The electromagnetic spectrum is the most contested resource in modern operations. We built the tools to make sense of it.
A strange loop is a hierarchical system in which the top level feeds back into the bottom — creating a self-referential structure that is more than the sum of its parts. It describes systems of emergent complexity that arise from simple rules acting on themselves.
We chose the name deliberately.
The electromagnetic spectrum is a strange loop. Every system that transmits alters the environment that all other systems depend on. Modern spectrum operations are the art of exploiting and controlling that feedback — faster than the adversary, at every level of the stack, simultaneously.
"The electromagnetic spectrum is the invisible infrastructure everything depends on. Whoever understands it first holds the decisive advantage."
Traditional spectrum systems are linear: sense, process, respond. But contested spectrum demands recursive awareness — your jamming changes the adversary's behavior, which changes your signal environment, which changes your decision calculus. A strange loop demands a strange loop solution.
Closing that loop at machine speed requires intelligence that lives where the signals are — not in a distant data center. Strangeloop puts the entire OODA loop inside the sensor itself. Observe, Orient, Decide, Act — all in under 2 milliseconds.
Modern AI can classify signals in under a millisecond on commodity hardware. But the systems deployed to operators in the field were still sending raw data to centralized cloud infrastructure built in the 2000s — introducing seconds of latency and total dependency on network connectivity that doesn't exist in contested environments.
The convergence of edge AI hardware, modern RF front-ends, and purpose-built neural architectures for signal processing creates a fundamentally new capability: a sensor that is also an intelligence analyst. Persistent, tireless, accurate to 99%, consuming 18 watts.
Edge AI inference hardware crossed a threshold. INT8 quantization on modern NPUs delivers performance that was impossible at any power budget three years ago. The physics of the problem hadn't changed — but our ability to solve it at the edge finally had.
We build for the operator who can't wait for a satellite pass, a secure network connection, or a data analyst to finish their shift. Our systems have to work in the worst conditions, on the worst terrain, under the most adversarial electromagnetic pressure imaginable. No exceptions.
We do not build cloud-native systems and port them to edge hardware. We design from the sensor out — every architectural decision starts with the assumption of zero connectivity. Cloud integration is an optional output, never a dependency.
Lab benchmarks on clean signals are necessary but insufficient. Strangeloop systems are evaluated and deployed against signals specifically designed to defeat classification — jammers, spoofed waveforms, adversarial emissions. Our accuracy numbers are from real-world contested environments, not anechoic chambers.
Operators are managing dozens of concurrent signals across thousands of frequencies while executing a mission. Our interfaces surface only the information that requires a human decision. Everything that can be classified, geolocated, or responded to automatically — is. The human is in the loop for intent, not mechanics.
Classified information can't leave a system that has no external network connection. Strangeloop's air-gapped operation model isn't a feature — it's the baseline design. All communications are opt-in, encrypted, and cryptographically authenticated. The system's default state is silent.