Film runs 0–12 s. The hold runs to 40 s and includes a night gap and a heat-free next-night re-exposure. Any change is written into the timeline at the playhead, so scrubbing back replays it exactly.
Sets IR irradiance at the skin. Only the IR, skin and report lanes should respond.
Dims the sky or screen. Skin temperature should not move beyond 0.0001 °C.
Changes appearance. The heater physics are identical either way.
Heat arrives this many seconds after the command.
Overrides the operator. The image stays on.
Six frames from the canonical film. Tap a frame to jump there.
Every claim the scene animates, with its status. Unknown rights are not treated as granted.
In-page invariant tests
Run in your browser on load against the canonical film. These are machine checks, separate from the critic.
Open questions
Real forearm skin under a radiant panel responds over tens of seconds, not 2.5 s. What compression is acceptable before the film misleads?
Does a viewer who reads no labels conclude that moonlight warmed the person? This needs a five-viewer test with no captions.
Which safe irradiance limit and exposure time should bound the physical rig? Not sourced here.
The two user-supplied anchors (the 1977 Georgia Tech mirror array, Ichihashi's thermal displays) need citations before any scene depicts them.
Next smallest proof
RECURRENCE with a real participant: run the next-night re-exposure (image, no heat) with five people who did and five who did not receive heat on night one. Expected observation if the hypothesis holds: higher warmth reports to the image alone in the heated group. Falsifier: no difference between groups.