Spatial AI, in plain words.

Spatial AI is artificial intelligence that reasons about real places: where things are, how they change over time, and what is true on the ground right now.

A general language model can describe the idea of a place. A spatial AI system can answer about a specific one, like this field or this address, because it's connected to observations of that place rather than to text written about it. Vortx is a spatial AI lab: we build the layer that holds those observations and hands them to an agent in a form it can use and cite.

Earth memory is a record of what has been observed at each place on Earth, stored so an AI agent can recall it later and prove where it came from.

Every spot on Earth gets a stable address about ten metres across. Each observation at that address — a vegetation reading, a water level, a change since last month — is stored as a small fact with its time, its source, and a signature. An agent asks by place and gets the fact back, along with the receipt. The open version is emem.dev; the private version is geo.qa.

Grounding is giving an AI agent a checkable source for what it says about the world, instead of letting it answer from memory or guesswork.

An ungrounded agent will state that an address is low flood risk because the sentence sounds right. A grounded one points to a signed observation of that place and a date. The difference matters most where being wrong is expensive: insurance, lending, logistics, defense. That is also where a stale dataset served as live truth does the most harm.

Each fact carries an ed25519 signature and a blake3 content hash. Anyone can recompute the hash and check the signature offline, without trusting the server that served it.

The hash is taken over the fact itself, so the address you ask for and the answer you get are bound together. Change either, and the hash stops matching. The signature ties the fact to the responder that produced it. Put together, an agent that quotes a fact is also handing you a receipt you can re-pull and check yourself. You can try it on the home page.

Imagery and maps show a place to a person. Earth memory gives an agent a signed answer about that place that it can cite.

You don't query a satellite photo; you look at it. A map tells you what is where, not what changed last week or whether a reading can be trusted. Earth memory sits a layer underneath: it turns observations from many sources — satellites, drones, vehicle cameras, fixed sensors — into facts an agent can ask for by place and time, and verify. That's the part that was missing.