What This Is //

The Spatial Systems Atlas is YesterdayWorldz's living map of the labs, models, and robots reshaping how machines understand and move through space — six clusters, one interactive scene, a glossary to learn the vocabulary in, and a Journal that makes sense of what's actually moving. Built to be explored, not just read.

Every node is a real company, model, or technique — not a placeholder. Click into one and you get what it builds, what it runs on, who buys it, how it's funded where that's public, and an editorial Reading: a specific, falsifiable prediction about where it's headed, not a restatement of the facts above it.

The Problem

Spatial AI is moving too fast to track by hand. World models, robotics, Gaussian splatting, simulation, mixed reality — six fields, dozens of companies, real developments every week. It's a real enough shift that major industry-intelligence firms have started tracking "physical AI" as its own category this year. Outside of that, the only way to keep up is scattered blog posts, X threads, and newsletters that go stale the day they're sent.

Why Being AI-Native Actually Matters Here

Plenty of products say "AI-powered" and mean a chatbot bolted onto a static page. This one means something checkable: an agent monitors real sources on a schedule, drafts new entries, and keeps the map from going stale the week after launch — the way almost every manually-maintained reference eventually does. A builder or developer gets a reference they can actually rely on without wondering when it was last touched. A student gets current, correctly-used vocabulary in the Glossary instead of scattered tutorials of uncertain age. An enthusiast catches developments a personal reading list would miss simply by not reading something that day.

Built to Prove AI Can Be Practical, Not Just Impressive

Nothing here publishes on its own. Every agent-drafted node, every Journal piece, every glossary entry gets checked by a person before it goes live — slower than full automation, on purpose. That review gate is a systems decision as much as a safety one: a deliberate control point governing what's allowed downstream, not a feature bolted on after the fact. The two agents behind this platform don't talk to each other directly either — they coordinate through shared data, kept loosely coupled rather than tightly wired together. Even the six-category taxonomy is a systems choice: how you classify a fast-moving, messy field into something navigable without flattening what makes it worth tracking in the first place.

If you're building with AI agents yourself, or just thinking in systems generally, this platform is a working example as much as it's a map: filtering before the expensive step, structured extraction instead of freeform guessing, loose coupling over tight coordination, a genuine review gate, and an eval suite built to catch failure, not just demonstrate success. Functional and practical was always the actual goal here — not AI as decoration.

Who It's For

• Builders and developers who want a spatial-AI reference that updates itself, not one they have to maintain by hand
• Students and newcomers getting oriented in the field, who need a map and a glossary before they need a firehose
• Spatial computing enthusiasts tracking a space moving faster than any single newsletter can
• Founders, researchers, and analysts sizing up who's actually building what, not just who's loudest