What this is
A sequence of substrate frames along a moving vantage is video of the twin, and unlike
generated footage every frame is checkable: this browser recomputes each frame's SHA-256 over the raw
bytes and compares it to the digest the cell declared before a single pixel is drawn. Pin the
recording to a sealed sky and the same path replays byte-for-byte on any cell — which is what turns a
pile of frames into a film you can seal.
Playback fps is a display choice, not a capture rate. Frames are
recorded ahead — a pinned frame measured about 2.0–2.2 s over the network — so nothing here is
streamed as it renders and no claim of real-time capture is made.
Record
path
Drift — 40° of longitude, across the sealed traffic
Zoom in — span × 7/8 per frame
Orbit — 360° of longitude (mostly empty at this seal)
Latitude sweep — −60° to +60° (5 of 8 near-empty)
frames
8 16 24 32
size
the stage's size (fills it)
128 192 256 384
stance
EXTERIOR INTERIOR
lanes
1 2 3 4
pin to a sealed sky
● Record
■ Stop
Re-verify from memory
Measured 2026-08-09 from a real browser: 8 pinned frames at 192²
took 10.4 s across 3 lanes — 1.30 s per frame. From the shell, a single pinned frame took
2.0–2.2 s, so lanes are overlapping network waits; they do not make a cell faster.
This frame
vantage —
marks drawn —
The pin
at_seq —
sealed at —
age —
aircraft —
stream —
Fetch newest sealed sky
Pinned , the aircraft come from one sealed snapshot named by
stream sequence, so every cell composes the identical frame and the recording is reproducible.
Unpinned , the frame is composed from the serving cell's own live feeds: the
recording follows a moving head, two cells disagree, and replaying it later will not give these
bytes back. That is a different kind of recording and this page will not label it reproducible.
This sky is concentrated, and it decides which path is worth recording.
Measured at seq 129254, span 40°: US east 9,205 marks in view · US central 17 · London 1 ·
Tokyo 1 · Gulf of Guinea 0, out of 9,219 sealed aircraft. A 360° orbit against this seal
therefore records one populated frame and seven empty ones — which is why the drift is the
default and the orbit says so in its own label rather than quietly disappointing you.
An empty frame is not a broken frame. A verified raster with no
aircraft in it is a true statement about a sky with nothing sealed in that view. It is drawn,
counted and kept. Only a frame that fails its digest or its geometry is refused, and that is a
different colour on the strip and a different word on the screen.
Chain
no sequence recorded
chain₀ = 32 zero bytes
chainᵢ = SHA-256(chainᵢ₋₁ ‖ digestᵢ)
where digestᵢ is the 32 raw bytes of frame i 's verified digest, taken in
recording order. Two machines that record the same path against the same pinned sequence must
arrive at the same chain — that agreement is what makes the sequence a sealable film rather
than a bag of frames. A refused frame stops the chain where it stopped; the reel is then not
sealable and says so.
Measured 2026-08-09, not asserted. The 8-frame drift at 192²
pinned to seq 129254 was recorded twice by different means: in this browser, and in a shell that
fetched from nbg-01 with curl and folded the chain with
shasum. Both reached
c929396e2ca00a18773f0acc69fe61e3bb1ba8c10cec018ee39051b2afd1c77b, and all eight
frame digests agreed one for one. Different code, different cell, same film.
Export
Download manifest (JSON)
include raw frames as base64
—
The manifest carries every request URL verbatim, every pose, every frame digest
and the chain. That is enough for another machine to re-issue the same requests and check whether
it reaches the same chain. Frames themselves are optional because the digests are the claim; the
pixels are only the evidence.
Honest limits
Not a video codec. Frames are independent full RGBA rasters with no inter-frame
prediction. Nothing here compresses; a 24-frame reel at 256² is 6.3 MB in memory.
Not real-time. About 2 s per pinned frame over the network, so a reel is built
and then played, never captured live.
Pinned frames show the sealed sky and airport anchors only. Measured at the
same pose: 9,205 marks pinned against 403 unpinned, because the pinned branch deliberately
excludes live satellites and weather rather than reintroduce per-cell sampling.
Heading does nothing. azimuth_milli is parsed by the route and does
not change the frame, so this page ships no heading control.
The seal head can be stale. The age is shown above and is not judged for you —
whether a two-day-old sky is fresh enough depends on what you are looking for.
A recording at the stage's size depends on the screen. It asks for the frame
that fills this stage, so a phone and an ultrawide record the same path at different sizes and
reach different chains — each reproducible from its own manifest. For a film two screens should
agree on, choose one of the fixed squares.
Elsewhere