Fragmented, Dumb, and Rented

Spatial persistence exists. ARKit has World Maps. ARCore has Cloud Anchors. Meta has Spatial Anchors. Niantic built persistence for Pokémon Go. You can save AR content and load it back later. The technology works.

What doesn't work is everything around it. Your spatial data lives in Apple's cloud or Google's or Meta's—siloed, incompatible, controlled by platforms you don't own. It stores bytes but understands nothing about what those bytes represent. It can't tell you what changed, when, or why. There's no history to query, no versions to compare, no way to ask "show me everything within ten meters" or "what did this room look like last month."

Spatial computing has storage. What it doesn't have is memory—the kind that remembers, understands, and belongs to you.

We've Solved This Before

In the mid-2000s, code had a similar problem. You could save files. But collaborating on them was chaos—tracking changes meant appending "_v2_final_FINAL" to filenames, sharing meant emailing zip folders, and if two people edited the same file, someone's work got silently overwritten.

Then Linus Torvalds wrote Git, and suddenly code had history. You could track every change, branch off to experiment, merge contributions from dozens of people, roll back mistakes. Code became permanent in a way that saving files never achieved.

But Git's real impact came from what it enabled next. GitHub turned code into a social object. Open source exploded because collaboration finally had infrastructure. Package ecosystems emerged. Fork-and-pull became a pattern that reshaped how software gets made. The primitive was version control. The consequence was an entire culture of building in public, together.

Spatial computing is stuck in the pre-Git era. Everyone building for AR is solving the same persistence problems—how to save, load, sync, query—and they're all solving them badly, independently, locked to platforms that don't interoperate. There's no shared infrastructure, so there's no ecosystem that could emerge from it.

The Walled Gardens

Build on Apple's persistence and you're locked to Apple devices, Apple's cloud, Apple's formats. Build on Meta's and you're locked to Meta. Want your app to work across both? Rebuild everything twice. Want users to keep their spatial history when they switch platforms? They can't.

This isn't a technical limitation—it's an incentive structure. Platform vendors have no reason to make spatial data portable. Fragmentation serves them. Your lock-in is their moat.

The open web won because no single company owned HTTP. Email works everywhere because SMTP isn't Gmail's proprietary protocol. Spatial computing is heading toward a future where your AR history fragments across walled gardens, controlled by platforms, inaccessible to anything outside their ecosystem. Unless someone builds infrastructure that doesn't belong to any of them.

What Memory Makes Possible

When spatial data becomes portable, versioned, and queryable, the obvious applications come first. Architects leave annotations on job sites that persist for months and work regardless of what devices contractors carry. Warehouse robots share maps that survive reboots, shift changes, and fleet updates without vendor lock-in. Your AR glasses accumulate a personal spatial layer across years—and you own it, can export it, can switch platforms without losing everything.

These are useful, but they're just "code that saves properly"—the version control layer. The interesting part is what emerges when spatial memory becomes ambient infrastructure that anyone can build on.

What Emerges

Cities accumulate layers of AR art, commentary, and memory that exist independent of any single app or platform. A new medium develops in the gap between physical and digital—persistent but intangible, public but invisible until you look. You walk into a coffee shop and find a message someone left three years ago on their worst day, telling whoever discovers it that things got better. A park bench becomes a memorial where a family has left tributes over a decade, visible to anyone who knows to look.

Games unfold over weeks instead of sessions, with territory that persists and history that accumulates. Content decays unless maintained, creating archaeology of abandoned digital spaces—the Geocities of spatial computing. Letters to your future self unlock in five years, anchored to places you might not even live anymore. Strangers leave things for other strangers. Places develop memory independent of any single person's intention.

This is the GitHub-level transformation—not persistence as a feature, but persistence as infrastructure for forms of culture, connection, and meaning that couldn't exist before. The primitive enables the ecosystem. The technology disappears, and what's left is people doing things they previously couldn't.

Beyond Storage

Interoperability and versioning are the foundation, but they're not the end state. Storage that doesn't understand what it contains is still just bytes—useful, but limited.

Imagine spatial memory that knows what it's looking at. You place an anchor and it recognizes the object: that's a chair, it's blue, it's near the window. You query your environment in natural language: "find everything I placed last month," "show me all the notes within ten meters," "how has this room changed since January?" You version space like you version code—branch to experiment with layouts, merge contributions from collaborators, roll back mistakes.

Eventually, spatial memory that reasons. "Where did I leave my keys?" "Based on your patterns, probably the entry table." "How has my desk setup changed this year?" "Who left this annotation and when?" "What usually happens in this space on Tuesday afternoons?" Not just storage, not just memory, but understanding.

That's where we're going. Persistence is the starting point. Intelligence is the destination.

We're building the open layer—spatial infrastructure that works across platforms, that developers own instead of renting from Apple or Google or Meta, that treats spatial memory as a primitive worth building ecosystems on.

The platforms will keep building their walled gardens. We're building the commons.