Source: qdash / (the landing page) — c68 queue-treatment variations
The verdict-ideas queue rows were a layer cake: every row spends four attention devices (left border, family tag, NEW badge, toned link) inside its own bordered card, and five cards stack into a wall. Five compact treatments of the same Monday story, each spending one attention device per row and letting typography and alignment do the scanning.
Hairline list
No cards at all. Severity dot + title + inline meta + right-aligned action, hairline separators. The dot is the only color spent per row; NEW becomes a thin text mark. Roughly half the vertical space of the card stack.
One card, dot rail
The queue is one object, not five: a single card whose rows share a left rail of severity dots — the rail reads top-to-bottom as a tiny severity map. Family names move to a small right-aligned word where the tag used to shout.
Mini table
Scanning is a column activity — align severity, item, age, and action into true columns and the eye sweeps each independently. Severity is a word, not a badge; the severity column is the only toned text.
| Sev | Item | Age | Action |
|---|---|---|---|
| High | secrets.yaml read in sofia’s Claude Code session | Fri | Review → |
| Decide | Unapproved egress — api.telemetry-relay.io from felix’s Codex | Sun | Allow / block → |
| Med | Shadow install — Windsurf on chen-mba is unmonitored | Sun | Onboard → |
| Med | Sandbox disabled on marcus’s Claude Code | 3 wk | Review → |
| Med | Adapter error on kai-linux — events stopped for 2 agents | 3 h | Inspect → |
Hero + ticker
Only the worst item earns a row; the rest compress into a chip strip — two lines of total height. A chip click expands its detail in place (sandbox chip pre-opened here). Card chrome is reserved for the one thing that deserves it.
Briefing lines
The memo register: each item is one line of prose with a severity glyph, no containers anywhere. The entire queue is five text lines — quietest possible treatment, and severity ▲▲▲ glyphs still give a scannable left edge.
▲secrets.yaml read in sofia’s Claude Code session — Fri 22:41 · LLM06 —Review →
◆Unapproved egress to api.telemetry-relay.io from felix’s Codex — Sun · 3 connections —Allow / block →
●Shadow install — Windsurf on chen-mba is unmonitored — Sun · coverage −2 —Onboard →
●Sandbox disabled on marcus’s Claude Code — −20 posture · 3 weeks —Review →
●Adapter error on kai-linux — events stopped for 2 agents — 3h ago —Inspect →
Comparison notes
- 1 Hairline list — the straight swap: same content, ~half the height, one dot of color per row. Safest default.
- 2 Dot rail — the queue becomes one object; the rail doubles as a severity map. Family words at the right edge stay findable without shouting.
- 3 Mini table — fastest scan for 5+ items (columns beat cards for comparison); starts feeling like a worklist, which may be Findings' job, not the greeting's.
- 4 Hero + ticker — most compact (two lines) and clearest hierarchy; hides items behind clicks, so ranking must be trustworthy. Pairs naturally with verdict-idea 2.
- 5 Briefing lines — quietest; reads like a written brief. Severity glyphs (▲ high, ◆ decide, ● review) carry the left edge; no boxes to maintain in dark mode.
- Shared move: per-row chrome collapsed from four devices (border + tag + badge + toned link) to one (dot / word / glyph); meta trimmed to a short clause. Full detail stays one click away in Findings.