Study Record · CN-2026-06
Executive Briefing

SubjectInteresting Design

Participants51 people

Images36, on 3 boards

Dots placed834

PublishedJune 15, 2026

What Makes a Design Interesting?

We showed 51 people a bunch of designs and asked one question: which of these feel interesting — and which don't? This is the plain-language record of what they told us.

Green dots — "this is interesting" Red dots — "this is not" Dark, quiet areas — people weren't sure what to do

Board 1

Logos

Green piles up on the drawn, character-filled logos. Red pools on the plain ones — the t-shirt and the simple wordmark took the most red dots of all.

Board 2

Ads

Green gathers on the clever-idea pictures. Red lands on the abstract rings and the trash-bag heart. The water words got almost no dots at all.

Board 3

App & Computer Screens

The toughest board — mostly red. Nearly all the green sticks to just two pictures: the whale and the bull.

1.0

How the Study Worked

We put 36 pictures on three big boards. One board had logos, one had ads, and one had app and computer screens.

Then 51 people looked at them. Each person got green dots and red dots. Green dot = "this is interesting." Red dot = "this is not interesting." All together, they placed 834 dots.

When we counted the dots, something surprising happened. People didn't just pick a favorite color or style. Almost everyone was following the same hidden rule — even if they didn't know it.

2.0

The Big Discovery

The winning pictures don't look alike at all. But they all do the same trick.

Every winning picture is a tiny puzzle. You look at it, and one second later you see something more — and figuring it out feels like a little prize for your brain.

Principal Finding

People find a design interesting when it gives their brain a small puzzle to solve — and the answer feels like a reward.

Here's the coolest part: the three top-scoring pictures — one from each board — all use the exact same trick. Each one takes a living animal and builds it out of something else, like letters or data.

Plate 0191% green

The bear and the letter B. The bear isn't standing next to the letter — the bear IS part of the letter. Two things in one body.

Plate 0291% green

The whale made of numbers. This chart doesn't just show a whale — the whale's body IS the chart. The facts become the animal.

Plate 0392% green

The computer-line bull. The highest score in the whole study. A bull drawn out of live computer lines and numbers.

People weren't voting for animals. They were voting for the trick: two different things fused into one. A bear + a letter. A whale + a chart. A bull + a computer screen. The moment you notice the mash-up — that's the "interesting" feeling.

Other winners play the same game in different ways. A paper airplane casts the shadow of a real fighter jet. A coffee cup turns out to be a bird's eye. A monkey hides inside a boring clothing tag. Each one has a second meaning waiting for you to find it.

Plate 0488% green

The paper mask. A famous play rebuilt out of cut paper. You can see the careful handwork — and that made people like it even more.

Plate 0566% green

The paper plane's big dream. A small paper plane casts the shadow of a fighter jet. You "get it" one beat after you look.

Plate 0670% green

The cup that's an eye. Look once: a bird. Look again: its eye is a cup of coffee. Surprise!

3.0

What People Did Not Like

The losing pictures failed in two opposite ways — and both break the same rule.

Way to fail #1: Nothing to figure out. Some designs are so plain and quiet that there's no puzzle at all. Your brain looks, finds nothing to do, and moves on.

Plate 0797% red

The t-shirt. A small, neat logo on a plain black shirt. Nothing is wrong with it — and that's the problem. 31 out of 32 dots were red. The biggest loser in the study.

Plate 0881% red

The plain logo. Simple white letters on black. Very clean, very professional — and people found nothing to discover in it.

Plate 0987% red

The mystery rings. Dark circles that don't stand for anything. A picture of nothing gives your brain nothing back.

Way to fail #2: Looking fancy but meaning nothing. Some designs dress up like they're full of information — glowing lines, numbers, tech-looking screens. But when your brain tries to read them, there's nothing actually there. We call this costume complexity. People rejected almost all of it.

Plate 1078% red

The sci-fi dashboard. Looks like a spaceship screen full of data. But the "data" is just decoration. It's wearing a costume.

Plate 1177% red

The water words. The words "World Water Day" made out of... water. When you solve the puzzle, the answer is just what you already saw. No prize.

Plate 1279% red

The trash-bag heart. A real puzzle with a real answer — but the answer is gross. This one teaches its own lesson (see below).

Note 3.1 — The Trash-Bag RuleThe trash-bag heart was a good puzzle, and people solved it. They still gave it red dots — because the answer felt yucky. So the rule has two parts: the design must give you a puzzle, and the answer must feel good. A clever idea with a gross ending scores just as badly as no idea at all.

4.0

The Twin Test

The clearest proof in the whole study: two pictures made the same way, with opposite scores.

Plate 03 (again)92% green

The bull. Glowing dots and lines that add up to something real — a bull. The style is showing you a subject.

Plate 1363% red

The face. Almost the same glowing-dots style — but here it's just wallpaper. The style is only showing off.

Same tools. Same look. Opposite votes. The only difference: the bull's dots and lines are about something, and the face's dots are just decoration. Style doesn't win by itself — style wins when it reveals a subject.

5.0

The Ones in the Middle

A few pictures split the room — half green, half red. They show exactly where the line is.

Plate 14Split · 52% green

The fancy bull. Beautifully drawn, tons of skill — but no twist. It looks like a hundred old-timey labels you've seen before.

Plate 01 (again)91% green

The bear again. Just as skillful as the bull — but with the surprise built in. That's the whole difference.

The bull and the bear are both drawn with great skill. But the bull got half the votes and the bear got almost all of them. The gap between them isn't talent — it's surprise. Skill alone makes people shrug. Skill plus a twist makes people point.

6.0

The Rule Behind Everything

Think of every design as making a deal with your eyes: "Spend a moment looking at me, and I'll give you something back."

The 51 people in this study judged every picture by whether it kept that deal. The plain designs asked for attention and gave nothing back — bad deal. The fake-fancy designs promised a lot and delivered nothing — worse deal. The winners always paid up: a hidden animal, a joke that clicks, a chart that turns into a whale.

In One Sentence

Interesting design is a fair trade: it takes a moment of your attention and pays you back with a small, happy discovery.

7.0

What To Do With This

Mash two things into one. A bear + a letter. A whale + a chart. The top three winners were all this one move.
Make data about something real. If people can say what your chart shows in one word ("a whale!"), you win.
Let the hard work show. Hand-drawn lines, cut paper, careful detail — visible effort tells people the puzzle is worth solving.
Hide one surprise. The sweet spot is a one-second puzzle: quick enough that everyone solves it, slow enough to feel like a discovery.

Not Recommended

Being plain on purpose. Super-simple logos got the worst scores of all. To this crowd, "quiet" read as "empty."
Costume complexity. Techy lines and fake data that mean nothing — people see through it fast.
Puzzles with no prize. Making the word "water" out of water gives nothing extra. So does a tease that says "coming soon" and shows nothing.
Gross endings. A clever puzzle with a yucky answer loses just like having no puzzle at all.

Note 7.1 — One Warning Before You Run With ThisThis study only asked about the word "interesting." A plain, quiet logo might lose at "interesting" but win at "trustworthy" or "calm" — and sometimes that's the trade you want. Also, these voters were probably design-savvy people who have seen a thousand plain logos before. A different crowd might vote differently. If it really matters, test again with the word you actually care about.

Interesting Design · A Constellations Study
51 completions · Published June 15, 2026 · Executive briefing
Need every plate, score & method note? Open the Full Technical Report — tab at the top of the page.

1.0

The Heatmaps

Aggregate dot placement across all 51 participants, before threshold sorting. These are the raw temperature maps the arrays below are cut from.

Green dots — associated with “interesting”Red dots — least “interesting”Quiet zones — low engagement, not rejection

Board 1 — BrandingDense green over the illustrated, character-driven marks (2, 8, 12); heavy red concentration on the minimal wordmark (3) and the t-shirt mockup (11). The photographic silhouette (4) drew the board’s largest crowd of dots — most of them red.

Board 2 — MarketingGreen clusters on the concept-driven pieces (2, 6, 7, 12); red pools on the abstract rings (1) and the trash-bag heart (8). The liquid lettering (3) sits in a near-quiet zone — the board’s least engaged image.

Board 3 — UI/UXThe most red-shifted board of the three: 7 of 12 images crossed the resistance threshold. Green concentrates almost entirely on the two data-as-creature pieces (4, 8) and the particle scatter (11).

Reading note: dot density reflects attention as well as valence. A crowded image is one people felt strongly about in either direction; the dark, quiet areas are where participants didn’t know what to do with an image — disinterest or confusion, which is not the same thing as rejection.

2.0

Executive Summary

Fifty-one participants sorted 36 images across three boards — branding, marketing, and UI/UX — against a single word: interesting. The result is unusually coherent. “Interesting” did not behave like a style preference. Participants did not reward a palette, a medium, or an era; the winning images share almost nothing visually. What they share is a payoff structure: each one stages a small act of decoding — a second reading that resolves a beat after the first glance — and rewards the viewer for completing it. Both failure modes sit at the ends of that same curve: images with nothing to decode (minimalist wordmarks, abstract patterns, sci-fi interface texture) and images that can’t be decoded at all (opaque diagrams, atmospheric teases).

Key insight

The three highest-scoring images in the entire study — one per board — are the same move made three ways: an animal fused with an information structure. A bear built into a letterform (91.3%), a whale drawn out of catch data (91.4%), a bull rendered as live wireframe (92.3%). Participants aren’t asking for animals; they’re rewarding the splice — a living subject caught inside a system of notation. That collision of categories is what this audience reads as “interesting.”

3.0

Resonance Array

15 images cleared the 60% positive threshold. They share no palette, medium, or era — they share a payoff structure.

Board 1 — Branding: craft with a twist

ResonanceBranding · Swatch 8

Bernhard — bear embracing the letter B, “Depuis 1827”

91.30% res (21/23) 8.70% ris (2/23) 8.21% of board points

ResonanceBranding · Swatch 12

Made in USA — woodcut monkey garment label

89.47% res (17/19) 10.53% ris (2/19) 6.79% of board points

ResonanceBranding · Swatch 2

Serif “A&C” lockup with insect ornament

81.25% res (13/16) 18.75% ris (3/16) 5.71% of board points

ResonanceBranding · Swatch 10

SUPER — vintage microphone lockup, “Top Secret”

68.42% res (13/19) 31.58% ris (6/19) 6.79% of board points

ResonanceBranding · Swatch 6

Hot-pink alligator illustration

68.18% res (15/22) 31.82% ris (7/22) 7.86% of board points

Every resonant branding image is illustrated — not one photograph clears the threshold — and every one carries high craft density: engraving texture, woodcut linework, ornamental detail. But craft alone doesn’t explain the array, because the Contested array contains equally well-crafted engravings (the bull, the “Be Frugal” dog) that failed to excite anyone. The variable that separates the winners is a conceptual splice. The bear doesn’t accompany the letter B; it embraces it, becoming part of the letterform (91.3%). The monkey isn’t a mascot; it’s embedded in the bureaucratic grid of a garment care label (89.5%). The ampersand hosts an insect. These images put two categories in one body, and the moment of noticing that is the “interesting” being measured.

The hot-pink alligator (68.2%) and SUPER microphone (68.4%) sit lower in the array and show the mechanism at reduced strength: the alligator’s twist is chromatic rather than structural (a swamp creature in shocking pink), and the microphone lockup is rich pastiche whose surprise is mostly stylistic. Still rewarded — but a full tier below the true hybrids.

Board 2 — Marketing: the two-step read

ResonanceMarketing · Swatch 2

Macbeth — red papercraft theatrical mask

88.46% res (23/26) 11.54% ris (3/26) 9.15% of board points

ResonanceMarketing · Swatch 6

Botanical letterform with red foliage

72.00% res (18/25) 28.00% ris (7/25) 8.80% of board points

ResonanceMarketing · Swatch 12

Bird silhouette with a cup for an eye, red field

70.00% res (14/20) 30.00% ris (6/20) 7.04% of board points

ResonanceMarketing · Swatch 10

Skull sculpture defaced with neon scribbles

66.67% res (16/24) 33.33% ris (8/24) 8.45% of board points

ResonanceMarketing · Swatch 7

Paper plane casting a fighter-jet shadow

65.52% res (19/29) 34.48% ris (10/29) 10.21% of board points

ResonanceMarketing · Swatch 5

“Age of Rage” — crumpled newspaper cover

62.50% res (10/16) 37.50% ris (6/16) 5.63% of board points

The marketing winners are visual metaphors — images that resolve into a second meaning: a paper plane casting the shadow of the fighter jet it “wants to be” (65.5%), a cup that becomes a bird’s eye (70.0%), rage literalized as a crumpled newspaper page (62.5%), Macbeth rebuilt as tactile papercraft (88.5%). Note that the Macbeth mask — the board’s runaway winner — is also the board’s most crafted object: the two-step read and visible handwork compound each other, echoing the branding board’s logic.

Scores here run lower than branding’s top end (mostly 62–72%), and the dissent share is real: the defaced skull carries a third of its dots in red. Wit resonates, but edgy execution splits the room — conceptual images buy attention at the price of consensus.

Board 3 — UI/UX: data given a body

ResonanceUI/UX · Swatch 8

Wireframe bull with live telemetry

92.31% res (24/26) 7.69% ris (2/26) 9.63% of board points

ResonanceUI/UX · Swatch 4

Whale-catch infographic — the dataset drawn as the whale

91.43% res (32/35) 8.57% ris (3/35) 12.96% of board points

ResonanceUI/UX · Swatch 11

Particle scatter with data panels

80.00% res (20/25) 20.00% ris (5/25) 9.26% of board points

ResonanceUI/UX · Swatch 6

Particle-wave scene with scene menu

66.67% res (10/15) 33.33% ris (5/15) 5.56% of board points

Only four images survived on the most hostile board of the study, and the top two are the study’s two highest scores overall. What they share is precise: the data is about something. The whale infographic doesn’t decorate a chart with a whale — the whale’s body is the chart, its length carrying national catch figures (91.4%, and the board’s highest attention share at 12.96%). The wireframe bull turns telemetry into anatomy (92.3%). The particle scatter (80.0%) earns its place by pairing visual complexity with legible data panels — the complexity appears to be of something. Participants reward information visualization with a nameable subject and punish the same visual vocabulary the moment the subject disappears, as the Resistance array makes brutally clear.

4.0

Resistance Array

15 images crossed the 60% negative threshold. The rejections are more instructive than the wins — they isolate exactly which offers this audience refuses.

Board 1 — Branding: the cost of restraint

ResistanceBranding · Swatch 11

Small logo on a black t-shirt mockup

3.13% res (1/32) 96.88% ris (31/32) 11.43% of board points

ResistanceBranding · Swatch 3

Athtelier — minimal white-on-black wordmark

18.52% res (5/27) 81.48% ris (22/27) 9.64% of board points

ResistanceBranding · Swatch 9

Geometric sticker mark over motorcycle photo

36.36% res (8/22) 63.64% ris (14/22) 7.86% of board points

ResistanceBranding · Swatch 4

Blurred figure photograph with abstract mark

37.84% res (14/37) 62.16% ris (23/37) 13.21% of board points

The most decisive verdict in the entire study is the black t-shirt mockup: 31 of 32 dots red — 96.9% resistance. Nothing about it is badly made; it’s a perfectly professional product shot of a small logo on a garment. That’s the point. Alongside the minimal Athtelier wordmark (81.5% resistance) and the geometric sticker (63.6%), it defines this board’s failure mode: sophisticated absence. Minimalist restraint — the default register of contemporary premium branding — reads to this audience as withholding: nothing to decode, no second reading, no reward. The blurred silhouette photograph is the interesting nuance: it drew the largest dot count on the board (13.21% of all points, 37 dots) before being voted down 62/38. It provoked — participants stopped, looked, engaged — and then rejected it. Attention and interest are separable measurements, and this image separates them.

Board 2 — Marketing: three ways to lose the reader

ResistanceMarketing · Swatch 1

Concentric rings, dark abstract

13.33% res (4/30) 86.67% ris (26/30) 10.56% of board points

ResistanceMarketing · Swatch 8

Trash bag knotted into a heart

20.83% res (5/24) 79.17% ris (19/24) 8.45% of board points

ResistanceMarketing · Swatch 11

“Coming Soon” — face behind fluted glass

21.74% res (5/23) 78.26% ris (18/23) 8.10% of board points

ResistanceMarketing · Swatch 3

“Dia Mundial da Água” — liquid 3D lettering

23.08% res (3/13) 76.92% ris (10/13) 4.58% of board points

The marketing rejections fail in three distinct ways, and distinguishing them matters. The concentric rings (86.7% resistance) are abstraction without a referent — a one-step image with nothing behind it. The liquid lettering (76.9%) is tautology — the word for water rendered in water; decoding it returns exactly what the first glance already gave, and participants barely engaged (4.58% of points, the study’s quietest image). The “Coming Soon” tease (78.3%) is marketing’s version of the minimalist wordmark: atmosphere that promises meaning and delivers none.

The trash-bag heart (79.2%) is the most diagnostically valuable rejection in the study. Structurally it is identical to the winners — a genuine two-step visual metaphor, well executed. Participants clearly decoded it. They docked it anyway, because the payoff is aversive: garbage as a heart resolves into disgust. This cleanly separates the mechanism into two requirements — the image must invite decoding and the decoded meaning must reward. Wit with a repellent punchline scores like abstraction.

Board 3 — UI/UX: complexity theater

ResistanceUI/UX · Swatch 7

Orange/black striped poster pair

14.29% res (2/14) 85.71% ris (12/14) 5.19% of board points

ResistanceUI/UX · Swatch 3

“LOC” — vector-editing grid view

19.05% res (4/21) 80.95% ris (17/21) 7.78% of board points

ResistanceUI/UX · Swatch 1

Dark HUD instrument panel

21.74% res (5/23) 78.26% ris (18/23) 8.52% of board points

ResistanceUI/UX · Swatch 9

Warm gradient wave field, 70°F

23.08% res (6/26) 76.92% ris (20/26) 9.63% of board points

ResistanceUI/UX · Swatch 12

Red technical diagram of a figure

23.53% res (4/17) 76.47% ris (13/17) 6.30% of board points

ResistanceUI/UX · Swatch 2

Iridescent orb on black

33.33% res (10/30) 66.67% ris (20/30) 11.11% of board points

ResistanceUI/UX · Swatch 10

Particle face with editorial layout

37.50% res (9/24) 62.50% ris (15/24) 8.89% of board points

Seven of twelve images rejected — the audience showed near-zero patience for what might be called complexity theater: the HUD panel, the vector grid, the striped posters, the gradient wave. These wear the costume of information — numbers, ticks, technical annotation — but nothing resolves; they are texture performing as data. The most surgical comparison in the study sits on this board: the particle face (62.5% resistance) and the wireframe bull (92.3% resonance) use nearly the same rendering technique. The face reads as a style demo — particles as wallpaper; the bull reads as a revelation — particles as anatomy. Same vocabulary, opposite verdicts, and the only changed variable is whether the technique is disclosing a subject.

The red technical diagram (76.5% resistance) marks the far end of the decoding curve: it has both a figure and data, but the notation is illegible — a puzzle with no solution offered. Too little to decode fails; so does too much.

5.0

Contested Array

Six images split the room — labeled “Neutral” in the platform export. These are the tipping points: images holding exactly one of the two ingredients resonance requires.

ContestedBranding · Swatch 5

“Be Frugal” — engraved dog badge on a coin motif

55.56% res (10/18) 44.44% ris (8/18) 6.43% of board points

ContestedBranding · Swatch 1

Engraved bull illustration, heritage style

52.17% res (12/23) 47.83% ris (11/23) 8.21% of board points

ContestedMarketing · Swatch 9

Glitch-erased face with error messages

54.17% res (13/24) 45.83% ris (11/24) 8.45% of board points

ContestedMarketing · Swatch 4

High-heel / sneaker hybrid product shot

46.67% res (14/30) 53.33% ris (16/30) 10.56% of board points

ContestedBranding · Swatch 7

Rotated bold-type poster, white on black

45.45% res (10/22) 54.55% ris (12/22) 7.86% of board points

ContestedUI/UX · Swatch 5

ŪRE — red brand poster, silhouetted figure

42.86% res (6/14) 57.14% ris (8/14) 5.19% of board points

The contested images are where exactly one of the two required ingredients is present. The engraved bull (52.2%) and the “Be Frugal” dog badge (55.6%) have all the craft of the resonant branding marks and none of the splice — they are competent performances of a familiar heritage genre, and the room split almost perfectly on whether well-executed familiarity counts as interesting. This is the clearest tipping point in the data: the distance between the contested bull and the 91% bear is not skill, it’s surprise.

The marketing pair inverts the formula: the glitch-face (54.2%) and heel-sneaker hybrid (46.7%) have concepts but ambivalent payoffs — unsettling rather than delightful, clever rather than resolved. And the ŪRE poster (42.9%, on the UI/UX board) reads as a category mismatch — brand-poster aesthetics judged against interface expectations. Contested zones, as usual, are where the open questions live; these six images are the natural targets for anchor segmentation in a follow-up pass.

6.0

Comparative Analysis

What actually separates green from red, across all three boards.

Dimension	Resonance array	Resistance array
Medium	Illustration, craft objects, data-drawing; photography only when it documents a constructed idea (papercraft, staged metaphor)	Photography as atmosphere or product context; digital texture and pattern
Information density	High and legible — detail that resolves into meaning	Either near-zero (minimalism, abstraction) or high but illegible (HUDs, opaque diagrams)
Second reading	Present and rewarding — a splice, metaphor, or revealed subject	Absent (one-step images) or aversive (decoded disgust)
Subject	Nameable — a creature, a character, a dataset about something	Unnameable — mood, texture, restraint, tease
Emotional register	Wit, warmth, discovery	Cool detachment, blankness, or repulsion
Production signal	Visible handwork and effort	Effortlessness as a value — polish in service of restraint

The underlying mental model looks like a cognitive transaction. Each image makes an implicit offer: spend attention here and you’ll get something back. Participants priced that offer with remarkable consistency. Minimalism and abstraction were rejected not as ugly but as bad deals — they ask for attention and return nothing. Complexity theater is a worse deal dressed better — it advertises a payload it doesn’t carry. The winners pay out: a hidden fusion, a metaphor that clicks, a dataset that becomes an animal. And the trash-bag heart establishes the contract’s fine print: the payout must be pleasurable, not merely present.

Predictor variables, in rough order of strength: (1) a categorical splice — two kinds of thing in one body — predicts resonance more reliably than any stylistic property in the study; (2) a nameable subject rendered through an information structure predicts resonance on visual-density images; (3) product-context realism (mockups, product shots) predicts resistance; (4) heritage-genre familiarity without a twist predicts a contested split, not resonance; (5) high attention (Total Points) predicts nothing by itself — the most-engaged image on Board 1 was voted down, the most-engaged on Board 3 was voted up.

7.0

Strategic Implications

Translating the decision logic into practice.

Embrace

Build the splice. Fuse the subject with a structure from another category — creature × letterform, dataset × anatomy, prop × symbol. The study’s three top scores are all this one move.
Give data a body. Visualization about a nameable subject beat visualization aesthetics by 25–30 points. If the audience can say what the data is of in one word, you’re on the right side of the line.
Let the handwork show. Engraving, papercraft, woodcut texture all pulled green — visible effort signals a payload worth decoding. But pair it with a twist: craft alone lands contested, not resonant.
Design the second reading. One beat between glance and click is the sweet spot — long enough to feel like discovery, short enough to guarantee arrival.

Avoid

Restraint as identity. Minimal wordmarks and quiet mockups took the harshest scores in the study (81–97% resistance). For this audience, on this prompt, understatement reads as absence.
Complexity theater. HUD chrome, particle wallpaper, technical texture without content — the UI/UX board rejected all of it, at 62–86%.
Tautology. Executions that illustrate their own literal content (water type made of water) return nothing on the second read — and get ignored, not just rejected.
The aversive punchline. A well-built metaphor with a repellent payoff (79% resistance) scores like having no idea at all. Provocation is a separate strategy from interest — choose it deliberately.
Teasing without content. “Coming soon” atmospherics promise a payload they don’t deliver; anticipation did not register as interest.

Conditional guidance. These findings answer one prompt — interesting — and should not be over-rotated into a general aesthetic verdict. Minimalism may well win the same audience on “trustworthy,” “premium,” or “calm”; the Athtelier mark isn’t bad branding, it’s uninteresting branding, and for many briefs that’s an acceptable trade. Likewise the audience composition matters: a design-literate sample (as this one likely is) has seen a thousand minimal wordmarks and a thousand particle renders; a general consumer sample might find novelty exactly where this group found cliché. Where the stakes justify it, re-run the boards against the competing attribute you actually need.

8.0

Methodology & Confidence

This study collected 51 completions across 111 opens, generating 834 dot placements over 36 images on three boards (280 on Branding, 284 on Marketing, 270 on UI/UX). Arrays were classified at a 60% threshold. At n=51, board-level patterns qualify as clear consensus signals, and the study’s headline findings — the rejection of minimalism, the strength of the creature-system splice, the collapse of complexity theater — rest on wide margins that small-sample noise cannot plausibly explain.

Per-image confidence is more uneven: individual images carry between 13 and 37 dots, so single-image scores in the 60–70% band (the pink alligator, the particle-wave scene, the “Age of Rage” cover) should be read as directional rather than settled. The near-unanimous verdicts (t-shirt mockup at 1/32 green; wireframe bull at 24/26 green) are stable at any reasonable read of the data.

Suggested next passes: anchor segmentation on the contested images — the engraved bull is a natural anchor for splitting heritage-craft lovers from heritage-craft skeptics, and would clarify whether the vintage-Americana contested zone is generational or dispositional. Point isolation (first-dot view) would show whether the splice images win on gut reaction or on consideration — the one-beat-decode hypothesis predicts they win on the second look, which is testable. Finally, a follow-up board pairing the same images against a different attribute (“trustworthy” or “premium”) would price the trade-off the strategic section describes.

Interesting Design · A Constellations Study
51 completions · Published June 15, 2026 · Analysis generated July 29, 2026
For the summary version, open the Executive Briefing — tab at the top of the page.