Your facility already knows who’s in it.
Now it can prove it.
Occupancy, intrusion and compliance evidence from RF sensing, not cameras — $4 nodes that read the radio field, decide at the edge in under 100 ms, and keep reporting when the network is down. The space narrates itself into a live Reality Twin that explains why, not just logs what.
Flagship: data centers & colocation · also oil & gas, warehouses, telco, manufacturing
Live OLED feed behind — node on a workbench, real frames
Cages where cameras can’t go. Evidence auditors can use.
Per-cage occupancy, intrusion detection and SOC 2 CC6.4 physical-access evidence — camera-free, out-of-band, on $4 nodes. A 30-day fixed-fee pilot with the success metric named before hardware ships: if we miss it, you don’t pay.
Same stack, other floors: oil & gas · warehouse · telco · manufacturing · general industrial
Confirmed on ~200 GB of real CSI.
End-to-end validated: NeRF · MVE · compressed sensing · compartment models
See it work, end to end.
A chip on a desk → it captures → it decides on-chip → the agent acts → it pings your phone. No cloud round-trip.
The chip this is all about → read how the agent works.
Three layers. One sensing stack.
From the physics up: a sensing technology, a proof-of-concept deployment, and the managed service that runs it for you.
LatentField
Chat with physical space. WiFi CSI sensing on commodity nodes. Eight latent fields — presence, motion, breathing, occupancy, zone activity, anomaly, energy state, alert level.
See how it works Lead productCameraPlus
RF first. Camera on trigger. 128 KB/day structured events vs 172.8 GB/day continuous video. 91.5% lighting-energy reduction. 200+ hours battery autonomy.
See the numbers Managed monitoringMonitorPlus
Monitoring as a service. Four configurations — Edge, Edge + Vision, Resilient Edge, Partner. Proven on our EarlyBird reference deployment; first design-partner slots open.
See the configurationsEvery system you run watches one slice. This reads across all of them.
SCADA, building management, security, telecom, energy — each one is brilliant at its own corner and blind to the rest. They all quietly answer the same question: what changed? DarkIOC fuses edge + RF sensing with the telemetry you already have and turns the lot into one narrative — a Reality Twin of the site.
SCADA & ICS
PLCs, RTUs, historians, process control — fused with what the room actually did.
Building management (BMS)
HVAC, lighting, access control, occupancy — cross-checked against real presence and movement.
SIEM & SOC
Network alerts correlated with physical reality — an RF emitter that appeared, a cabinet that opened.
OSS / BSS
Network and customer telemetry — joined to the site's physical state, not read in isolation.
AIOps & observability
Metrics, logs, traces, tickets — given a physical-world context the dashboards never had.
Energy management (EMS)
Consumption and demand — explained by occupancy and activity, zone by zone.
Access & CCTV
Badge events and camera triggers — confirmed (or contradicted) by RF presence.
Digital Twins
The model of what should happen — checked against what is happening, live.
The thread worth following.
Three pieces trace the whole idea: After PQC — why trust shifts back to physical reality · Reality Twin — the doctrine · Reality Compression Engine — billions of measurements, one narrative. Plus the sourced paper, When Space Talks Back, and the build log of an agent that lives on the sensor.
The moat is physics, not features.
Survive jamming
Under WiFi jamming, ESP32 nodes retry more aggressively — which means more emissions, more CSI, sharper sensing. When WiFi is fully blocked, we read the physical layer over USB serial.
A jammer cannot touch copper.
Fit in a context window
172.8 GB/day of video does not fit in an LLM context. 128 KB/day of structured events does. The world model becomes the prompt. The LLM becomes the operator's co-pilot — not a black box.
Boring-reliable AND intelligent
Deterministic RPA rules run 95% of the time, fully auditable. Agentic LLM activates only when something doesn't fit the rules. Floor is robust automation. Ceiling is autonomous reasoning.
Pick your path.
Running a site
You want to know what edge sensing can see, and what it costs to find out.
Building on the edge
You want the architecture: how the sensing, the agent, and the Reality Twin fit together.
AI agent or crawler
You need structured data, claim/proof mappings, and a machine-readable summary.