The world needs offline access. That's why we built Apocalypse Mode.
The internet feels permanent until it isn't. A disaster cuts power and fiber. A war isolates a region. A government blocks access. A storm floods a data center. A cable breaks, a provider fails, or an ordinary family simply loses the connection it depended on. The unexpected does not schedule an appointment—and useful knowledge should not disappear with the signal.
The network is powerful, not guaranteed
Modern software quietly assumes that the network will always answer. Our documents live in clouds. Our search boxes are remote. Our AI assistants send every question to a data center. Even simple reference material is often wrapped in an application that becomes an empty shell when its servers cannot be reached.
Most days, that design is convenient. On the worst days, it is a single point of failure.
Disasters and wars are the sharpest examples, but they are not the only ones. Connectivity can disappear during travel, in rural areas, after infrastructure damage, under censorship, or because a household cannot afford a reliable connection. Sometimes the failure lasts minutes. Sometimes it lasts much longer. Whatever the cause, access to basic knowledge and reasoning tools should not depend entirely on a live account, a healthy route, and someone else's computer.
That is why we created Apocalypse Mode.
A small, self-contained core
Apocalypse Mode prepares a working local core inside WebBrain before you need it:
- A 2.6-billion-parameter text model for local conversation and reasoning.
- A compact vision model that can understand screenshots on the device.
- A searchable Wikipedia archive that opens full articles without the web.
- An Emergency Box for OpenStax textbooks, emergency-health references, field manuals, and other practical PDFs.
On supported Chromium systems, the two models run through WebGPU. The browser uses the GPU already in the computer; there is no separate inference server to install and no remote model endpoint required for the local run. Screenshots can stay on the device. Wikipedia and downloaded documents are read from browser-managed storage.
Once those resources are downloaded, the core is self-contained in the browser profile. Local inference, archive search, article reading, and installed PDFs do not need an internet connection.
The preparation step does. Models, archives, and books cannot arrive after the route to them is gone. That is the point of the mode: move the useful bytes onto hardware you control while the network is still ordinary.
Why WebGPU changes what a browser can keep
Offline knowledge alone is valuable, but a pile of files still asks the person under pressure to know exactly where to look and how to interpret what they find.
Local inference adds another layer. The text model can answer from the information already available to it and retrieve compact, attributed passages from the installed Wikipedia archive. The vision model can inspect a screenshot locally. Together they provide a modest but useful reasoning surface without turning every question into a network request.
We deliberately chose models small enough for consumer hardware. They are not frontier data-center systems, and we do not present them as such. Their job is resilience: be present, private, and usable when the alternative is no model at all.
WebGPU is what lets us ship that capability through the browser. It keeps the setup approachable—open WebBrain, enable the mode, let the downloads finish—while using on-device acceleration for inference. No cloud token. No API balance. No round trip.
Knowledge should be readable, not merely stored
“Downloaded” is not the same as “ready.” A backup nobody has opened is a hope, not a plan.
Apocalypse Mode includes an offline Wikipedia reader because an archive is only useful if a person can search it and open an article. The basic setup starts with a smaller Simple English, text-only edition. People who have more space can add another language, a fuller edition, or an archive with images. WebBrain checks the current archive metadata and verifies a replacement before removing the older working copy.
The Emergency Box applies the same principle to documents. It organizes a curated shelf of education, health, and field references; downloads PDFs into local storage; and gives them an offline reader. Downloads can be paused and resumed. Resources can be removed when storage matters more.
No collection can anticipate every emergency, and a document is not a substitute for a doctor, trained responder, or local authority. Medical guidance also ages. The Emergency Box says this plainly because resilience requires honest boundaries, not fantasy.
Offline does not mean invulnerable
Apocalypse Mode removes the live network from the core path after setup. It cannot remove every dependency from the physical world.
The device still needs power. The browser profile and its storage need to survive. Clearing extension data can erase the kit. Hardware can fail. Live news, current warnings, hosted providers, and websites you never downloaded still need connectivity. Firefox provides the offline library and Emergency Box, but it does not expose Chromium's WebGPU model stack.
These are reasons to practice, not reasons to give up. Open the local chat with the network disconnected. Search Wikipedia. Read an installed PDF. Check the kit again after browser migrations and storage cleanup. Keep external copies of source files where licensing and circumstances allow. Treat WebBrain as one layer in a larger preparedness plan that includes power, communication, supplies, and human expertise.
Download it while you can
Software normally asks you to download it because a launch is exciting. Apocalypse Mode asks for a more serious reason: availability tomorrow depends on preparation today.
Install WebBrain. Open Apocalypse Mode. Let the text model, vision model, and Wikipedia archive reach Ready. Choose the books and field references that make sense for your language, region, family, and work. Then test them offline.
The world needs access that survives disasters, wars, infrastructure failures, censorship, and the unexpected. It needs tools that can remain useful when a server cannot be reached. This is our first practical answer: local inference through WebGPU, local knowledge in open formats, and a browser experience designed to keep both close.
Read the complete Apocalypse Mode setup guide, then download WebBrain while you can.
Keep WebBrain in your browser. Keep it ready.
Tags: #WebBrain #ApocalypseMode #OfflineAI #WebGPU #OfflineAccess #LocalAI #Wikipedia #Preparedness