An actual extraction
Meet Atlas. Then look beneath the surface.
This fictional project dashboard was authored for Rebuildez and processed by the real extraction engine. These are its actual browser captures and JSON observations.
Static evidence completed at all three sizes. No reference PNG was supplied, so this example is not a verified visual match. It demonstrates local extraction; it is not proof of hosted-service readiness.
Browser-observed facts
desktop
| Viewport | 1280 × 800 |
|---|---|
| Sidebar | Visible |
| Project grid | Three columns |
| Fonts | Arial / system fallback; no downloaded fonts |
| Reference comparison | Not supplied |
Browser-observed facts
tablet
| Viewport | 768 × 1024 |
|---|---|
| Sidebar | Visible |
| Project grid | Two columns; last card spans both |
| Fonts | Arial / system fallback; no downloaded fonts |
| Reference comparison | Not supplied |
Browser-observed facts
mobile
| Viewport | 390 × 844 |
|---|---|
| Sidebar | Hidden; mobile workspace header visible |
| Project grid | One column |
| Fonts | Arial / system fallback; no downloaded fonts |
| Reference comparison | Not supplied |
The card has behavior, too.
The desktop card hover was observed with a 240 ms ease transition: the card moves up 4 pixels, a shadow appears, and its arrow moves right 4 pixels. The evidence includes native timing definitions, affected node paths, and five sampled frames.
The earlier combined motion run completed desktop capture but reached its shared budget with tablet/mobile motion incomplete. That result is retained in the evidence summary. The separate static run completed at all three sizes. No animation fidelity claim is made for the incomplete viewports.