The internet was supposed to be decentralized. That was the original promise. Packets would route around damage, information would flow freely, and no single point of failure could bring the whole thing down. Fifty years later, most of the world's digital communication funnels through a handful of cloud providers and cellular carriers who can flip a switch and make you disappear.
Mesh networks offer a different architecture. Instead of routing traffic through centralized servers, devices in a mesh network communicate directly with each other, passing messages from node to node until they reach their destination. No towers required. No ISP approval. Just radios talking to radios.
Bluetooth Mesh and the Local Web
Bluetooth mesh has been around since 2017, originally designed for industrial IoT applications like smart lighting and building automation. But developers have started repurposing the technology for human communication. The range is limited compared to LoRa or WiFi mesh, but the hardware is everywhere. Every smartphone, laptop, and wireless earbud already has a Bluetooth radio inside.
Tools like Bitle are building on this foundation. The project focuses on offline-first communication, letting users send messages and share data without ever touching the public internet. Chat protocols like Bitchat take a similar approach, treating connectivity as optional rather than mandatory.
This matters more than it might seem. During natural disasters, protests, or infrastructure failures, cellular networks often fail first. Mesh networks can keep communities connected when everything else goes dark. They're also harder to surveil and censor, which makes them attractive to privacy-focused builders and activists operating under hostile governments.
Compute Without the Cloud
Communication is only one piece. A more ambitious vision involves pooling computational resources across mesh networks to run software that typically requires server farms.
Michael Neale at Block recently released AnarchAI, a framework for distributing large language model inference across multiple devices. The idea is simple but powerful: instead of sending your prompts to OpenAI or Anthropic, you split the workload across a network of participating machines. Each node contributes a fraction of the compute, and the model runs without any single party controlling the process.
This approach sidesteps several problems at once. It reduces dependence on centralized AI providers. It keeps data local rather than shipping it to distant servers. And it creates a foundation for AI systems that don't require permission from anyone to operate.
Hardware for the Offline Future
Software alone can't build a mesh network. You need radios, and consumer smartphones have limitations. OfflineProtocol is betting there's a market for purpose-built hardware with their IntelligenceMesh device, a dual-protocol mesh node designed to bridge different network types. The device is currently available for preorder.
Dedicated hardware like this extends range, improves reliability, and creates fixed infrastructure that persists even when users walk away. Think of it as the difference between a walkie-talkie and a repeater tower. Both are useful, but one makes the other more effective.
What Gets Built on This
The real question is what applications emerge when connectivity becomes truly peer-to-peer. Mesh networks enable offline markets, local social networks, emergency coordination systems, and communication channels that exist entirely outside the surveillance apparatus.
None of this replaces the internet. But it creates an alternative layer that works when the internet doesn't, or when the internet works against you. In a world where platforms increasingly act as gatekeepers, that alternative matters.
The infrastructure is still nascent. Coverage is spotty, tooling is rough, and adoption remains niche. But the technical foundations are solid, and the use cases are clear. Sometimes the most important technology is the kind that works when nothing else does.
