An architectural decision that looks lazy and is not: the whole thing runs on one Mac, locally, with no server involved.

What the alternative would look like

Stream video to a backend, process it there, push results to a display client. This is how a lot of modern event tech is built, and it buys real things — heavier processing, central logging, remote support.

It also means a show depends on venue internet.

Why that is disqualifying

Venue networks are the least reliable component of any event, and they fail worst at exactly the moment of peak demand, which is also the moment your game is running in front of everybody.

The failure mode is not degraded quality. It is a black screen while nine hundred people watch, and no graceful path back.

Everything else in a production rack is deliberately local for this reason. Lighting desks, media servers, playback — the whole discipline of live production is built on not depending on anything you cannot see. A crowd game that needs the internet is asking to be the exception, and it has not earned that.

What running locally costs

Processing has to fit in a real-time budget on a machine that might be a few years old. That constrains the computer vision quite hard — we are working with aggregate silhouette measures rather than anything per-person, partly on principle and partly because the frame budget does not allow more.

We also give up central logging and remote diagnostics, which makes supporting a show harder.

What it buys

A system with one dependency: the machine in front of you. No network, no account, no service that can be down, nothing to authenticate at 8:55am. Plug in a camera, plug in a screen, run it.

It also removes an entire category of privacy problem. Video never leaves the machine, because there is nowhere for it to go — which is a much easier thing to tell a venue than any amount of policy about how carefully we would have handled it.

Why macOS

Because that is what is already in the rack at the events we are building for, and because a single well-understood platform is worth more right now than partial coverage of three.

Not a permanent position. It is the right first platform, not the right only platform.