{ "protocol": "agentic-next-v1", "model": "gpt-6.1-sol", "kitSha256": "4e37eb9e33cabef3bd6b723cdee5e1df23c159351fde34802fd5df1c292e8f74", "summary": { "pairs": 10, "wins": 1, "ties": 3, "losses": 6, "meanPreference": 3.35, "dimensions": { "hierarchy": 3.4, "genericSignals": 3.3, "productSpecificity": 3.25, "typography": 3.35, "spacing": 3.7, "color": 3.45, "copy": 3.7, "states": 4, "mobile": 3.6, "distinctiveness": 3.65 } }, "domains": [ { "id": "saas-dashboard", "label": "Studio billing dashboard", "route": "/dashboard", "task": "Identify the outstanding invoice and filter the ledger by period and status. Keep its exact invoice records and totals.", "result": "Plain", "votes": [ 3, 3 ], "ballots": [ { "order": [ "plain", "kit" ], "preference": 3, "dimensions": { "hierarchy": 3, "genericSignals": 2, "productSpecificity": 3, "typography": 3, "spacing": 5, "color": 3, "copy": 3, "states": 4, "mobile": 5, "distinctiveness": 2 }, "reason": "A’s “Invoice overview,” payment-focused subtitle and prominent Dovetail alert establish the task more clearly than B’s generic “Dashboard.” Both retain the displayed $25,450 total, $18,900 paid and $6,550 awaiting payment, with matching visible invoice records and accessible period/status controls. Both verified September/Overdue filters correctly show zero rows on desktop and mobile. B’s compact phone alert and ledger expose more records sooner; A spends more vertical space on context and the overdue action. Neither shows the calibrated two-line period pattern.", "cohortReason": "B repeats the purple branding, pale lavender surfaces, rounded cards and compact sans-serif treatment visible across its client portal, research desk, settlements and booking products. A shares muted green and cream with several cohort peers, but its restrained ledger, small section labels, accounting alignment and amber payment notice form a more coherent invoice-specific composition. Its distinction comes from task organization rather than extra decoration.", "displays": [ "/agentic/saas-dashboard/judge/ballot-1-desktop.png", "/agentic/saas-dashboard/judge/ballot-1-mobile.png", "/agentic/saas-dashboard/judge/cohort-plain.png", "/agentic/saas-dashboard/judge/cohort-kit.png" ] }, { "order": [ "kit", "plain" ], "preference": 5, "dimensions": { "hierarchy": 5, "genericSignals": 5, "productSpecificity": 5, "typography": 5, "spacing": 3, "color": 5, "copy": 5, "states": 4, "mobile": 3, "distinctiveness": 5 }, "reason": "B makes the invoice task clearer with “Invoice overview,” a prominent Dovetail overdue notice, project context and a readable action. Both show the same totals ($25,450 total, $18,900 paid, $6,550 awaiting payment), matching visible records, period and status controls, and verified empty September / Overdue results on desktop and mobile. A uses space more efficiently: its phone view exposes more ledger rows, while B spends more height on introductory copy and the alert. Neither visible period selector demonstrates the calibrated two-line comparison.", "cohortReason": "A closely repeats its cohort’s purple branding, pale lavender panels, rounded cards and conventional dashboard typography, especially the fintech and booking examples. B shares muted green accents, small uppercase labels and restrained surfaces with its research, training and marketplace peers, so its identity is not wholly unique. Its compact financial summary, overdue-payment treatment and structured ledger nevertheless make a more coherent product-specific composition.", "displays": [ "/agentic/saas-dashboard/judge/ballot-2-desktop.png", "/agentic/saas-dashboard/judge/ballot-2-mobile.png", "/agentic/saas-dashboard/judge/cohort-kit.png", "/agentic/saas-dashboard/judge/cohort-plain.png" ] } ], "dimensions": { "hierarchy": 3, "genericSignals": 2.5, "productSpecificity": 3, "typography": 3, "spacing": 5, "color": 3, "copy": 3, "states": 4, "mobile": 5, "distinctiveness": 2.5 }, "rounds": { "kit": 1, "plain": 1 }, "images": { "kit": { "desktop": "/agentic/saas-dashboard/kit/desktop.png", "mobile": "/agentic/saas-dashboard/kit/mobile.png" }, "plain": { "desktop": "/agentic/saas-dashboard/plain/desktop.png", "mobile": "/agentic/saas-dashboard/plain/mobile.png" } }, "checks": { "kit": { "desktop": { "passed": true, "observed": "September / Overdue shows zero invoice rows." }, "mobile": { "passed": true, "observed": "September / Overdue shows zero invoice rows." } }, "plain": { "desktop": { "passed": true, "observed": "September / Overdue shows zero invoice rows." }, "mobile": { "passed": true, "observed": "September / Overdue shows zero invoice rows." } } }, "stops": { "kit": "target reached", "plain": "target reached" } }, { "id": "client-portal", "label": "Design client portal", "route": "/projects", "task": "A client reviews the Atlas proof and finds the approval action. Preserve search, project detail and approval.", "result": "Plain", "votes": [ 1, 2 ], "ballots": [ { "order": [ "plain", "kit" ], "preference": 1, "dimensions": { "hierarchy": 1, "genericSignals": 1, "productSpecificity": 1, "typography": 2, "spacing": 2, "color": 2, "copy": 2, "states": 3, "mobile": 2, "distinctiveness": 2 }, "reason": "Both sides verified search and persistent approval on desktop and mobile. A makes the Atlas proof the primary project, shows its identity artwork, explains that review is needed and exposes “Review proof & approve” directly. B gives three projects equal weight beneath a large generic gradient welcome banner; its initial render provides no visible approval action. A’s serif headline, restrained palette and project-specific artwork support a studio portal. Both show readable status pills and dates. On mobile, A preserves the proof, review context and approval button together, although its taller cards require more scrolling.", "cohortReason": "B repeats the lavender gradient banner, purple accents and white cards visible in its fintech, documentation and rehearsal cohorts, with similarly broad welcome copy. A shares warm neutrals, serif display type and muted green accents with its coffee, print and tools cohorts, so its styling is not wholly unique. However, its oversized Atlas artwork, featured review composition and explicit approval action give this portal a more specific identity and workflow.", "displays": [ "/agentic/client-portal/judge/ballot-1-desktop.png", "/agentic/client-portal/judge/ballot-1-mobile.png", "/agentic/client-portal/judge/cohort-plain.png", "/agentic/client-portal/judge/cohort-kit.png" ] }, { "order": [ "kit", "plain" ], "preference": 6, "dimensions": { "hierarchy": 7, "genericSignals": 6, "productSpecificity": 7, "typography": 6, "spacing": 6, "color": 6, "copy": 6, "states": 5, "mobile": 6, "distinctiveness": 5 }, "reason": "B makes Atlas the primary project, shows a recognizable proof preview, explains that the identity direction is ready, and exposes “Review proof & approve” on desktop and phone. A gives all projects equal weight with placeholder monograms and requires discovering approval through project detail. B's larger mobile card costs scanning space but keeps the next action visible. Both preserve search and approval after reload on both devices; B's advantage is clarity, not verified behavior. Review status and due dates are legible in both.", "cohortReason": "A repeats its cohort's purple accents, pale gradient welcome banner, emoji headline and white cards, particularly the finance and rehearsal products. B also shares a cohort style: cream backgrounds, green accents, serif editorial headings and fine rules recur in coffee, print exchange and the marketplace. Its Atlas proof artwork, project-specific palette and prominent review action give that shared vocabulary a stronger product purpose, though its identity is not wholly distinct from the cohort.", "displays": [ "/agentic/client-portal/judge/ballot-2-desktop.png", "/agentic/client-portal/judge/ballot-2-mobile.png", "/agentic/client-portal/judge/cohort-kit.png", "/agentic/client-portal/judge/cohort-plain.png" ] } ], "dimensions": { "hierarchy": 1, "genericSignals": 1.5, "productSpecificity": 1, "typography": 2, "spacing": 2, "color": 2, "copy": 2, "states": 3, "mobile": 2, "distinctiveness": 2.5 }, "rounds": { "kit": 1, "plain": 1 }, "images": { "kit": { "desktop": "/agentic/client-portal/kit/desktop.png", "mobile": "/agentic/client-portal/kit/mobile.png" }, "plain": { "desktop": "/agentic/client-portal/plain/desktop.png", "mobile": "/agentic/client-portal/plain/mobile.png" } }, "checks": { "kit": { "desktop": { "passed": true, "observed": "Search narrowed to Atlas. Approval survived reload." }, "mobile": { "passed": true, "observed": "Search narrowed to Atlas. Approval survived reload." } }, "plain": { "desktop": { "passed": true, "observed": "Search narrowed to Atlas. Approval survived reload." }, "mobile": { "passed": true, "observed": "Search narrowed to Atlas. Approval survived reload." } } }, "stops": { "kit": "regression", "plain": "target reached" } }, { "id": "landing", "label": "Coffee recipe calculator", "route": "/", "task": "A brewer adjusts the dose, reads the water targets and saves the recipe. Preserve the working calculator.", "result": "Tied or split", "votes": [ 5, 3 ], "ballots": [ { "order": [ "plain", "kit" ], "preference": 5, "dimensions": { "hierarchy": 6, "genericSignals": 4, "productSpecificity": 5, "typography": 4, "spacing": 5, "color": 4, "copy": 5, "states": 5, "mobile": 5, "distinctiveness": 3 }, "reason": "B groups dose, total water and ratio together, making the calculator’s immediate result easier to read, and places Save after the pour targets with an explicit unsaved message. A makes Save visible earlier on mobile and has a more distinctive serif-and-orange identity, but separates dose from its outputs. Both verified desktop and mobile actions correctly produced 288 g from 18 g and persisted after reload. B’s clearer calculation hierarchy earns a slight overall preference.", "cohortReason": "A shares warm paper backgrounds, serif editorial headings, muted green panels and fine rules with its print exchange, marketplace and client portal cohort, though orange italic typography gives Pourbook some identity. B avoids the lavender banners and purple controls repeated across much of its cohort, but its green branding, cream background and restrained cards resemble the strength log and marketplace. Neither is wholly interchangeable; A’s typographic treatment is slightly more distinctive within its cohort.", "displays": [ "/agentic/landing/judge/ballot-1-desktop.png", "/agentic/landing/judge/ballot-1-mobile.png", "/agentic/landing/judge/cohort-plain.png", "/agentic/landing/judge/cohort-kit.png" ] }, { "order": [ "kit", "plain" ], "preference": 5, "dimensions": { "hierarchy": 6, "genericSignals": 5, "productSpecificity": 4, "typography": 5, "spacing": 5, "color": 5, "copy": 3, "states": 3, "mobile": 6, "distinctiveness": 2 }, "reason": "B puts Save beside the dose control and separates ratio from total water clearly; on mobile, saving is visible while A places it below the targets and outside the shown viewport. B’s restrained serif hierarchy and orange action color improve scanning. A offers more concrete copy about grams and clearer unsaved-state feedback. Both retain the pour chart, timed water targets and verified calculator behavior: 18 g produces 288 g and persists after reload on desktop and mobile. B narrowly wins for the main working action.", "cohortReason": "A’s green brewing interface departs substantially from its cohort’s recurring lavender gradient banners, purple actions and dashboard cards. B removes those generic treatments, but its cream background, editorial serif headings, muted green details and fine rules recur across its client portal, research desk, print exchange and marketplace. B is polished and appropriate, while A has greater visual independence within its own cohort.", "displays": [ "/agentic/landing/judge/ballot-2-desktop.png", "/agentic/landing/judge/ballot-2-mobile.png", "/agentic/landing/judge/cohort-kit.png", "/agentic/landing/judge/cohort-plain.png" ] } ], "dimensions": { "hierarchy": 4, "genericSignals": 3.5, "productSpecificity": 4.5, "typography": 3.5, "spacing": 4, "color": 3.5, "copy": 5, "states": 5, "mobile": 3.5, "distinctiveness": 4.5 }, "rounds": { "kit": 1, "plain": 1 }, "images": { "kit": { "desktop": "/agentic/landing/kit/desktop.png", "mobile": "/agentic/landing/kit/mobile.png" }, "plain": { "desktop": "/agentic/landing/plain/desktop.png", "mobile": "/agentic/landing/plain/mobile.png" } }, "checks": { "kit": { "desktop": { "passed": true, "observed": "18 g dose produced 288 g water and survived reload." }, "mobile": { "passed": true, "observed": "18 g dose produced 288 g water and survived reload." } }, "plain": { "desktop": { "passed": true, "observed": "18 g dose produced 288 g water and survived reload." }, "mobile": { "passed": true, "observed": "18 g dose produced 288 g water and survived reload." } } }, "stops": { "kit": "target reached", "plain": "target reached" } }, { "id": "ai-chat", "label": "Research conversation desk", "route": "/chat/field-notes", "task": "A researcher sends a follow-up about the source brief and reads the answer in context. Preserve the thread and its source records.", "result": "Plain", "votes": [ 2, 2 ], "ballots": [ { "order": [ "plain", "kit" ], "preference": 2, "dimensions": { "hierarchy": 2, "genericSignals": 1, "productSpecificity": 2, "typography": 2, "spacing": 2, "color": 2, "copy": 2, "states": 3, "mobile": 2, "distinctiveness": 3 }, "reason": "A connects the conversation to an attached source record, brief excerpt and local-save status, supporting follow-up research in context. Its restrained hierarchy keeps the answer, citation and composer close together. B preserves the same useful production facts but gives substantial space to a gradient title banner and oversized message cards; its source is a plain reference without the visible record context. On mobile, A fits the answer and complete composer onscreen while B pushes the writing area below the visible fold. Both passed sending and reload persistence on desktop and mobile; A additionally makes saved state visible.", "cohortReason": "B repeats the lavender gradient banners, purple accents and rounded white cards seen in its client portal, settlements and rehearsal products. A shares serif headings, muted greens and editorial spacing with its portal, coffee and marketplace cohort, so its identity is not wholly unique. However, its conversation rail, document context panel and compact source citation adapt that vocabulary more specifically to research work.", "displays": [ "/agentic/ai-chat/judge/ballot-1-desktop.png", "/agentic/ai-chat/judge/ballot-1-mobile.png", "/agentic/ai-chat/judge/cohort-plain.png", "/agentic/ai-chat/judge/cohort-kit.png" ] }, { "order": [ "kit", "plain" ], "preference": 6, "dimensions": { "hierarchy": 6, "genericSignals": 6, "productSpecificity": 6, "typography": 6, "spacing": 6, "color": 6, "copy": 6, "states": 5, "mobile": 6, "distinctiveness": 5 }, "reason": "B connects the conversation to a visible source record, proof number and brief excerpt, supporting a researcher’s follow-up in context. Its restrained message hierarchy and compact source link keep the answer readable, while the phone render brings the composer into view sooner than A’s tall cards. A has larger, accessible text but gives disproportionate space to the gradient title and separates the brief from the working context. Both passed sending and reload persistence on desktop and mobile; B additionally makes local saving and brief-informed replies visible, though some supporting text is small.", "cohortReason": "A repeats the lavender gradient banners, purple accents, pale backgrounds and rounded white cards visible in its client portal, settlements and rehearsal cohorts. B replaces that composition with a research sidebar, conversation column and source-record rail. Its serif headings, muted green palette and fine rules still resemble B’s coffee, marketplace and strength interfaces, so its identity is more product-specific without being wholly distinct from its cohort.", "displays": [ "/agentic/ai-chat/judge/ballot-2-desktop.png", "/agentic/ai-chat/judge/ballot-2-mobile.png", "/agentic/ai-chat/judge/cohort-kit.png", "/agentic/ai-chat/judge/cohort-plain.png" ] } ], "dimensions": { "hierarchy": 2, "genericSignals": 1.5, "productSpecificity": 2, "typography": 2, "spacing": 2, "color": 2, "copy": 2, "states": 3, "mobile": 2, "distinctiveness": 3 }, "rounds": { "kit": 1, "plain": 1 }, "images": { "kit": { "desktop": "/agentic/ai-chat/kit/desktop.png", "mobile": "/agentic/ai-chat/kit/mobile.png" }, "plain": { "desktop": "/agentic/ai-chat/plain/desktop.png", "mobile": "/agentic/ai-chat/plain/mobile.png" } }, "checks": { "kit": { "desktop": { "passed": true, "observed": "Sent a source question. The seeded answer survived reload." }, "mobile": { "passed": true, "observed": "Sent a source question. The seeded answer survived reload." } }, "plain": { "desktop": { "passed": true, "observed": "Sent a source question. The seeded answer survived reload." }, "mobile": { "passed": true, "observed": "Sent a source question. The seeded answer survived reload." } } }, "stops": { "kit": "regression", "plain": "target reached" } }, { "id": "mobile-feed", "label": "Print exchange feed", "route": "/", "task": "A printmaker finds the Ink Lab entry, saves the post and sees it on their profile. Keep the real listings and routes.", "result": "Plain", "votes": [ 2, 3 ], "ballots": [ { "order": [ "plain", "kit" ], "preference": 2, "dimensions": { "hierarchy": 3, "genericSignals": 2, "productSpecificity": 3, "typography": 2, "spacing": 3, "color": 3, "copy": 3, "states": 4, "mobile": 4, "distinctiveness": 2 }, "reason": "Both interfaces successfully support opening Ink Lab, saving it and finding it on the profile on desktop and mobile. A has a more intentional print-publication character through serif typography, restrained red accents, small uppercase process labels and outlined save controls. Both preserve meaningful artwork, processes and maker credits. B makes saving more prominent with solid purple buttons and larger mobile controls; A offers a tighter mobile title/action arrangement and larger desktop artwork. Saved-state styling is not visible, so neither receives an advantage for it.", "cohortReason": "B repeats the purple action buttons and broadly neutral interface typography seen across its billing, project, research, fitness and booking cohort, although its serif heading adds some identity. A shares editorial serif typography and muted colors with its coffee, client portal and marketplace peers, but combines them with red ink accents, process captions, thin rules and artwork-led spacing that feel more specific to a print exchange.", "displays": [ "/agentic/mobile-feed/judge/ballot-1-desktop.png", "/agentic/mobile-feed/judge/ballot-1-mobile.png", "/agentic/mobile-feed/judge/cohort-plain.png", "/agentic/mobile-feed/judge/cohort-kit.png" ] }, { "order": [ "kit", "plain" ], "preference": 5, "dimensions": { "hierarchy": 5, "genericSignals": 5, "productSpecificity": 5, "typography": 6, "spacing": 5, "color": 5, "copy": 3, "states": 4, "mobile": 5, "distinctiveness": 5 }, "reason": "Both interfaces verified the complete Ink Lab save-to-profile journey on desktop and mobile. B gives the prints more visual weight, uses more deliberate serif typography and restrained controls, and keeps the mobile save action beside the entry details. A's purple save buttons are easier to spot, and its introductory copy explains the task more directly than B's slogan. B's stronger print presentation earns a slight overall preference; successful behavior is equal.", "cohortReason": "A repeats the purple actions and conventional heading-plus-grid structure seen across its client portal, settlements and booking cohort, although its artwork adds identity. B shares editorial serif headings, muted backgrounds and fine rules with its coffee and marketplace peers, so it also shows a house style. Its red accents, large print reproductions and compact technique labels nevertheless give this exchange a somewhat more specific visual identity.", "displays": [ "/agentic/mobile-feed/judge/ballot-2-desktop.png", "/agentic/mobile-feed/judge/ballot-2-mobile.png", "/agentic/mobile-feed/judge/cohort-kit.png", "/agentic/mobile-feed/judge/cohort-plain.png" ] } ], "dimensions": { "hierarchy": 3, "genericSignals": 2.5, "productSpecificity": 3, "typography": 2, "spacing": 3, "color": 3, "copy": 4, "states": 4, "mobile": 3.5, "distinctiveness": 2.5 }, "rounds": { "kit": 1, "plain": 1 }, "images": { "kit": { "desktop": "/agentic/mobile-feed/kit/desktop.png", "mobile": "/agentic/mobile-feed/kit/mobile.png" }, "plain": { "desktop": "/agentic/mobile-feed/plain/desktop.png", "mobile": "/agentic/mobile-feed/plain/mobile.png" } }, "checks": { "kit": { "desktop": { "passed": true, "observed": "Opened Ink Lab, saved it and located it on the profile." }, "mobile": { "passed": true, "observed": "Opened Ink Lab, saved it and located it on the profile." } }, "plain": { "desktop": { "passed": true, "observed": "Opened Ink Lab, saved it and located it on the profile." }, "mobile": { "passed": true, "observed": "Opened Ink Lab, saved it and located it on the profile." } } }, "stops": { "kit": "target reached", "plain": "target reached" } }, { "id": "fintech", "label": "Fintech settlements", "route": "/", "task": "A finance operator matches an unmatched payment to its bank deposit and confirms the remaining exception count.", "result": "Tied or split", "votes": [ 4, 4 ], "ballots": [ { "order": [ "plain", "kit" ], "preference": 4, "dimensions": { "hierarchy": 4, "genericSignals": 4, "productSpecificity": 4, "typography": 4, "spacing": 4, "color": 4, "copy": 4, "states": 4, "mobile": 4, "distinctiveness": 4 }, "reason": "A and B appear equivalent: both show the $4,710 incoming total, one unmatched payment, TX-091 for $1,840 and corresponding deposit BNK-407 with a Match deposit action. Both verified matching successfully on desktop and mobile, but the supplied evidence does not confirm the remaining count after matching. Both prioritize a large generic welcome banner over the working action. Mobile stacks summary cards with generous spacing while the payment table overflows horizontally, limiting scanning of statuses.", "cohortReason": "Both repeat their cohorts’ purple branding, pastel gradient hero, rounded white cards and centered promotional copy, particularly the developer documentation and rehearsal booking examples. The finance table and deposit references provide relevant content, but neither interface establishes a distinctive finance workflow through layout, typography or density.", "displays": [ "/agentic/fintech/judge/ballot-1-desktop.png", "/agentic/fintech/judge/ballot-1-mobile.png", "/agentic/fintech/judge/cohort-plain.png", "/agentic/fintech/judge/cohort-kit.png" ] }, { "order": [ "kit", "plain" ], "preference": 4, "dimensions": { "hierarchy": 4, "genericSignals": 4, "productSpecificity": 4, "typography": 4, "spacing": 4, "color": 4, "copy": 4, "states": 4, "mobile": 4, "distinctiveness": 4 }, "reason": "A and B are effectively identical. Both show meaningful customer names, payment references, amounts, an unmatched count of 1 and a deposit matching action, but give the generic gradient hero more prominence than the exception workflow. Both successfully match TX-091 to BNK-407 on desktop and mobile; neither provides visible evidence of the remaining count after matching. Both phone tables overflow their cards, compromising readability. Typography, spacing, color and copy offer no material preference.", "cohortReason": "Both targets repeat the centered purple branding, pastel gradient banner, rounded white cards and promotional headline seen in their respective rehearsal booking interfaces. A also shares these patterns with its client portal and research desk; B with its developer documentation. B’s other cohort products more often use editorial typography, muted colors and denser task layouts, but the finance target retains the same generic composition as A. Neither establishes a distinct finance identity.", "displays": [ "/agentic/fintech/judge/ballot-2-desktop.png", "/agentic/fintech/judge/ballot-2-mobile.png", "/agentic/fintech/judge/cohort-kit.png", "/agentic/fintech/judge/cohort-plain.png" ] } ], "dimensions": { "hierarchy": 4, "genericSignals": 4, "productSpecificity": 4, "typography": 4, "spacing": 4, "color": 4, "copy": 4, "states": 4, "mobile": 4, "distinctiveness": 4 }, "rounds": { "kit": 1, "plain": 1 }, "images": { "kit": { "desktop": "/agentic/fintech/kit/desktop.png", "mobile": "/agentic/fintech/kit/mobile.png" }, "plain": { "desktop": "/agentic/fintech/plain/desktop.png", "mobile": "/agentic/fintech/plain/mobile.png" } }, "checks": { "kit": { "desktop": { "passed": true, "observed": "TX-091 matched bank deposit BNK-407." }, "mobile": { "passed": true, "observed": "TX-091 matched bank deposit BNK-407." } }, "plain": { "desktop": { "passed": true, "observed": "TX-091 matched bank deposit BNK-407." }, "mobile": { "passed": true, "observed": "TX-091 matched bank deposit BNK-407." } } }, "stops": { "kit": "regression", "plain": "regression" } }, { "id": "fitness", "label": "Strength training log", "route": "/", "task": "A lifter records the next completed set and sees how many of today’s prescribed sets remain.", "result": "Plain", "votes": [ 3, 3 ], "ballots": [ { "order": [ "plain", "kit" ], "preference": 3, "dimensions": { "hierarchy": 4, "genericSignals": 3, "productSpecificity": 2, "typography": 3, "spacing": 3, "color": 4, "copy": 3, "states": 5, "mobile": 3, "distinctiveness": 3 }, "reason": "Both passed set logging on desktop and mobile, advancing front squat from two to three of four. A gives weight, reps and the next-set prescription greater prominence, keeps eight remaining sets in a compact progress panel, and exposes more of the exercise queue on phone. B has larger set targets and clearer current-exercise labeling; A confusingly labels the active squat 'Up next.' B spends more mobile height on introductory copy and separated progress sections. A narrowly wins through prescription clarity and efficient workout scanning.", "cohortReason": "A shares cream surfaces, muted green accents and restrained borders with its coffee, marketplace and client-portal cohort, but its bold sans-serif exercise names, compact set markers and prescription hierarchy provide a recognizable workout structure. B avoids the gradient banners and emoji headings repeated across its portal, finance and booking examples, yet its purple accents, pale badges and spacious white cards retain more of that cohort's common app styling. Both adapt their shared visual language to the task; A does so somewhat more specifically.", "displays": [ "/agentic/fitness/judge/ballot-1-desktop.png", "/agentic/fitness/judge/ballot-1-mobile.png", "/agentic/fitness/judge/cohort-plain.png", "/agentic/fitness/judge/cohort-kit.png" ] }, { "order": [ "kit", "plain" ], "preference": 5, "dimensions": { "hierarchy": 5, "genericSignals": 5, "productSpecificity": 5, "typography": 5, "spacing": 5, "color": 4, "copy": 5, "states": 3, "mobile": 5, "distinctiveness": 4 }, "reason": "Both successfully advance front squat from two to three completed sets on desktop and mobile. B groups remaining sets and progress clearly, emphasizes load and reps, and places the next set prescription beneath the logging button. Its compact mobile cards expose more upcoming work. A labels the active exercise more accurately: B says “Up next” on the exercise already in progress. A also gives Undo clearer prominence.", "cohortReason": "A shares purple accents and conventional white cards with its billing, finance and booking peers, though its left-aligned workout hierarchy avoids their gradient hero banners. B shares cream backgrounds, green accents, small uppercase labels and editorial typography with its coffee, marketplace and client portal peers. Both adapt those recurring styles with set counters and a prominent logging action; neither has a strong distinctiveness advantage.", "displays": [ "/agentic/fitness/judge/ballot-2-desktop.png", "/agentic/fitness/judge/ballot-2-mobile.png", "/agentic/fitness/judge/cohort-kit.png", "/agentic/fitness/judge/cohort-plain.png" ] } ], "dimensions": { "hierarchy": 3.5, "genericSignals": 3, "productSpecificity": 2.5, "typography": 3, "spacing": 3, "color": 4, "copy": 3, "states": 5, "mobile": 3, "distinctiveness": 3.5 }, "rounds": { "kit": 2, "plain": 1 }, "images": { "kit": { "desktop": "/agentic/fitness/kit/desktop.png", "mobile": "/agentic/fitness/kit/mobile.png" }, "plain": { "desktop": "/agentic/fitness/plain/desktop.png", "mobile": "/agentic/fitness/plain/mobile.png" } }, "checks": { "kit": { "desktop": { "passed": true, "observed": "Logging a set advanced the front squat from two to three of four." }, "mobile": { "passed": true, "observed": "Logging a set advanced the front squat from two to three of four." } }, "plain": { "desktop": { "passed": true, "observed": "Logging a set advanced the front squat from two to three of four." }, "mobile": { "passed": true, "observed": "Logging a set advanced the front squat from two to three of four." } } }, "stops": { "kit": "target reached", "plain": "target reached" } }, { "id": "docs", "label": "Developer tool documentation", "route": "/", "task": "A maintainer finds the correct install and health-check commands for their package manager.", "result": "Kit", "votes": [ 6, 7 ], "ballots": [ { "order": [ "plain", "kit" ], "preference": 6, "dimensions": { "hierarchy": 6, "genericSignals": 7, "productSpecificity": 6, "typography": 6, "spacing": 6, "color": 6, "copy": 6, "states": 4, "mobile": 6, "distinctiveness": 6 }, "reason": "B makes installation and workspace checking a clear numbered sequence, with prominent commands, readable terminal panels and useful workspace-root guidance. Its headline describes the actual maintenance task; A spends substantial space on a vague welcome gradient. Both expose the package-manager selector and verified pnpm switching works on desktop and mobile. B keeps the install command on one line on the phone, while A wraps the package name. A does make exit-code meanings visible earlier, a useful operational detail.", "cohortReason": "A repeats the centered purple branding, pastel gradient welcome banner, emoji and rounded white cards seen in its finance and rehearsal products, with similar framing in the client portal. B shares restrained typography and muted surfaces with its print, coffee and marketplace cohort, but its numbered documentation flow, compact runtime prerequisite and dark command panels give Latch a more specific identity. Its distinction serves command discovery rather than adding decoration.", "displays": [ "/agentic/docs/judge/ballot-1-desktop.png", "/agentic/docs/judge/ballot-1-mobile.png", "/agentic/docs/judge/cohort-plain.png", "/agentic/docs/judge/cohort-kit.png" ] }, { "order": [ "kit", "plain" ], "preference": 1, "dimensions": { "hierarchy": 1, "genericSignals": 1, "productSpecificity": 2, "typography": 2, "spacing": 2, "color": 2, "copy": 2, "states": 4, "mobile": 2, "distinctiveness": 2 }, "reason": "A prioritizes the maintainer’s workflow with numbered install and health-check steps, a clear package-manager selector, and readable terminal commands. Its mobile install command stays on one line, while B wraps it and spends substantial space on a vague welcome banner. B does provide useful exit-code guidance. Both correctly update the command after selecting pnpm on desktop and mobile, so interaction evidence is tied.", "cohortReason": "A shares restrained typography and terminal styling with its cohort but adapts them to CLI documentation through numbered steps, workspace-root labels, and command panels. B repeats the centered purple gradient banner, rounded white cards, and promotional phrasing visible in its finance and rehearsal products. A’s composition is more specific to the task, though its purple accents also recur across its cohort.", "displays": [ "/agentic/docs/judge/ballot-2-desktop.png", "/agentic/docs/judge/ballot-2-mobile.png", "/agentic/docs/judge/cohort-kit.png", "/agentic/docs/judge/cohort-plain.png" ] } ], "dimensions": { "hierarchy": 6.5, "genericSignals": 7, "productSpecificity": 6, "typography": 6, "spacing": 6, "color": 6, "copy": 6, "states": 4, "mobile": 6, "distinctiveness": 6 }, "rounds": { "kit": 1, "plain": 1 }, "images": { "kit": { "desktop": "/agentic/docs/kit/desktop.png", "mobile": "/agentic/docs/kit/mobile.png" }, "plain": { "desktop": "/agentic/docs/plain/desktop.png", "mobile": "/agentic/docs/plain/mobile.png" } }, "checks": { "kit": { "desktop": { "passed": true, "observed": "Selecting pnpm displayed the correct install command." }, "mobile": { "passed": true, "observed": "Selecting pnpm displayed the correct install command." } }, "plain": { "desktop": { "passed": true, "observed": "Selecting pnpm displayed the correct install command." }, "mobile": { "passed": true, "observed": "Selecting pnpm displayed the correct install command." } } }, "stops": { "kit": "target reached", "plain": "regression" } }, { "id": "marketplace", "label": "Used tools marketplace", "route": "/", "task": "A bookbinder saves an inspected listing to their shortlist and checks the pickup terms.", "result": "Plain", "votes": [ 3, 3 ], "ballots": [ { "order": [ "plain", "kit" ], "preference": 3, "dimensions": { "hierarchy": 3, "genericSignals": 3, "productSpecificity": 2, "typography": 4, "spacing": 4, "color": 3, "copy": 3, "states": 4, "mobile": 3, "distinctiveness": 4 }, "reason": "A slightly wins: equipment illustrations, inspection badges and location details make the inspected Adana chase easy to identify, and its full-width mobile save button is prominent. B offers clearer desktop pickup terms and a more compact text inventory, but loses useful equipment recognition. Both expose dimensions, condition, prices and pickup locations; verified saving and saved feedback work on desktop and mobile for both.", "cohortReason": "A repeats its cohort’s muted green, cream, editorial serif headings and restrained cards, particularly the print exchange, client portal and training tracker. Its equipment drawings and collection details give that vocabulary a relevant purpose. B avoids the lavender banners and purple controls common across much of its cohort, but closely resembles its coffee calculator and print exchange in cream palette, serif typography and thin rules. Neither establishes a wholly unique visual system.", "displays": [ "/agentic/marketplace/judge/ballot-1-desktop.png", "/agentic/marketplace/judge/ballot-1-mobile.png", "/agentic/marketplace/judge/cohort-plain.png", "/agentic/marketplace/judge/cohort-kit.png" ] }, { "order": [ "kit", "plain" ], "preference": 5, "dimensions": { "hierarchy": 5, "genericSignals": 4, "productSpecificity": 5, "typography": 4, "spacing": 4, "color": 5, "copy": 3, "states": 4, "mobile": 5, "distinctiveness": 4 }, "reason": "B makes the inspected chase easier to identify through equipment illustrations and gives its save action a prominent full-width mobile button. Its desktop pickup panel also repeats condition and collection details. A is more compact, exposes the next listing sooner on mobile, and uses clearer pickup language: B’s “A local handover” is less direct than “Pickup terms.” Both successfully save the listing and visibly mark it saved on desktop and mobile; neither demonstrates fuller pickup arrangements.", "cohortReason": "A’s restrained list and serif heading differ from its cohort’s frequent lavender banners and white dashboard cards, although its warm editorial styling resembles the coffee and print products. B’s olive palette, serif headlines, small uppercase labels and fine rules recur across its coffee, print and project interfaces. Its equipment drawings add useful product evidence, but the surrounding editorial treatment remains recognizable as a cohort style. Neither has a decisive distinctiveness advantage.", "displays": [ "/agentic/marketplace/judge/ballot-2-desktop.png", "/agentic/marketplace/judge/ballot-2-mobile.png", "/agentic/marketplace/judge/cohort-kit.png", "/agentic/marketplace/judge/cohort-plain.png" ] } ], "dimensions": { "hierarchy": 3, "genericSignals": 3.5, "productSpecificity": 2.5, "typography": 4, "spacing": 4, "color": 3, "copy": 4, "states": 4, "mobile": 3, "distinctiveness": 4 }, "rounds": { "kit": 1, "plain": 1 }, "images": { "kit": { "desktop": "/agentic/marketplace/kit/desktop.png", "mobile": "/agentic/marketplace/kit/mobile.png" }, "plain": { "desktop": "/agentic/marketplace/plain/desktop.png", "mobile": "/agentic/marketplace/plain/mobile.png" } }, "checks": { "kit": { "desktop": { "passed": true, "observed": "The Adana chase entered the shortlist and the control marked it saved." }, "mobile": { "passed": true, "observed": "The Adana chase entered the shortlist and the control marked it saved." } }, "plain": { "desktop": { "passed": true, "observed": "The Adana chase entered the shortlist and the control marked it saved." }, "mobile": { "passed": true, "observed": "The Adana chase entered the shortlist and the control marked it saved." } } }, "stops": { "kit": "target reached", "plain": "target reached" } }, { "id": "booking", "label": "Rehearsal booking", "route": "/", "task": "A quartet selects an available rehearsal time and confirms its booked slot. A full slot cannot be booked.", "result": "Tied or split", "votes": [ 4, 4 ], "ballots": [ { "order": [ "plain", "kit" ], "preference": 4, "dimensions": { "hierarchy": 4, "genericSignals": 4, "productSpecificity": 4, "typography": 4, "spacing": 4, "color": 4, "copy": 4, "states": 4, "mobile": 4, "distinctiveness": 4 }, "reason": "A and B appear visually identical at both sizes. Both clearly show Room B, four seats, upright piano, £24/hour, a time selector and Reserve room action. Phone layouts remain readable and usable, though the large promotional banner takes prominence over availability. Verified interactions equally confirm booking at 14:00 and prevent booking full slots.", "cohortReason": "Both repeat the centered purple branding, pastel gradient hero, broad promotional copy and rounded white cards seen in their respective finance dashboards. A also closely resembles its developer documentation cohort; B shares treatments with its client portal and research desk. Rehearsal details provide useful specificity, but neither target develops a distinct booking-oriented visual identity.", "displays": [ "/agentic/booking/judge/ballot-1-desktop.png", "/agentic/booking/judge/ballot-1-mobile.png", "/agentic/booking/judge/cohort-plain.png", "/agentic/booking/judge/cohort-kit.png" ] }, { "order": [ "kit", "plain" ], "preference": 4, "dimensions": { "hierarchy": 4, "genericSignals": 4, "productSpecificity": 4, "typography": 4, "spacing": 4, "color": 4, "copy": 4, "states": 4, "mobile": 4, "distinctiveness": 4 }, "reason": "A and B are visually equivalent on desktop and phone. Both clearly present Room B, four seats, upright piano, £24/hour, a time selector and Reserve room action. Both give substantial prominence to a generic gradient headline, while booking details remain legible and mobile controls usable. Verified interactions equally demonstrate disabled full slots and successful booking at 14:00.", "cohortReason": "Both repeat the centered purple branding, pastel gradient hero, rounded white cards and neutral sans-serif typography visible in the fintech settlements cohort. A also closely resembles its client portal and research desk; B resembles its developer documentation. B’s other cohort products show more varied editorial layouts, but the target itself has no visible distinguishing advantage. Rehearsal details provide useful specificity without establishing a distinctive booking identity.", "displays": [ "/agentic/booking/judge/ballot-2-desktop.png", "/agentic/booking/judge/ballot-2-mobile.png", "/agentic/booking/judge/cohort-kit.png", "/agentic/booking/judge/cohort-plain.png" ] } ], "dimensions": { "hierarchy": 4, "genericSignals": 4, "productSpecificity": 4, "typography": 4, "spacing": 4, "color": 4, "copy": 4, "states": 4, "mobile": 4, "distinctiveness": 4 }, "rounds": { "kit": 0, "plain": 0 }, "images": { "kit": { "desktop": "/agentic/booking/kit/desktop.png", "mobile": "/agentic/booking/kit/mobile.png" }, "plain": { "desktop": "/agentic/booking/plain/desktop.png", "mobile": "/agentic/booking/plain/mobile.png" } }, "checks": { "kit": { "desktop": { "passed": true, "observed": "Full slot disabled. Room B booked at 14:00." }, "mobile": { "passed": true, "observed": "Full slot disabled. Room B booked at 14:00." } }, "plain": { "desktop": { "passed": true, "observed": "Full slot disabled. Room B booked at 14:00." }, "mobile": { "passed": true, "observed": "Full slot disabled. Room B booked at 14:00." } } }, "stops": { "kit": "target reached", "plain": "target reached" } } ] }