The first time we shipped Ajna's mesh, three iPhones in a coffee shop in Berlin discovered each other in 380ms and finished a 4MB exchange before the espresso landed. The second time, in a hotel basement with no internet, it just worked. That second moment is the one we kept chasing.
Going from five devices in a lab to five thousand in the wild took eighteen months and three protocol revisions. Along the way we threw out two transports, learned that BLE is not a Wi-Fi replacement (it's a different animal entirely), and discovered that the hardest part of a peer network isn't the peers — it's recovering gracefully when half of them disappear at once.
The web spent twenty years assuming a stable connection. Most of those years, that assumption held — in cities. The rest of the time, especially on the move, we ran into the same wall: latency, captive portals, rate limits, paywalls, deplatforming. Mesh networks have always offered an alternative, but the bandwidth used to be too low and the discovery too painful.
Two things changed. Bluetooth 5.4 made device-to-device throughput practical (we're seeing 600KB/s sustained on iPhone 14+), and DIDComm v2 gave us a clean identity primitive that doesn't require a server. With both in hand, it became cheaper to ship a mesh than to keep building around its absence.