No internet, no grid, no cloud account. The Archive is a curated body of tools, apps, and knowledge packaged to be useful under exactly those conditions — which also makes it a genuinely good offline library on an ordinary day. It sizes to computers small and large, and serves to any phone over local Wi-Fi.
In the Arctic at Svalbard, humanity keeps a vault of seeds and a vault of data on century-scale film — both answers to the question "what would we regret losing?" But neither holds the thing you'd need first after a real rupture: working knowledge. How to purify water. How to treat an infection. How to build the machine that builds the machines. The Archive is our answer to that gap — not one vault at the end of the world, but many small living copies close to the people who might need them.
| Tier | What it holds |
|---|---|
| days–weeks | Evacuation and shelter information, offline maps, ways to reach people when networks are down, water treatment, first aid, food safety, local emergency references. Small, fast to copy, worth having on every phone and station. |
| months | Deeper medical and public-health references, sanitation, repair manuals, agriculture for the current season, and tools for a community to coordinate and decide. |
| years | The rebuild corpus: an operating system that runs tools and can build new ones, local AI models, the Open Source Ecology civilization starter kit, open machine designs and CAD libraries, manufacturing, metallurgy, chemistry, medicine, education — the knowledge that turns survival back into civilization. |
The tiers nest: a full Archive node carries all three as well as many user selected libraries; a phone might carry only the first. The short tier is small, contextual, and localized; the long tier is vast and has to be well organized and searchable by an agent.
A library you can't navigate is a warehouse. The Archive includes a small language model that runs entirely on the node's own hardware — no internet — and can search the corpus, summarize it, translate it, and walk someone through it. "How do I make this water safe?", answered from the library, with citations back into it, in the asker's own language. It's an index with a voice: every answer points back to the source it came from. Co-located with a comms station, it helps a small team answer far more questions during a disaster than they could alone, and can serve as an automated dispatcher over a radio.
An archive is only as good as its last update, and the places that need one most have the worst links. Good Ancestor funded a binary-diff updater for the Butter Box community-node software: instead of re-sending a whole library, a node receives only the bytes that changed — small enough to arrive over a thin, intermittent, or one-way link.
That's what connects the Archive to the rest of the system. The same mechanism that updates a node over a slow rural connection can, in principle, update it from a broadcast overhead — content, and eventually the librarian's model itself, arriving to every node under the footprint at once.
Archive nodes aren't only for disasters. They already serve schools without reliable internet, boats, off-grid homes, and anyone who wants a library no subscription can take away.
Curious what's inside? The content is served with Kiwix — you can browse a live Kiwix library to explore the kind of offline material it collects. Read more about the initiative at our Archive project.