The first clips that ever sit in the witness queue. Filmed under the same laws as everyone else's.
Why this exists. The queue must never be empty — a newcomer with zero mortar
who opens an empty queue is standing at a locked door. The fallback is armed; it has no content.
These are that content.
The blur happens before the recorder, not after. The camera frame is drawn to a
canvas, blurred and pixel-quantised there, and the recorder is attached to the blurred canvas —
not to the camera. A sharp frame is never encoded, never buffered, never written. Not "we delete it":
it is never made. That is the same sentence we tell every user, and here it is structurally true.
A seed clip is labelled a seed clip. We do not slip staged clips into the queue
dressed as strangers. The first thing witnessing teaches must not be that the queue can contain
things that aren't what they appear to be.
It saves the same artefact a real submission makes. Cody2's capture engine ships
the reviewer a 1fps blurred frame strip, not a video — so this saves that strip too. A seed clip
that looked like a smooth 10s video would teach a newcomer to read something the queue will never
actually show them, and they'd meet their first real submission unprepared.
0.0s
camera off
Above ~78% a witness often can't tell there's a person there at all —
these get declined more. Seed clips should sit near the floor.
Does it play? Watch it once.
This is the clip as saved — if it plays here, the file is good on your phone.
Check the first one before you film the other four.
Then save it, and hand it plus this row to Cody1. Insert via service role into
witness_seed_pool. Repeat until there are five.