ZipF / Blog / The Physical Event Graph
Product
The Physical Event Graph
The Physical Event Graph is ZipF's core dataset. Every scene the network writes is a structured event. The graph is what you query. The products are how different buyers consume it.
An event is observed physical activity at a place and a second. ZipF records it with a stable schema: class, track, box, mask, lat/lon, heading, speed, inertial state, weather, rarity, and change. Identity is stripped before a pack ships.
Events become signals when they are normalized across corridors, weeks, and weather. Signals become products when a lab, a map desk, an insurer, or a city can ask for them. The path is Event to Signal to Decision. Different customers pay at different layers.
What sits on the graph
- Licensed scene packs and corridor subscriptions
- The Long-Tail Index, VRU Pulse, Surface Distress, Change Rate, Weather Stress
- World-model eval suites and VLM grounding scores
- The Scene Query API: /events, /query, /aggregates, /change, /eval, /alpha
Point-in-time fields travel with production feeds: event_time, ingest_time, published_time, correction_time. Latency is a SKU. Distribution can be open, limited, or exclusive.
How the graph grows
Density first: corridors where the tail is thick. Time second: the same kilometer next week. Layers third: train, refresh, risk, city, sim. Every channel writes into the same normalizer. The join across them is a product no single survey pass has from its own data.
The graph is not only objects. Scene events - mixed traffic, pedestrian in carriageway, animal in scene, impact candidates - sit on top of tracks. Uncaptured tail - the joins a public dump never sees - is a first-class product. The same ontology is what a planner queries when it asks what is true on this meter, right now.
Questions
What is the Physical Event Graph?
ZipF's live graph of the physical world. Every scene is a structured event; the graph is the queryable join of those events across corridors and weeks.
What fields are in a ZipF physical event?
Class, track, box, mask, lat/lon, heading, speed, inertial state, weather, rarity, and change, with point-in-time stamps and a privacy cut.
Who buys graph access?
World-model labs, ADAS and OEM teams, HD-map desks, cities, parametric insurance, every embodied stack, spatial computing, and simulation. Price and restriction rise from packs to exclusive seats.