Video is where the bytes are.
Video is the majority of bytes in motion. Every increase in resolution, frame rate or catalogue size is paid for in bandwidth and storage, and the codecs that carry it are frozen standards shipping on a decade-long cycle.
A codec generation takes about a decade, and ships frozen.
AV1 is frozen. AV2 will be frozen the day it ships. Between generations the only levers a video business has are to send fewer titles, send them smaller, or pay for more pipe. That is a long time to wait for the next few percent, and the next few percent is what the whole industry is fighting over.
Quality costs bandwidth, linearly
Every step up in resolution or frame rate lands directly on delivery cost, and the audience never asks for less.
The catalogue never shrinks
Back catalogue, mezzanine masters and contribution feeds sit on disk permanently and are re-encoded every time a new format arrives.
Standards move slower than demand
A frozen codec cannot learn anything about your library. It picked its compromises once, for everyone at once.
We start where the decoder is ours to place.
Last-mile playback is governed by hardware decode on devices you do not own, and no startup has ever shipped a proprietary codec into that path. So we do not start there. We start in the parts of a video business where adding a decoder is a deployment decision you make on Tuesday.
- 01deployable nowOrigin storage
Masters, mezzanine and back catalogue at rest. Written once, read on your schedule.
- 02deployable nowContribution
Feeds between facilities and into the encoding tier, on links you pay for by the gigabit.
- 03deployable nowShield and mid-tier
Cache fill between origin and edge, inside infrastructure you operate.
- 04standards-boundLast-mile playback
Device hardware decode. Governed by adopted standards, not by us.
Highlighted stages are inside your own infrastructure, where the decoder is software you control. Last-mile playback stays standards-bound until a format is adopted, which is a decade-scale process we are not betting the company on.
What changes
- 01More catalogue per petabyteThe same storage footprint holds more titles at the same quality, or the same titles at higher quality.
- 02Cheaper cache fillBytes moving between origin, shield and edge are bytes you are billed for on every miss.
- 03Higher quality inside the same pipeDelivery budgets are fixed long before the content is. Fitting more picture inside them is the whole game.
What we have measured, and what we have not.
Our published result is a text benchmark: enwik9, lossless, decompressor counted. We reached #1 on August 30, 2026 on the public LTCB. It proves the engine, not the video product. We have not published a video number and we will not publish one informally. When we do, it will be BD-rate against a named encoder on a named test set, under a measurement protocol registered before the run. Until that happens, treat video as our direction and the benchmark as our evidence.