Three engines for the places data has always been thin — the floor, the answer engine, and the product market.
Moonlight is Ali and Nik. The stack does one thing on purpose — and it's almost too simple: run the models on the data nobody bothers to collect (vision on the floor, probes across the AI answer engines, signal scoring on the market) and turn those raw feeds into a real-time intelligence layer you can query before anyone else can.
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. If we can't verify a number independently, it doesn't ship.
Physical space is the biggest blind spot. Cameras are everywhere; useful data from them is almost nowhere. Janus turns the cameras you already have into a live model of the floor.
Search is fragmenting fast. Google still matters — 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 products.
Better together. The three engines share one job: close the gap between what's happening and what you actually know about it.
Each engine is live, independently deployable, and solving a distinct intelligence problem.
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 health, competitor moves, 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.
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 ten. Each engine reaches production quality before the next one starts.
Intelligence should sit below the app layer — a service you wire into decisions, not a dashboard you remember to check.
Janus reads zones, flow, and dwell. It doesn't store faces, build profiles, or identify individuals. The moat is operational intelligence, not biometrics.
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.
