How GhostStrats Builds an Offline Survival Computer With Project NOMAD

Thank you to GhostStrats for documenting a practical way to build a portable offline knowledge server. His Project NOMAD survival computer video combines a compact ZimaBlade, Ubuntu, an external boot drive, locally stored reference material, and optional local AI in one deliberately portable setup.

Credit also belongs to the developers and contributors behind Project NOMAD. As GhostStrats explains, he did not create the project. His contribution is showing how the community-built software can be installed on a small x86 computer and prepared before an outage, remote trip, or other situation in which a normal internet connection may not be available.

The phrase “offline internet” needs one important qualification. This system does not preserve the live web, current news, cloud accounts, or websites that were never downloaded. It creates a searchable local collection of selected knowledge, maps, courses, documents, tools, and AI models. What remains available after disconnection depends entirely on what was installed and synchronized beforehand.

GhostStrats introduces the idea of keeping essential digital resources accessible even when a conventional internet connection is unavailable.

Watch before building: The video shows GhostStrats’s physical hardware, external-drive workflow, Ubuntu preparation, Project NOMAD installation, and final setup in sequence. Watching the original demonstration helps distinguish his portable configuration from Project NOMAD’s standard installation path.

Source and safety note: This article reorganizes the process and observations shown in GhostStrats’s video and checks the software details against the official Project NOMAD repository. It is not a promise that every computer, USB enclosure, BIOS, AI model, or content collection will work identically. Back up existing data before writing an image or installing an operating system, and verify every target drive before confirming a destructive disk operation.

Why Build an Offline Knowledge Server?

Most of the information people use every day lives behind infrastructure they do not control. Search engines, cloud drives, streaming platforms, mapping services, and AI assistants normally require internet access plus functioning remote services. Even a fully charged laptop becomes much less useful when those connections disappear.

GhostStrats approaches the problem as a preparedness project rather than an attempt to duplicate the whole internet. The goal is to choose valuable resources in advance, store them locally, and make them available through a browser to the server itself or other devices on the same network.

The project begins with a familiar workstation, but its purpose is to prepare information that can remain usable away from normal online infrastructure.

A prepared Project NOMAD system can serve several different needs:

• Emergency and medical reference material selected before an outage
• Wikipedia, ebooks, repair guides, and other offline reading
• Regional street maps downloaded for local browsing
• Educational lessons and courses for places with unreliable connectivity
• Personal PDFs, notes, manuals, and technical documentation
• An optional local AI assistant that does not require a cloud API after its model is downloaded

The system is useful only if it is prepared, tested, powered, and updated before it is needed. Project NOMAD still needs an internet connection during initial installation and while downloading new applications, content packages, updates, and AI models.

The Portable Hardware Behind GhostStrats’s Build

GhostStrats uses a ZimaBlade compact x86 server as the host. Its small footprint makes it easy to place beside a portable display and external storage, while the x86 architecture supports standard Debian-based operating systems such as Ubuntu.

GhostStrats pairs the compact ZimaBlade with a small display, portable power equipment, cables, and external storage to create a transportable field setup.

-47% OFF

ZimaBlade - Single Board Server for Cyber Native

Single board computer

$69.00 - $179.00

Buy Now

ZimaBlade is a logical platform for the base knowledge server because it provides x86 compatibility, replaceable DDR3L memory, onboard eMMC, native SATA connections, and PCIe expansion. The official ZimaBlade setup documentation can help owners identify ports, storage options, memory limits, and third-party operating-system paths.

That does not make every ZimaBlade configuration a high-performance AI workstation. Project NOMAD’s core management software is comparatively lightweight, but local large language models can require far more memory, storage, CPU performance, and GPU acceleration. On modest hardware, the sensible priority is the offline library, maps, documents, and education tools. Local AI should be treated as an optional workload and matched to the model size the machine can actually sustain.

How Project NOMAD Organizes Offline Resources

Project NOMAD is a browser-accessed command center that coordinates a collection of containerized applications. According to its official documentation, the available modules can include:

• Information Library: Kiwix-based Wikipedia, medical references, ebooks, repair material, and other ZIM archives
• Offline Maps: Downloaded regional map data that can be browsed without a live mapping service
• Education Platform: Kolibri-based lessons, videos, and course progress
• AI Assistant: Optional local models through Ollama, with document upload and semantic search
• Data Tools: CyberChef functions for encoding, hashing, encryption, and analysis
• Notes and utilities: Local notes plus additional applications selected through the command center

Project NOMAD downloads resources while a connection is available, then serves the stored content locally to computers, tablets, and phones after the internet link is removed.

This architecture explains why a browser is enough for normal use. The applications run on the host, while another laptop, phone, or tablet on the same trusted network opens the host’s local address. The client devices do not each need a duplicate copy of the full library.

Step 1: Decide What Must Remain Available Offline

Start with a content plan rather than downloading every available collection. A practical library might combine regional maps, first-aid references, repair manuals for equipment you own, operating-system documentation, a concise encyclopedia, educational material, and personal documents.

Storage requirements rise quickly. The Project NOMAD application alone can run with modest free space, but Wikipedia archives, maps, videos, courses, local documents, and AI models can consume tens or hundreds of gigabytes. Leave unused capacity for updates, temporary downloads, container data, and backups.

Content selection also determines usefulness. A huge library is not automatically better if the material is in the wrong language, covers the wrong region, or cannot be searched efficiently. Build around realistic scenarios and verify that critical pages, maps, and manuals actually open.

Step 2: Prepare the Ubuntu Installation Media

GhostStrats’s distinctive choice is to keep the operating system, Project NOMAD, downloaded content, local AI, and personal files together on a dedicated external bootable drive. This makes the environment more portable and avoids permanently replacing the internal operating system of every compatible host computer.

His Windows preparation stage uses Rufus to write an Ubuntu installation image to removable media.

GhostStrats uses Rufus to select an Ubuntu image and prepare bootable installation media before moving to the dedicated external system drive.

Drive-selection warning: Writing boot media and installing Ubuntu can erase the selected device. Disconnect unrelated removable drives when practical, compare drive model and capacity carefully, and back up anything important before continuing. The installer media and the external drive that will hold the finished Ubuntu system may be separate devices.

Step 3: Install Ubuntu on the Dedicated External Drive

Boot the target computer from the Ubuntu installer through its BIOS or UEFI boot menu, then choose the dedicated external SSD as the installation destination. Exact menu names vary by motherboard. Some systems also restrict external booting or require changes to boot order and Secure Boot settings.

The safest approach is to identify the destination by both capacity and model number. A portable installation should keep its bootloader and system partitions on the intended external device rather than accidentally modifying another internal disk.

Once Ubuntu starts from the external drive, connect networking, install current system updates, confirm that the storage has the expected free capacity, and test at least one full reboot before adding the NOMAD application stack.

Step 4: Install the Project NOMAD Command Center

Project NOMAD currently supports Debian-based operating systems and uses a terminal-based installer. The official quick-start documentation requires administrator privileges, installs the required container stack, and then exposes the Command Center through a local browser.

Because software changes over time, copy the current command from the official Project NOMAD installation guide instead of relying on an old command reproduced in a third-party article. When installation completes, the interface is normally reached through the host’s local port 8080, using either localhost on the server or the server’s local IP address from another device.

Step 5: Download the Knowledge, Maps, Courses, and Models

This is the stage that turns a normal Ubuntu server into an offline resource. While internet access is still available, use the setup wizard and content tools to choose the collections required for the project.

Download one category at a time and confirm that it finishes successfully. Large map regions, video courses, full encyclopedia archives, and AI models can take substantial time and disk space. A partially downloaded collection may not be useful when the connection disappears.

If local AI is enabled, choose a model appropriate for the hardware. A model that technically loads but consumes nearly all available memory may leave the interface slow or unstable. The Project NOMAD documentation recommends significantly stronger hardware for its optimal AI experience than it requires for the bare command center.

Step 6: Add Personal Material and Create a Backup

Public reference collections become more useful when combined with information specific to the owner: equipment manuals, contact lists, medication details, local procedures, certificates, repair diagrams, and copies of important documents.

Protect sensitive material appropriately. Project NOMAD’s current design does not include built-in user authentication, so anyone who can reach the exposed services on the same network may be able to access them. Do not store private documents without understanding that access model, and do not expose the server directly to the public internet.

An external boot drive is also a single point of failure. Keep a second tested copy of irreplaceable personal files and, where practical, a backup of the application data or the complete system drive. A portable server that exists on only one SSD is convenient, but it is not resilient.

Step 7: Disconnect the Internet and Test the Real System

The decisive test is not whether the dashboard works while the server still has internet access. Disconnect the WAN connection, leave only the trusted local network in place, and try the tasks the system is supposed to support.

The final validation should confirm that the prepared resources remain reachable from the Project NOMAD setup without depending on an outside service.

A useful offline acceptance test includes:

• Restarting the host while disconnected and confirming that required services return
• Opening the command center from both the host and a second local device
• Searching several downloaded encyclopedia and reference pages
• Zooming and moving through the downloaded map regions
• Opening courses, videos, PDFs, notes, and personal documents
• Starting the selected local AI model and asking it questions without a cloud endpoint
• Checking power consumption and expected runtime on the chosen backup-power equipment

If a feature fails only after disconnection, it may still depend on a cloud API, remote font, login service, un-cached asset, or content that was never fully downloaded. Discovering that dependency during a controlled test is far better than finding it during an actual outage.

What Continues Working—and What Does Not

Security, Maintenance, and Reliability Matter as Much as Installation

Project NOMAD is intentionally easy to reach on a local network and currently has no built-in authentication layer. Use router rules, firewall controls, and a trusted isolated network to decide which devices can reach it. The project’s own documentation strongly advises against exposing the instance directly to the internet.

An offline library also ages. Reconnect on a controlled schedule to install software fixes, refresh reference collections, update maps, and replace AI models when appropriate. After every major change, repeat the offline test instead of assuming the update preserved the previous behavior.

Finally, plan for the physical system: protect the board and exposed connectors, secure the external drive, label the required cables, document the boot procedure, and verify that the selected power bank or backup supply can deliver the voltage and current the complete setup needs. Knowledge stored locally is valuable only while the storage remains readable and the hardware can start.

Who Is This Project For?

GhostStrats’s build is especially relevant to people who want a compact, self-contained reference system rather than another cloud-dependent service.

• Preparedness users can keep carefully selected emergency, medical, communications, and repair information available locally.
• Travelers and off-grid users can carry maps, manuals, documents, and educational material into places with unreliable connectivity.
• Families and educators can serve learning resources to several local devices without each client storing the whole library.
• Homelab builders can explore Docker, local content services, offline-first design, and local AI on hardware they control.
• Privacy-focused users can keep selected documents and AI interactions on local hardware, provided they also secure the local network.

It is less suitable for anyone expecting live search results, automatically current information, cloud-scale AI performance, or a maintenance-free emergency appliance. The project rewards deliberate preparation and regular testing.

Frequently Asked Questions

Does Project NOMAD work without internet access?

Yes, after the software and desired resources have been fully downloaded. Initial installation, updates, additional content, and new AI models require internet access.

Is the external Ubuntu drive required?

No. It is a portability choice used in GhostStrats’s build. Project NOMAD can also be installed on the normal internal storage of a compatible Debian-based computer.

Can ZimaBlade run the local AI features?

Lightweight models may run on some configurations, but experience depends on the exact processor, installed memory, model size, and available acceleration. The offline information library and other non-AI tools are more realistic priorities for modest hardware.

How much storage does an offline server need?

The base application needs relatively little space, but the selected content determines the real requirement. Maps, encyclopedia archives, courses, personal files, and AI models can increase the total from a small installation to hundreds of gigabytes.

Is Project NOMAD a backup of the internet?

No. It is a locally stored collection of chosen resources and tools. It cannot retrieve live pages or information that was never downloaded.

Can the server be exposed directly to the internet?

It should not be. Project NOMAD currently has no built-in authentication and is intended for controlled local access. Follow the project’s security guidance and use network-level controls.

A Survival Computer Is Built Before It Is Needed

The strongest lesson from GhostStrats’s project is not that one small box can replace the internet. It is that a compact x86 server can preserve a useful, searchable subset of digital knowledge under the owner’s control.

ZimaBlade provides the portable hardware foundation, Ubuntu supplies a familiar operating-system environment, and Project NOMAD organizes downloaded knowledge, maps, courses, tools, and optional local AI behind a browser-based interface. GhostStrats’s external-drive design adds another layer of portability, allowing the prepared environment to move between compatible computers.

The result becomes dependable only after the owner chooses relevant content, completes every download, secures the local network, creates backups, and tests the system with the internet physically disconnected. Watch GhostStrats’s complete Project NOMAD build for the original hardware and installation walkthrough.

Building your own portable offline server or experimenting with local knowledge and AI? Join the ZimaSpace community to compare hardware choices, share your finished build, and learn from other home-server makers.