[Epic] What's That Card: full-page repass (layout, question content, feed intelligence)

[WilfordGrimley]

Owner walkthrough finding, verbatim: "I am not satisfied with many elements. We have done a lot of patchwork and not a full repass in a while."

This epic tracks a full-page repass of the "What's That Card" surface (frontend/src/features/questionFeed/QuestionFeed.tsx, frontend/src/features/printingTags/cardPanel.tsx, frontend/src/features/attributeChips/*, frontend/src/components/ArtistSupportLink.tsx, MPCAutofill/cardpicker/question_feed.py), split into three tracks. All nine children below were individually audited against current code and each carries its own code-evidence + verdict.

Track A — Layout / space

Ships as ONE coherent repass, not six separate patches — several of these issues touch the same shared styled-components (cardPanel.tsx's CandidateButton/ArtPlaceholder/ZoomableThumbnail, QuestionFeed.tsx's Btn/ActionGrid/SubjectArtTitle), so fixing them independently would mean multiple diffs churning the same files. Each child issue has its own "Layout-pass coupling" note flagging exactly which other items it shares surface area with.

• What's That: hover-zoom on card/illustration is clipped by its container frame #705 hover-zoom on card/illustration is clipped by its container frame
• What's That: chips don't collapse their space when a selection invalidates them #707 chips don't collapse their space when a selection invalidates them
• What's That: reference image title overlays the bottom corner of the image #708 reference image title overlays the bottom corner of the image
• What's That: replace the MTGAC applet with a compact link to the artist's MTGAC page #709 MTGAC applet should become a compact link, not a full embed-sized panel
• What's That: reference card should stay visible while scrolling at every viewport; snip page uses vertical space poorly #710 reference card should stay visible while scrolling at every viewport; page uses vertical space poorly (the organizing thesis for this track — the others are symptoms of the same inattention to space)
• What's That: answer-row control hierarchy — the Yes button is oversized #711 answer-row control hierarchy — the Yes button is oversized

Track B — Question content

• What's That: illustration questions render every candidate in a group instead of one representative image (first group also hidden) #706 illustration questions render every candidate in a group instead of one representative image (plus a second, distinct "first group hidden" symptom under the same title — see that issue for whether the two share one root cause)

Track C — Feed intelligence

• What's That: 'Not sure' and 'Skip' are the same action — differentiate and record ambiguity #712 "Not sure" and "Skip" are the same action today — differentiate them and record ambiguity as real data about the image
• What's That: users are re-asked questions they have already answered on the same image #713 users are re-asked questions they've already answered on the same image

A broader information-gain-ordered question-selection redesign is planned separately and is explicitly out of scope for this epic and its children — #713 is the concrete, immediately-actionable slice of that larger direction, not the redesign itself.

Audited against ProxyPrints/ProxyPrints.github.io@17345a48050cd6734f312d033b6ef1ed4f1af014 (origin/master, 2026-08-05).

Children

1. What's That: hover-zoom on card/illustration is clipped by its container frame #705 — hover-zoom on card/illustration is clipped by its container frame
2. What's That: illustration questions render every candidate in a group instead of one representative image (first group also hidden) #706 — illustration questions render every candidate in a group instead of one representative image (first group also hidden)
3. What's That: chips don't collapse their space when a selection invalidates them #707 — chips don't collapse their space when a selection invalidates them
4. What's That: reference image title overlays the bottom corner of the image #708 — reference image title overlays the bottom corner of the image
5. What's That: replace the MTGAC applet with a compact link to the artist's MTGAC page #709 — replace the MTGAC applet with a compact link to the artist's MTGAC page
6. What's That: reference card should stay visible while scrolling at every viewport; snip page uses vertical space poorly #710 — reference card should stay visible while scrolling at every viewport; page uses vertical space poorly
7. What's That: answer-row control hierarchy — the Yes button is oversized #711 — answer-row control hierarchy — the Yes button is oversized
8. What's That: 'Not sure' and 'Skip' are the same action — differentiate and record ambiguity #712 — "Not sure" and "Skip" are the same action — differentiate and record ambiguity
9. What's That: users are re-asked questions they have already answered on the same image #713 — users are re-asked questions they have already answered on the same image
10. Make AI-generated art and external-IP human-votable attributes #719 — Make AI-generated art and external-IP human-votable attributes
11. What's That: the Level 1 Yes button is still visually dominant — block width and accent colour, not font size #740 — the Level 1 Yes button is still visually dominant — block width and accent colour, not font size
12. What's That: the subject card's title label sits inside the artwork's bounding box, covering its bottom edge #741 — the subject card's title label sits inside the artwork's bounding box, covering its bottom edge
13. What's That: exclusion-group chip rows stretch to the full panel width, with the Yes/No controls stranded at the far right #742 — exclusion-group chip rows stretch to the full panel width, with the Yes/No controls stranded at the far right
14. What's That: the attribute chip panel renders two independent exclusion groups as one unlabelled list #743 — the attribute chip panel renders two independent exclusion groups as one unlabelled list
15. What's That: at Level 1 the pinned subject card has no scroll slack, so sticky never engages #744 — at Level 1 the pinned subject card has no scroll slack, so sticky never engages
16. What's That: illustration groups at Level 2 stack vertically instead of flowing side by side #745 — illustration groups at Level 2 stack vertically instead of flowing side by side
17. What's That: illustration and Scryfall artwork render at too small a default size, inside a card-shaped frame that does not match their aspect #746 — illustration and Scryfall artwork render at too small a default size, inside a card-shaped frame that does not match their aspect
18. What's That: the artist's name appears more than once, and the MTGAC disclosure should live inside the link button #747 — the artist's name appears more than once, and the MTGAC disclosure should live inside the link button
19. What's That: a rejected Level 1 suggestion is permanently removed from the Level 2 candidate grid #748 — a rejected Level 1 suggestion is permanently removed from the Level 2 candidate grid

---

Owner decisions (2026-08-05 review)

Owner's strategic framing for the epic:

1. The redesign is an INVERSION. Derive everything derivable from stored evidence first; route human questions ONLY to where the machine abstained or where channels conflict. The feed's job shifts from "ask about everything" to "ask only where automation couldn't resolve it."
2. "Resolution" does NOT require identifying an exact printing. Confirming machine-derived data on other high-value fields is equally a resolution. This decouples the confirmation goal from the printing-identification coverage goal — a question that confirms, say, a border/frame/bleed field the machine already leans toward counts as a resolved question in its own right, not merely a step toward identifying a printing.
3. The 51% consensus gate protects soundness and stays unchanged. No change to the existing threshold mechanics.
4. Difficulty may be user-selectable or ramp per session. Owner leans undecided between the two; session-ramp (difficulty increases within a session rather than being a persistent user setting) is the working default until decided otherwise.
5. New children added to this epic (filed 2026-08-05):
   • What's That: contextual report button — flag an image for moderators against relevant tags, with optional per-user hide #714 — "What's That: contextual report button — flag an image for moderators against relevant tags, with optional per-user hide"
   • What's That: some controls require two clicks to register #715 — "What's That: some controls require two clicks to register"
   • What's That: data-driven question-selection policy (expected information gain per question) #716 — "What's That: data-driven question-selection policy (expected information gain per question)" — this is the data-driven spec for the broader information-gain-ordered selection redesign this epic's Track C children (What's That: 'Not sure' and 'Skip' are the same action — differentiate and record ambiguity #712/What's That: users are re-asked questions they have already answered on the same image #713) already flag as "planned separately... out of scope for this epic."
   • Make AI-generated art and external-IP human-votable attributes #719 — "Make AI-generated art and external-IP human-votable attributes" (filed 2026-08-05, addended 2026-08-06 with the owner's context-dependent-surfacing decision and a code-verified finding on the SENSITIVE-tag consensus gate)

---

Existing spec reconciliation (2026-08-05)

Spec: docs/proposals/mockups/wtc-rebuild/SPEC-wtc-rebuild.md (+ companion wtc-mockup.html
in the same dir). Verdict: CURRENT, and implemented for its own scope — shipped via PR #446
(2026-07-24); the container-first hero (WtcHero/Subject/QPanel, @container hero,
--wtc-* tokens) is still live in QuestionFeed.tsx as of 17345a4 (2026-08-05), and #705/#710's
own code-evidence sections independently cite matching structure. Scope caveat: the spec is a
visual/layout/token rebuild only — it does not address question-content logic (Track B) or
backend feed-intelligence/vote logic (Track C).

Issue	Classification	Note
#704 (this epic)	MIXED	Track A (layout) largely ALREADY SPECIFIED; Track B/C largely NOT COVERED — spec never reaches question-content or feed-intelligence logic
#705	NOT COVERED	spec's cardPanel.tsx file-change row doesn't address ZoomableThumbnail/overflow clipping
#706	NOT COVERED	spec doesn't address illustration-group tile-count rendering
#707	CONFLICTS (sub-symptom A) / NOT COVERED (sub-symptom B)	spec §6 checklist keeps AttributeChipPanel "opt-in... behavior-preserved"; owner's 2026-08-05 decision wants it shown automatically by question type
#708	CONFLICTS	spec §1c BINDING "subject art title" row prescribes exactly the bottom-gradient overlay #708 reports as a defect
#709	NOT COVERED	spec never mentions ArtistSupportLink/MTGAC
#710	CONFLICTS	spec §3 "Page scroll model" (WD4) retires sticky/bounded-height in favor of an ordinary scrolling document; issue demands the reference card stay sticky-visible at every viewport. Issue does cite the spec's container-first policy (§185-190/198-206) as the implementation method to use for the eventual fix — agreement on method, conflict on the core mechanism
#711	NOT COVERED	spec's .btn.big token exists generically; nothing binds the Level-1 Yes button to combining .big+.block specifically
#712	NOT COVERED	spec's §2 "preserved flow" documents current Not-Sure=no-vote behavior as frozen for the visual rebuild's contract, but doesn't forbid a future backend change — out of spec's scope
#713	NOT COVERED	spec's Annex A explicitly disclaims backend served-mix/exclusion logic as "not built by this design"
#714	NOT COVERED	no mention of reporting/flagging UI
#715	NOT COVERED	no mention of double-click/interaction bugs
#716	NOT COVERED	no mention of question-selection ranking; Annex A's served-mix seam is adjacent but distinct

Gap found in the other direction: spec's Annex A calls for a backend seam — logging served-mix
composition per session in question_feed.py so click-latency/agreement-rate can be correlated
against a session's easy-question exposure — explicitly "not built by this design." No issue on
this board (#704-#716) tracks that ask.