{
  "authorship": "Five self-authored fictional cases, not vendor results or actual observations.",
  "cases": [
    {
      "id": 1,
      "text": " 1. Two tiny caption buttons in a crowded toolbar  Fictional input: Mira builds a player with adjacent caption and translation buttons. Each clickable box is 18 by 18 CSS pixels, with centers 20 CSS pixels apart horizontally. The glyphs are crisp, so the design review initially concentrates on appearance. In this exercise, however, each activation area is smaller than the baseline in both dimensions.  Original reasoning: two centered circles with radius 12 have a combined radius of 24; centers only 20 apart make them overlap. That pair cannot use the spacing route described above. This is geometry on invented values, not a browser observation. The record should state the two region sizes, center distance and the route being evaluated, rather than merely saying that the buttons look small.  Proposed change: enlarge the clickable boxes to at least the baseline dimensions and lay them out without overlap. Keep the glyphs visually compact if needed. Pending check: activate both buttons around their edges in the revised layout and verify that neither triggers the neighboring action. Do not mark the proposal passed before that check. "
    },
    {
      "id": 2,
      "text": " 2. Small language options with deliberate separation  Fictional input: Ellis uses two 20 by 20 CSS pixel language targets. Their centers are 32 CSS pixels apart on one row. The surrounding area is otherwise empty in the simplified drawing. The targets remain undersized even though they are more separated than Mira's buttons.  Original reasoning: the two 24-diameter circles do not intersect because 32 exceeds 24. That establishes only the circle-to-circle relationship in this invented pair. A real menu must also be checked for intersection with other targets, including a close button or a larger neighboring row. One favorable pair does not establish that the entire menu qualifies for the spacing exception.  Proposed change: consider full-width menu rows with comfortable activation areas to reduce dependence on narrow gaps. Pending check: inspect the expanded menu at the actual shipping width, map every neighboring target and record its geometry. Save a separate state if the language list wraps. Do not transfer the simplified drawing's conclusion to that changed arrangement. "
    },
    {
      "id": 3,
      "text": " 3. A duplicate action with a larger control  Fictional input: Rowan places a small caption shortcut over the video and a larger caption control below it. The lower control is on the same page. In the design brief both controls toggle captions, but the implementation has not been reviewed. The shortcut is convenient while the lower control has an explicit label.  Original reasoning: the equivalent-control exception requires a qualifying alternative for the same function. A general settings button is not automatically equivalent just because it is large. The reviewer must establish what action the alternative actually makes available and that its target meets the criterion. Here the brief is an intention, so the worksheet should say equivalence is proposed, not confirmed.  Proposed change: retain the clear lower caption control and document the relationship. Pending check: use both paths to enable and disable the same caption track, verify their states stay synchronized, and inspect the lower control's actual clickable dimensions. If one path only opens an unrelated help page, revise the equivalence claim. Keep that functional finding separate from the size measurement. "
    },
    {
      "id": 4,
      "text": " 4. A transcript link inside a sentence  Fictional input: Jules adds the sentence \u201cRead the transcript for this song\u201d below a player, with only the transcript phrase linked. A compact standalone download icon sits nearby. The mockup author wants to apply the inline exception to both because they share the same visual row.  Original reasoning: placement on one row does not make every control an inline text target. The sentence link has a different context from the independent icon. Record the sentence and its text flow, then examine the separate download control on its own. Do not extend an exception from one element to every neighbor in the surrounding container.  Proposed change: keep the meaningful sentence link, and give the standalone download action a clear activation area and label. Pending check: verify where each target begins and ends after layout changes. Confirm that selecting nearby text does not unexpectedly activate the download. A text-flow exception does not remove the practical value of making a link easy to recognize, nor does it replace the separate keyboard review. "
    },
    {
      "id": 5,
      "text": " 5. Browser controls replaced by custom artwork  Fictional input: Devon begins with unmodified browser-supplied video controls. Later, a designer draws a custom toolbar over the player and hides the original controls. The project note still says that all controls use the user-agent exception because the media element itself is native HTML.  Original reasoning: the relevant question is who determines the target size and whether the author modified it. A native media element underneath a custom toolbar does not establish that the custom targets are browser-determined. The review record should identify the actual visible control implementation in each revision, rather than carrying an old exception forward indefinitely.  Proposed change: split the record into the original browser-controls revision and the new custom-controls revision. Leave the former's browser-dependent details recorded as such; measure the latter's activation areas. Pending check: confirm which controls are exposed in the delivered page, then review their sizes, neighbors and actions. Do not describe an uninspected browser control as measured, or call custom styling exempt solely because playback uses a standard video element. "
    }
  ]
}