us revise it
If blind studio comparisons showed that precisely named traditions match substance-rich briefs on the outputs students can defend in crit, we would rewrite the card around vague against precise and drop the style/substance line. We already accept that a named tradition is substance in disguise. The open question is whether the split still helps a beginner more than it misleads them. We check it against evidence each edition.
Try it
Take a real brief you have sent. Split every descriptive word into two columns: style (mood, era, adjective) and substance (a space, a material, a direction, a count, a sequence). Most first attempts run nine to one. Rebalance toward substance and run both.
Take it to crit
Does the student's brief carry programme and geometry, or only mood words that leave the decisions to the model — and can they point to one substance word that visibly changed the output?
What this idea builds on
What this idea opens up
Sources
- prompt-as-program
- briefing-ai-field-guide
- The Seduction of the Machine
Open this idea on the map · The complete map · Logika · RBDS AI Lab, India · revised every edition.