The Offline Atlas: A Peer-to-Peer Learning Mesh for Privacy-First Education

The Offline Atlas is an ambitious plan to reimagine educational technology from the ground up, putting privacy and autonomy first. Imagine a school, library, or community center where students connect their devices (old phones, Raspberry Pis, laptops) to a local mesh network — no internet required. The Atlas software creates a distributed library of curated learning modules, discussion boards, and project templates that replicate and synchronize across devices as nodes join the mesh. Every piece of content is stored locally; nothing is ever sent to a central server. The system uses Sybil-resistant identities (self-signed keys with optional physical meetup verification) to prevent spam while preserving anonymity. The curriculum is modular, covering everything from algebra to blacksmithing, contributed by teachers and peers, and rated by local reputation. The goal is to create an educational ecosystem that works in disaster zones, remote villages, or tech-free classrooms — and that cannot be monetized, surveilled, or algorithmically manipulated. This project never got built because the team scattered after a funding gap, but the prototypes proved the mesh routing and content syncing worked over five kilometers.

Traditional EdTech platforms collect vast amounts of student data, often without meaningful consent, and require constant internet access. This excludes communities with poor connectivity and forces learners into algorithmic funnels that optimize for engagement, not curiosity. Teachers and parents are increasingly uncomfortable with their children's data being used for profit or surveillance. There is a need for an educational tool that is intrinsically private, works offline, and fosters genuine exploration without a corporate agenda.

• Build a fully offline peer-to-peer mesh network that can span up to 5 km using Wi-Fi ad-hoc mode and LoRa backup.
• Create a content distribution protocol that syncs modules, assignments, and discussion posts across all nodes without a central server.
• Develop a modular, open-source curriculum covering K-12 core subjects plus trades and arts, with version control and local forking.
• Implement Sybil-resistant identity system using self-signed certificates and optional in-person verification via QR codes.
• Achieve zero data collection: no telemetry, no analytics, no login tracking; everything is anonymous and ephemeral.
• Package the software for low-cost hardware like Raspberry Pi Zero W and old Android devices, making it deployable for under $50 per node.

• We will not build a web-based version or any cloud dependency; the system must work with zero internet.
• We will not include any tracking, analytics, or user profiling, even anonymized.
• We will not create a marketplace or any monetization layer; all content is free and share-alike.
• We will not support real-time video streaming; the focus is on asynchronous text, images, and small files (under 10 MB per module).
• We will not enforce a rigid curriculum; teachers and learners can modify and remix any module.

• Mesh Networking: batman-adv (B.A.T.M.A.N. Advanced) for Wi-Fi mesh routing; LoRa (RFM95) modules for long-range low-bandwidth backup.
• Content Syncing: IPFS (InterPlanetary File System) with local-only pinning and custom CRDT for offline-first edits.
• Identity: libsodium for Ed25519 key pairs; QR code exchange for manual verification.
• Hardware: Raspberry Pi Zero W (primary node), Android 5.0+ devices (client nodes), USB power banks.
• Frontend: Lightweight React Native app for mobile, Electron desktop app, both offline-capable with SQLite local database.
• Curriculum Format: Markdown + YAML frontmatter, with assets embedded as IPFS CIDs.

The system is organized as a mesh of nodes, each running the Atlas Daemon. When a node boots, it scans for nearby Atlas peers using Wi-Fi ad-hoc. Once connected, it exchanges a list of content CIDs (content identifiers) and syncs missing modules via a gossip protocol. Each node maintains a local copy of the entire curriculum (if storage permits) or selectively syncs based on user preferences. Discussion posts are stored as append-only logs signed by the author's key; reputations are computed locally from the graph of endorsements. The architecture is fully decentralized: any node can leave and rejoin without data loss. There is no single point of failure, no central authority, and no connection to the internet required.

• Mesh networks can be unstable with many nodes; need to ensure effective routing and conflict resolution.
• Content moderation is difficult without central authority; local reputation systems may be gamed or lead to echo chambers.
• Physical meetups for identity verification may limit usability; alternative methods (e.g., web of trust) need exploration.
• Offline-first sync can lead to stale data; user education about manual sync triggers may be necessary.
• Legal uncertainty around distributing certain educational content (copyright) — must rely on open-licensed material only.

• How do we handle large media files (e.g., video lectures) within the 10 MB per module limit? Possibly use external USB drives as 'cold storage' nodes.
• Should we implement a 'teacher node' with admin privileges for classroom management, or keep all nodes equal?
• How do we prevent spam in the discussion system without revealing identities? Possibly use hashcash-style proof-of-work for posting.
• What is the optimal trade-off between storage and accessibility for low-memory devices?

The core team of three developers and two educators prototyped the mesh networking and basic sync over one summer, but then two members moved for jobs and the third had a child. The project wasn't abandoned — it just lost the critical mass to push through production. The code lives on GitHub, awaiting someone to rekindle the flame.

This project was inspired by the concept of 'sneakernet' and the resilience of mesh networks used in protest movements. The curriculum design intentionally mirrors the structure of Wikipedia's Wikibooks, but with a focus on hands-on projects. A notable early success was syncing a full science module across five Raspberry Pis in a park, with zero data leaving the local mesh.