One continuous dive on an infinite canvas: from 38 dashboards as clouds of dots, into one cluster, one chart, one data point — the single wrong number Acme Pulse found — and all the way back out as every dot turns teal. A zoom pill, breadcrumb and minimap keep the viewer oriented at every altitude.
readability ●●●trust ●●●effort high
38 dashboards. Where is the problem?
checkout
signups
latency
errors
revenue
3%· fit all
1 · HOOK0:00–0:08
The canvas drifts right→left at 3 % zoom (slow constant pan, 8 px/frame) as 38 dot-clusters fade in one per 1/16 note; one cluster flickers coral for a single beat on the bar-2 downbeat — nobody would see it at this altitude.
Acme Pulse
12%· 2,418 events/sec
2 · REVEAL0:08 (drop)
On the drop every dot flies from its cluster to a point on the pulse line (0.7 s expo.inOut, stagger by distance) so the whole dataset draws the Acme mark; the coral dot lands last on the downbeat and the lockup plate fades up beneath it.
Product › Checkout › conversion ›09:42
38 dashboards
checkout
conversion · 1 h
Checkout conversion −18%
since 09:42 · after deploy #2231
Roll backAcknowledge
6400%· one data point
3 · PROOF MOMENT0:12–0:32
A 5-second log-space dive (interpolating log(scale) so speed feels constant, sine.inOut) passes through nested frames — dashboards › checkout › conversion — each level cross-fading in 0.2 s; the dive stops dead on the coral 09:42 point on the snare, the card rises, and the cursor takes Roll back on the next downbeat.
Acme Pulse
Every metric. Every minute.
Start free38 dashboards · 1 watcher
100%· all clear
4 · CTA0:36–0:45
The camera pulls all the way back out (expo.out, 1.6 s) and a teal ripple radiates from the fixed cluster, turning every dot teal with a distance-based stagger; the plate settles and Start free lands on the final downbeat as the zoom pill reads “100% · all clear”.
COMPONENT KIT
Dot field — every metric/event as a dot; 38 clusters = 38 dashboards.
Cluster labels — checkout, signups, latency… readable only at mid altitude.
Nested frames — receding outlines showing how deep we are.
Data-point macro — the single coral 09:42 point with its value.
Minimap with viewport rectangle — orientation; shrinks to a coral speck at max zoom.
Log-space zoom — interpolate log(scale) so perceived speed stays constant; 4–6 s per dive (2–3 bars at 120 BPM), sine.inOut, landing on a downbeat.
Level-of-detail cross-fade — dots → cluster labels → chart → number swap at scale thresholds over 0.2 s, each threshold crossed on a beat.
Pull-back reveal — expo.out zoom-out over 1.6 s with a teal ripple radiating across the dots, stagger by distance 0.004 s/px.
WHY / RISK · ADAPT IT
Why: it proves the agent's coverage physically — it watches everything and still finds one data point — and one continuous move is spectacular and highly readable at 1:1. Risk: text and SVG blur at extreme scale; build it as nested layers you swap at thresholds instead of scaling a single DOM tree, and budget real build time.
Adapt to your product: map your own hierarchy onto the dive (workspace › project › item › field) and let the deepest level be the one value your product catches.
Acme Pulse