TALAMANA · THE AI LITERACY MAP FOR ARCHITECTURE AND DESIGN · The Brief · AGE 18—20 · POSITIONAL · HELD
The brief-to-crit loop
Every crit must produce a next action, and most often that is a rewritten brief.
The dilemma
An output comes back and it is not right. You can press generate again and hope — the re-roll — or you can stop, critique the output, and carry the critique back into the brief. The re-roll takes ten seconds. The loop takes ten minutes and a rewritten document.
The choices
Re-roll until something you like appears, then select it. Or crit the output as you would a junior's drawing — what held, what drifted, what the brief never said — and revise the brief to v2 before running again.
The consequence
The re-roll treats the model as a slot machine and the designer as a picker. When something good arrives you cannot say why, and you cannot make it happen again. The loop treats every failure as information about the brief: the brief improves, the outputs converge, and the reasoning is on paper. A crit that only judges the image stalls. A crit that rewrites the brief turns generation into iteration.
The case
A student shows a courtyard house after forty generations and says the light "finally came right." Asked what changed between the thirty-ninth and the fortieth, they cannot say: same brief, different seed. A second student shows three images and three briefs — v1, v2, v3 — with the crit notes between them: "v1: light arrives from nowhere. Add north skylight, morning." The second student can do it again tomorrow, on another building.
Our position
The studio loop closes through the brief: output, crit, revised brief, again. A crit must produce a next action, and most often that action is a change to the document, not a choice between images.
Why we hold it
It is the difference between gambling and iterating. The gambler has results. The iterator has a method, and the method carries over to the next project, the next tool, and the jury room where someone asks why. Sometimes the next action is a sketch over the output, a change made directly to the geometry, or a different tool altogether; the brief records that decision too, so the record stays whole.
The strongest objection
Dreaming engines are sampled, not steered, and taking many draws from a good brief is a legitimate method; photographers shoot hundreds of frames and select. Treating every re-roll as a lapse in discipline may be too strict about what is sometimes the correct use of a stochastic tool: hold the brief, widen the draw, then curate.
What would make us revise it
Evidence from studio that students who sample widely from a stable brief and curate produce work they can defend as well as students who revise the brief between runs. We already teach curation as the middle phase; if curation alone carried authorship, this card would soften to "loop or sample, but know which you are doing." Reviewed each edition.
Prove it
Show one full loop — an output, the crit notes on it, the brief rewritten from those notes, and the next output — where the crit changed the brief and not just the image you selected.
Take it to crit
After critiquing the output, did the student go back to the brief or to the re-roll button — and can they show one crit finding that became a specific edit in the next version of the brief?
What this idea builds on
What this idea opens up
- Nothing yet names this as a foundation.
Sources
- briefing-ai-field-guide
- curriculum-spine
- The Seduction of the Machine
- The AI Design Director Protocol
Open this idea on the map · The complete map · Logika · RBDS AI Lab, India · revised every edition.