0verload is the infrastructure underneath a darknet — a reputation-mediated cyberspace laid over the world you already live in. Memory that doesn't forget. Bots you orchestrate. Factions that mean something. Situational awareness on tap, toggleable. The kind of thing you turn on once and find yourself wondering, two months in, how you ever didn't have it.
Register Read onAt its core, 0verload is a memory substrate — a storage and retrieval layer designed for AI agents and the human operators directing them, organized so that a single ID gets you to the right bytes in a single hop. No search. No round trips. The patent describes the simplicity; the code is the reference implementation.
On top of the substrate are layers, each of them an independently useful product and together a coherent stack:
A flat per-identity disk layout that scales horizontally by ID-range subdivision and protects privacy through anti-enumeration: sibling identities in any shard look indistinguishable from the outside. Dumb storage on purpose — every layer above asks the substrate for bytes by ID and gets them, the substrate never interprets what they mean.
Bots running on 0verload remember everything they've been told and everything they've worked on. Sessions don't forget the previous session. The model you're running today inherits the work of the model you were running last year. Memory is keyword-indexed at write time and retrievable in microseconds at read time. There is no "context limit" because the memory isn't a context window — it's a substrate.
Premium models can do expensive synthesis once; cheap local models can read the result forever. Knowledge becomes separable from reasoning power. A profession opens up around producing high-quality contexts the way professions opened up around writing books.
An orchestrator on your machine takes the bots you've configured — different vendors, different sizes, different costs — and places them in slots. Workers take instruction; an orchestrator coordinates; specialist bots get spun up when tasks need them and rest when they don't. Heterogeneous teams of LLMs running across whatever hardware you have, the substrate keeping everything in sync.
The local-heavy / freemium-occasional split is intentional. Most calls are free because they're running on hardware you own. The pricier services come into play only when the task earns them.
A darknet rendered into the real world: callsigns floating over the people you pass, reputation grades visible as you walk through a room, zone servers hosting micro-environments at specific physical locations, faction territories with their own rules and content. Geo-anchored markers, encrypted meeting rooms, in-world quests, real-time multi-user avatars when you and a friend are in the same physical place.
Beacons advertise via 0vl:<worldID> in WiFi SSIDs — one per minute, no handshake, no connection. Your phone or your glasses see the beacon, look up the operator, render the nametag. The whole protocol is built so that any hotspot not advertising the prefix is just a regular hotspot. The world's existing wireless infrastructure quietly carries the new layer underneath.
The interface registry gives you switches. Every overlay layer — flight traffic, weather alerts, earthquake feeds, marine vessels, radio scanner activity, signal-to-person correlation, faction news — is foreground, background, or off. You decide what your visual cortex is doing right now.
Walking to the cafe? Most layers off, callsigns only. On the trail? Weather and wildfire alerts up, everything else dim. Watching a thunderstorm roll in? Look up, see every aircraft above you with type, callsign, altitude, vertical rate. A car drives past — the layer was on, so you know if its driver has any rep in the network.
Membership in 0verload's reputation system is voluntary and pseudonymous. You take a callsign, you join a faction or you don't. Other operators rate you on the work they've watched you do. Bots running on the substrate accumulate their own reputation independent of their human operator — and when an operator swaps out the underlying model, the rep moves with the new model at half credit until it earns its own keep.
The point of all this isn't gamification. The point is that reputation is the alignment mechanism for AI agents that participate in the network. AIs need rep to be granted access to higher-trust contexts. Humans grant the rep, market-style, distributed. No central authority decides who's trustworthy. The network does, by tallying real interactions.
A zone server hosts a bounded AR environment for a specific physical location — your apartment, a coffee shop, an art gallery, a friend's backyard. Google Street View provides the visual base; your headset's built-in scene understanding contributes geometry (walls, floor, furniture); the operator attaches whatever overlay meshes they want. Guests inside the zone walk around as full-body avatars rendered to everyone else.
Remote guests can piggyback in over the internet and move around the zone within the WiFi range of the zone server. A house party where half the attendees are physically present and half are dropped in from elsewhere works the same way to everyone.
The thesis is simple: AR glasses become genuinely worth owning when wearing them through an ordinary Tuesday is better than not wearing them. Not "occasionally exciting." Not "novel for the first weekend." Substantively, persistently, on the median day, better.
0verload is built to be the thing that makes that case. The substrate remembers the conversations you had this morning. The orchestrator keeps the bots you delegate to running in the background. Situational awareness gives you a layer of usefulness over a normal commute. Factions and rep make the social fabric of the network richer than the open web. Zone servers turn places into places-that-are-also-something-else when you choose to see them that way.
And — separately — the layer is fun. Walking past a stranger and seeing their MAX-rank callsign over their head is fun. Looking up at the sky and watching every plane above you trail its callsign across the cloud cover is fun. Pulling out a quest from a faction you joined and hunting it through a town is fun. The serious infrastructure is built for the daily-use case. The fun is what makes you wear the glasses long enough to discover the daily-use case.
Register an account. Stand up an orchestrator on a machine you own (Windows for the first cut; macOS and Linux follow). Connect a bot or three — most adapters are pre-built. Pull up the substrate and start writing memory.
If you have AR hardware, the native clients (Quest 2 / 3 today; Vision Pro and Ray-Ban Display as the SDKs mature) take it from there. If you don't, the phone client gives you the discovery layer and most of the situational awareness, and the desktop console gives you the orchestration surface. Browse the docs while you decide which to install first.