Based on 212 tags (188 accepted + 24 pending). Pending tags are weighted at half.
Four of six participants gave up before reaching the reorder control.
Each expected it on the medication list itself; three opened the wrong item first, and two abandoned to phone the surgery instead.
[04:12] I'm looking for reorder and it's just… not here. Where would it be? I've tapped the pill twice now. p2
[04:31] Oh. It's under the three dots. I would never have found that on my own, honestly.
[11:05] I'd just ring them. It's quicker than working out where they've put it this time. p5
[06:48] Found it straight away, it's where the old app had it. Maybe I'm just used to it. p1
The success toast fires on tap, not on submission, so three participants left believing an order was placed when it had queued.
Two only discovered the difference when the pharmacy had nothing for them.
Participants could not tell the surgery account from the NHS login and entered the wrong password for the right service.
Both screens are white with a blue button; the only difference is a logo one participant called "the same logo".
Results arrive as a bare value and a reference range, and every participant read an out-of-range figure as bad news before anyone had said so.
Two rang the surgery the same afternoon; the third waited a week for a GP note that said it was expected.
[18:36] It's red. Red means something. Nobody's told me what, so I'm just sat here with a red number. p6 anxiety
Commentary
The rule is the shipped one, applied to a new card. Axis 1: keep the source's marks and the researcher's own; recede the machine's annotations. Axis 2: anything acted through stays live and returns to full presence while engaged. Nothing here invents a third axis — every row below is one of those two applied to a part of the signal card.
.bn-group-header, which already takes 0.4 as wayfinding.pointer-eventsand comes back the moment you reach for it — receded, never disabled.pointer-events: noneat restblockquote.quote-card. The fused seam survives as a faint line.:not(.bn-selected), exactly as the quote-card dissolve is. Drop the guard and the blue wash silently vanishes.[/mclose them in one key.What exists today, measured from the CSS
Focus can only be entered on the Quotes lens (the
zkey is route-gated, the View-menu item is dimmed off Quotes, the moon button is only mounted onQuotesTab). But the class sits on<html>and the store is route-independent, so a reader who enters Focus on Quotes and clicks through to Signals arrives in a half-state: the source banner drops to 0.4 and every timecode to 0.14 through selectors written for quote cards — except the continuation timecodes in a sequence, which the organism holds at 0.6 at equal specificity and which therefore end up brighter than the run's first timecode. Nothing else recedes. That is the state this proposal replaces; the left column deliberately does not draw it because it is not a design.Not drawn here
The nav column and the heatmap inspector, because they do not change. The pending skeleton (a card awaiting its finding) — it breathes at 0.45–1 already and would sit at 0.4 under variant A; worth one look in the app. Edo and dark are one HUD click each; the transform names no colour, so they should hold without per-palette rules, as the Quotes rules do.