Build the future of AI video

What we work on

Generated video is easy to demo and hard to make useful. A single clip is a solved problem; twenty clips that look like one video is not. Most of our work sits in that gap — keeping a character's face the same across twenty scenes, stopping a room rearranging itself between shots, making a cut land where an editor would put it, and making all of it cheap enough that a person can afford to change their mind.

Concretely, that means prompt and pipeline design, model routing and cost control, rendering infrastructure, and a product surface where a person stays in charge of something that would otherwise run away with their money. If you have opinions about any of that, we would like to read them.

How we work

• Small team, so what you build ships and you own it afterwards.
• We would rather cut scope than ship something we would have to describe carefully.
• Written thinking over meetings — decisions land in the repo, not in a calendar.
• The product bills real money on every action, so correctness is not optional.

What to send

Something you made, and a paragraph on why it was harder than it looks. That tells us more than a CV does. If it is relevant to generated video, better — but a well-argued piece of work in any field reads the same way.