Intelligence, across
every surface
that matters
We build intelligence engines for the places data has always been thin — physical space, search visibility, and the product market. Three engines. One layer.
The thesis
One intelligence layer
across the full picture
Most analytics tools see one slice of a business: the click, the ranking, the dashboard. They miss what happens before the click, in the room the camera watches, in the answer engine that never links back. Moonlight was built to close that gap.
We are a small studio — deliberate by design. Every engine is built to depth, not breadth: real data, real proof, and a thesis that holds up in a pitch room and in production.
Physical space is the largest blind spot. Cameras are everywhere; useful data from them is almost nowhere. Janus turns existing infrastructure into a live operating model of the floor.
Search is fragmenting faster than most teams realise. Google rankings matter. So do the AI answer engines now fielding millions of queries. Caposeo tracks both in one place.
The product market moves invisibly. Breakouts and quiet deaths happen weeks before anyone notices. Launch Sentinel surfaces the signal early, across 79,573 tracked products.
Together, not apart. The three engines share a single intelligence thesis: close the information gap between the world and the decisions being made about it.
The three engines
What we've built
Each engine is live, independently deployable, and solving a distinct intelligence problem.
Janus
Physical-space intelligence. Turns ordinary cameras into a live operating model of the floor — zones, dwell, queues, and role recognition — without surveillance creep.
Search and AI-search intelligence. Tracks Google rankings, technical site health, competitor movement, and AI citations across ChatGPT, Perplexity, Claude, and Gemini — one workspace.
Product-market intelligence. Maps 79,573 tech products with signal scoring, liveness probes, press velocity, and product-death modeling — so you know before everyone else.
How we work
Built to a standard,
not a schedule
A small studio with a narrow mandate: build things that hold up in production, with data that holds up under scrutiny.
Every metric we surface — in a deck, on a page, or in a dashboard — is measured, not modeled. If we can't verify it independently, it doesn't ship.
Three engines, not thirty. We'd rather own a problem completely than skim the surface of ten. Each engine is built to production quality before the next one starts.
Intelligence should sit below the application layer — a service you wire into decisions, not a dashboard you remember to check. That's the architecture we aim for.
Janus reads zones, flow, and dwell. It does not store faces, build profiles, or identify individuals. The moat is operational intelligence, not a biometrics play.
Work with us
Ready to close the
information gap?
If you're building something that needs deeper intelligence on physical space, search visibility, or the product market — let's talk.
No deck required. A one-paragraph description is enough.