PRODUCTION SYSTEM · TRL 6–7 · FIELD-PROVEN
A fleet that thinks
for itself.
FrogNet is a sovereign, self-forming mesh with a real-time network shared memory — one live picture that supports simultaneous AI control of every connected device — no cloud, no base station, and a link that holds even when it's constrained or contested.
A complete, sovereign network — mobile, self-forming, self-healing.
Each node is a whole network on its own, the size of a deck of cards. Nodes discover each other and mesh into a pond you own end to end — forming and healing themselves with no operator, even on the move — and its shared memory is a live database every node reads and writes.
A machine that depends on a link to base is dead the moment the link is jammed.
Jam one link and the vehicle goes blind, deaf, and dumb. Every layer it depends on is someone else's.
Each vehicle is a complete node. They mesh directly, share state, and keep coordinating with no base, no cloud, no link to cut.
The network sees every vehicle — and steers them all.
Every vehicle publishes what it senses and reads back what to do. A network-hosted AI sees the whole fleet's state in one place — and writes control to any vehicle in real time, with no message-passing an adversary can jam, spoof, or sever.
The sense-act loop closes through the shared state — and it keeps running through the loss of any vehicle, any link, or the internet entirely.
A closed sensor → AI → actuator loop, through a database 2,400 miles away, over a 4,800-baud link at 20% loss — in under 2 seconds.
If it closes transatlantic over jammed dial-up, it closes across a field of vehicles.
Coordination that does not stop.
Conventional links are binary — working, or dropped. FrogNet's video steps down by resolution, then frame rate — then sheds to audio, text and canned messages, climbing back as the link recovers.
It never cliffs — it drops a rung and keeps the link alive. Only total jamming — zero bytes through — ends it.
Vehicles discover each other and mesh with no configuration, merge with other units on contact, and heal when one is lost. There is no central node for an adversary to take out.
A 900 MHz HaLow node held the mesh continuously through dense urban driving — across multipath, building occlusions, and the metal of the vehicle itself. The exact RF nightmare of a moving ground vehicle.
The same architecture runs a private family network and a first-responder mesh — genuinely dual-use, civil through defense.
Dan has already proven the hard parts.
An autonomous, re-taskable fleet has to sense and act, stay meshed on the move, pull in any sensor, join from anywhere, and be reprogrammed in the field. Daniel Tone has demonstrated each on FrogNet, on his own nodes in New York — an independent reproduction of the whole system, the latest completed today.
Dan drove a physical actuator from a web UI — every step routed through a database 2,400 miles away, over a 4,800-baud link at 20% loss, in under 2 seconds. The closed control loop of an autonomous vehicle, proven over a worst-case link.
Dan ran a trunk-mounted HaLow node through dense urban driving in Queens — no dropped links across multipath, building occlusion, and the vehicle's own metal. A moving platform that never loses the swarm.
Dan bridged MeshTastic and LoRa sensor data straight into the shared memory, end to end — RF-delivered readings the whole mesh sees at once. Field sensors over the radios a unit actually carries.
Dan stood up a node on nothing but a Starlink uplink — it joined the mesh as a full member in minutes, SSH-reachable and writing to the shared database. Drop a unit anywhere with a sat link and it's on the network.
Dan pushed a remote firmware update to the ESP32 boards — repurpose any sensor or actuator across the fleet with no physical access. Re-mission the hardware from afar — a soil sensor becomes a perimeter tripwire.
The building blocks of an autonomous, re-taskable fleet — already working, independently. Next is the fleet: a pilot on a handful of units, then the 500.
And he took it from concept to execution in a few days — the leverage is in the ideas, not the labor. Shift the paradigm, stay adaptable, and a few people do what used to take an army.