Atom

Record

A count of what the agent got right on one tenant. 88 instances across 16 screens, in the third column of a fleet row. It is the reason the fleet is the differentiator rather than a page of logos.

Design principle 2 says a number names its claim, its scope and its window and comes with an absolute count. voice/docs/voice.md forbids a bare percentage in as many words, which is why it reads 31 of 36 and never 86%.

Anatomy

dark, shipped

Meridian Healthcontain endpoint31 of 36was 34 of 36
Halden Freightinvestigate4 of 4too few to compare

light, the pair

Meridian Healthcontain endpoint31 of 36was 34 of 36
Halden Freightinvestigate4 of 4too few to compare

Mono and right aligned, and both are the same decision: a column of counts is compared downwards rather than read across, so the digits have to line up and the units have to sit on one axis. It takes the reading ink rather than the secondary ink its neighbours use, because the record is the thing the row exists to show.

The bench is 288px wide because the pane is. The detail pane is 320px below 1560 and its content is 288, and at that width the tenant name wraps to two lines and the whole row is two lines tall. A bench wider than the pane would show a component that does not occur: design/kit/checks/geometry.mjs reported the row 2.19px short until the bench was narrowed to the real thing.

Variants

AxisValuesThe rule
headbody 72, head 16The row's decision, not the cell's. In the head row of the fleet the cell is a column label rather than a count and takes the head's type. That rule lives in frow--head, which is a molecule: rec does not know whether it is in a head
direction–Prohibited, and it is the strongest prohibition on this page. A green count rising and a red count falling is the obvious move. The accent already does four jobs and no fifth; the severity ramp is spoken for; and above all a record painted good or bad decides for the analyst what she is being asked to judge
size–It is 12px, the row's size, because it is read in a column of rows rather than on its own

When to use it

In a fleet row, for the agent's record on one tenant, and nowhere else. This is the concrete element that carries the strategic dimension: CLAUDE.md names calibrated trust as the only surviving differentiator, and says the fleet has to read at a glance rather than on a configuration page. A column of these is that glance.

Where she meets it. Whenever the detail pane is at rest, which is most of the shift. The fleet is the resting state of the right pane rather than a menu item, so this column is in front of her without being opened, and the riskiest assumption in the whole project is that reading it makes her faster rather than giving her forty things to remember.

Rule and anti-rule

Do

31 of 36was 34 of 36

The count with its denominator, and the trend under it. Both are absolute, both name their window, and neither is a percentage.

Do not

86%

Never a bare percentage. 86% of four decisions and 86% of four hundred are the same string and not the same evidence, and the analyst is being asked to decide how much rope to give an agent on that number.

States

It has none, and it is not interactive. The fleet row has a hover; the count sits on the new ground and changes nothing.

dark, shipped

Meridian Healthcontain endpoint31 of 36was 34 of 36
Halden Freightinvestigate4 of 4too few to compare

light, the pair

Meridian Healthcontain endpoint31 of 36was 34 of 36
Halden Freightinvestigate4 of 4too few to compare

What it reads, and where it stands

RoleSurfaceDarkLight
--text-primaryink 4.5:1

Copy this

<span class="rec">31 of 36<span class="was">was 34 of 36</span></span>

Where it stands