The problem
Picture a store manager on a Tuesday afternoon, revenue down fifteen percent from the week before, with no obvious reason. The only way to actually find out is to scrub through six hours of security footage hoping something jumps out. Nobody has six hours for that, so the number gets chalked up to a slow week.
The camera was already running. The footage existed. It simply wasn't being turned into answers.
The idea
Physical space never got its analytics layer.
Web analytics made digital environments measurable, where people clicked, how long they stayed, where they gave up and left. Physical space never got that, even though most of it already has cameras pointed at everything that happens.
Brelisk is trying to make physical environments understandable the same way, using cameras that are already installed, without asking anyone to bolt on new hardware first.
What we're building
Continuous lightweight tracking.
Detect what is unusual.
Use deeper reasoning only when something deserves attention.
Why retail first
Retail is a clean, well-defined problem: clear zones, clear questions, clear stakes. Start narrow, make it work, prove the model on real footage in a real space. Then expand to whatever physical space has a camera and a question nobody's answering.
The goal isn't more cameras. It's better understanding of the spaces we already have.
All of this is still early. Brelisk is a working prototype today, tested on recorded footage rather than live spaces, and built for retail first.