Antifragile definition
Antifragile describes something that gets stronger when it is stressed. Nassim Nicholas Taleb introduced the word in his 2012 book Antifragile to name the opposite of fragile, which he argued was not robustness or resilience. The book’s own summary puts it in one line: the resilient resists shocks and stays the same; the antifragile gets better and better.
Applied to networks, the distinction is practical:
- Fragile systems break when stressed, and one failure can take everything down.
- Resilient systems absorb a failure and recover to where they were.
- Antifragile systems keep failures small and use each one to get harder to break, so the whole improves as parts are tested.
Internet fragility and shared dependencies
Today’s internet is fragile because much of it is built on large shared dependencies and on devices that anyone can reach.
Shared dependencies fail together. On November 18, 2025, a feature file loaded by Cloudflare’s traffic-routing software on every machine doubled in size, exceeded a limit, and caused failures across the network. One file, pushed everywhere, became a fault everywhere. What is a single point of failure? collects more cases like it.
Reachable devices are recruited in bulk. In 2016 the Mirai botnet scanned the internet for devices answering on Telnet and tried a list of 62 common default usernames and passwords. CISA reported one attack drawing on about 380,000 infected routers, cameras and video recorders. Each was reachable from anywhere, so one automated tool could find them all.
Central stores are central targets. A service that holds everyone’s data is worth attacking once to reach everyone. Apple made the point when describing its private cloud: data that exists only on user devices is by definition disaggregated and not subject to any centralized point of attack.
AI raises the stakes on all three. Automated tools find reachable systems and exploit published flaws faster, so anything that is exposed everywhere is found everywhere sooner.
Antifragile network design
An internet becomes antifragile when it is made of small pieces that work on their own, joined by narrow doors.
- Local coordination by default. Devices that need to work together find and talk to each other directly on the local network or over short-range radio, without a round trip to a distant server.
- One narrow gateway per site. Each local network reaches the wider internet through one controlled point that carries what the work needs. Nothing behind it is reachable from outside.
- Identity that lives on the device. Each device proves who it is with a key it generated and never shares, so there is no central directory whose loss or breach breaks trust everywhere.
- Data that lives with the people using it. The working copy sits on the devices, and the backend is the system of record, not the only copy.
- Many paths. Bluetooth, Wi-Fi, local IP, relays and the internet are all routes, so losing one degrades the network instead of stopping it.
The effect is containment. A bad update, an outage or an attack hits one site, not every site at once. The rest keep running, the failure is studied and fixed, and the fix can reach the others before the same fault does. That is what makes the whole stronger under stress rather than weaker.
CISA’s guidance for operational technology reads like the same blueprint applied to industrial sites: remove internet connections, segment networks, and keep the ability to run systems manually.
Offline Protocol as an antifragile layer
Offline Protocol builds the local layer that an antifragile internet needs. Devices generate their own Ed25519 identity keys and derive their addresses from them, so two devices verify each other directly with no lookup and no certificate authority. Every session is end-to-end encrypted with MLS, including groups. Devices discover each other and route over Bluetooth LE, Wi-Fi and local IP, carry messages for each other across multiple hops, hold them until a path appears, and keep shared state in replicated documents that merge deterministically when peers reconnect. A local mesh needs no internet, no portal account and no API key.
The wider internet is still there, through an internet relay or a gateway you provision against a defined contract, and your backend stays the system of record. Offline Protocol has reached more than 350,000 devices in more than 140 countries, and the local networks they form keep working on their own when the link above them drops. That is an internet that shrinks every shock to the size of one site.