TALAMANA · THE AI LITERACY MAP FOR ARCHITECTURE AND DESIGN · Studio Practice · AGE 19—22 · POSITIONAL · HELD
Discipline when time is short
Changing one named thing per cycle produces a result with a reason attached.
The dilemma
The jury is at nine tomorrow. The render is not right. You can run fifty variations in an hour and pick one, or run five, changing one named thing each time, and know why the fifth is better. Which?
The choices
Flood the engine: fifty images, scroll, pick the one that feels right, hope it holds at nine. Or hold the discipline: name the variable that is wrong — light direction, material, camera height — change only that, judge fast, log the change, go again. Fifty feels faster. Five usually is.
The consequence
Flooding produces a pick you cannot explain and cannot reproduce; the jury asks "why this one?" and the honest answer is "it came up." One-variable cycles produce a result with a reason attached, and a log that lets you go back to cycle three when cycle five overshoots.
The case
One in the morning, a courtyard render, the shadow is wrong. The student generates forty, scrolls, picks the warmest. At the jury the shadow is right, but the courtyard's proportion has quietly changed between generations — the engine drifted the geometry while the student was judging light. Nobody noticed, because nothing was held constant.
Take it to crit
Under a mock deadline, watch whether the student tightens the brief or floods the engine. The panic move is the tell. Ask for the log.
How it works
Each generation is a draw from a range. Change one input and you learn what that input controls; change five and you learn nothing you can use next time. The log is the studio's memory of the search. Studies of designers using generative tools report fixation — settling early on what the engine shows — as a real cost. So the discipline here guards against fixation under pressure, not only against waste.
What this idea builds on
What this idea opens up
Sources
- 08_render-to-reel
- iteration discipline matters most when time and cost are tight
- Wadinambiarachchi et al., 2024 (CHI)
- Understanding Design Fixation in Generative AI
Open this idea on the map · The complete map · Logika · RBDS AI Lab, India · revised every edition.