Do Assembled Screens Look Generic?
Tonight's five agent screens come from films eight, twelve, and thirty-three. Two tutor-card states from film twelve: the confident answer and the approval gate. One quiz-results screen from film eight. And two variants of the same confident state from film thirty-three, Path A and Path B. Five scre
Do Assembled Screens Look Generic?
Tonight's five agent screens come from films eight, twelve, and thirty-three. Two tutor-card states from film twelve: the confident answer and the approval gate. One quiz-results screen from film eight. And two variants of the same confident state from film thirty-three, Path A and Path B. Five scre
Series: Figma for Educational AI Topic: FIGMA · DESIGN QUALITY
Chapters
0:00 Introduction 0:24 Introduction 2:11 Your Turn
YOUR TURN
Your turn. Paste this into Claude: Here are descriptions of five educational screens built by an AI agent. Rate each from one to five on distinctiveness. Then list one thing that would push each screen up by one point. Then check two things yourself. Show both the agent screen and a community screen you admire side by side: what is the gap? And pic
@NikBearBrown
Part of the Figma for Educational AI series.
Sources and citations: see the reel's SOURCES.md in the open textbook.
▶ The full series playlist — Brutalist https://www.youtube.com/playlist?list=PLG9H-C6rp5RU
What this video covers
- 0:00 — Introduction
- 0:24 — Introduction
- 2:11 — Your Turn
Transcript
A lightly formatted transcript of the video, with timestamps so you can jump to any section on YouTube.
Introduction
[0:00] Sawaddi, this is Liam in for Bear. Do assembled screens look the same? Better question, do assembled screens look generic? Three terms, agent assembled, a screen built from a Figma design spec by a headless AI with no human touching the layout. Designed, made by a human designer and published to Figma community. And distinctiveness, how memorable or unique a screen looks at first glance, independent of its content. Tonight's five agent screens
Introduction
[0:25] Come from films 8, 12, and 33. Two tutor card states from film 12, the confident answer and the approval gate. One quiz result screen from film 8, and two variants of the same confident state from film 33, path A and path B. Five screens all built by an agent from a Figma spec with no human designer's hand on the layout, rendered at one size, ready to shuffle. The rating form is one page, 10 numbered screens shuffled, one question for each. On a scale from one to five, how distinctive is this screen? One means forgettable, could have come from any educational app. Five means you would recognize this designer's point of view. Students fill it in
Without knowing which screens are agent made. A tally script counts the result. Three headless model runs predict how a student would rate each agent screen. Not students' predictions labeled as such. Each run got a text description of the five screens in the same scale. Run 1 2 3 4 1 1. Run 2 3 5 4 2 1. Run 3 2 3 4 1 2. The tally, the quiz result screen from film 8 gets a four in every run. That is the only agreement. The approval state from film 12 swings between three and five depending on the framing. It stakes three as distinctiveness when the rater is thinking like a student and as content when thinking
Like a designer. The two film 33 cards score one or two in every run, too small, too similar. The other five screens are pending Bear. Five design screens from Figma community, cited, exported at the same size, and shuffled in. Without them, the test can't run. You need the baseline to know if the agent screens score lower on average. The citations are in the evidence folder. The task is to open each community file, export one frame, and drop it in. Verdict pending, the protocol is set, the agent screens are ready, the form is printed. What it is waiting for is five design screens, students filling in the form, and a count. Pending bear, one class, 10 minutes.
Your Turn
[2:11] Your turn, paste this into Claude. Here are descriptions of five educational screens built by an AI agent. Rate each from one to five on distinctiveness. Then list one thing that would push each screen up by one point, then check two things yourself. Show both the agent screen and a community screen you admire side by side. What is the gap? And pick the agent screen you rated highest. What one change would close half that gap? Do assembled screens look generic at Nick Bear Brown?
