What's actually being worked on
Open questions, active work, and the stuff that turned out to be a dead end. Updated when things move, not on a schedule.
01
How something makes the list
A question earns a spot here if three things line up: someone running their own setup would actually hit it, it can be measured with gear one person can realistically own, and there's a specific way the answer could turn out to be no. That last one is the part we're looking for.
Every piece of work states upfront what would prove it wrong, before any data gets collected. A boring answer still gets published, since it still saves the next person some time. Boring work still counts as work.
02
Active
| Track | Question | Status |
|---|---|---|
| MuxNode | How many ways can a common measurement tool produce a confident wrong number, and what catches each one? | Building |
| Cuebus | Do timing cues in caption tracks survive a normal encode and package chain? | Spec |
| Genlock | Where does each contribution protocol stop delivering a usable feed, and what does the recovery cost in latency? | Queued |
03
Dropped on purpose
Ad-insertion signalling was genuinely interesting engineering, but the market it serves isn't the one this lab is built for, so the research moved toward other kinds of timed metadata instead.
More will land here as things get shelved. Publishing what didn't work is half the point.