ULAP ONE · Command

Sim load is a counter, not an attach: subscriber-sim generates demand and nothing registers on a radio for it. Attached UEs is the evidenced count from /api/ran/ue/attached — a UE holding an IPv4 address on a kernel tunnel interface that exists right now — and it is the same endpoint Control Plane and Co-tenancy read, so the three routes agree by construction. Until 2026-08-04 this tile relayed /ue/list through subscriber-sim, which is every UE we track, attached or not; both read 3 that day, so the mislabel was latent and would have diverged the moment a UE was tracked without an address. The relay also returned real_ue_count: 0 both when nothing was attached and when ran-manager never answered; the canonical endpoint keeps those two apart and prints no number at all for the second. Uplink and downlink come from /api/ran/metrics/host, labelled with that producer's own source field; they used to be read from /api/mode/status, which has never returned either field.

Three states, not two. Answering, deployed and not answering, and opt-in and never deployed. Absence is neither a pass nor a failure, and the tile names the command that would bring it up.

TVWS sensing is a research preview, not a spectrum-access authority. A healthy chip here means the service responded, not that it is correct. On the 28-channel UHF survey of 2026-08-04 — 560 records, 10 occupied / 15 vacant, ground truth derived from measured physics rather than annotation — this checkpoint detected 2 of 10 live multiplexes, including neither of the two strongest carriers in the band, and called 48 of 300 vacant records occupied, at 0.93 to 0.99 confidence. With the physics detector wired in (same unchanged checkpoint, both arms in one process, differing only in whether a calibrated control floor is passed) it detects 10 of 10 multiplexes and calls 0 of 15 vacant channels occupied. Still not clean: UHF 28 asserts OCCUPIED on 8 of 20 records, and that threshold was not moved. This caveat used to blame the threshold — calibrated on synthetic noise rather than real receiver noise, the same defect one layer down — and a measured receiver-noise null falsified that on 2026-08-05: real receiver noise tops out at comb z = 2.289 against the synthetic 2.19, and 0 of 300 records reach the 3.5 threshold. The firing arrives through the antenna and its cause is now open: a genuine emission below the energy floor, which would mean the ground truth needs revising rather than the detector, or front-end intermodulation. The RX-gain sweep that separates them has not been run. The earlier “40 of 80” figure came from a 2026-07-29 set that was entirely occupied: a detector answering OCCUPIED unconditionally also scores 80 of 80 on it, so it measured no discrimination. Rollout is opt-in and the running service was not restarted — with no calibrated control floor configured the engine falls back to the legacy CNN verdict and reports occupancy_source: cnn-confidence, so any live occupancy verdict here is the classifier’s unless a floor is set. Model confidence is not accuracy and not reliability: high confidence has been measured on wrong answers, so the theta_sense = 0.85 gate does not catch them. PAWS grant state is the authority; sensing is advisory.

The two players are the degraded uplink and the SPAN super-resolution output, side by side. Neither panel probes its own geometry, so the captions state the pipeline's configured geometry rather than a measured one.

No service reports which profile is applied, so none of the four above is highlighted and this page does not poll for it. Apply one from the operator host: make ntn-simulate MODE=leo|meo|geo|clear. The profile is applied on the UPF N6 uplink with tc netem; the link becomes high-latency and low-rate while the 5G core keeps serving UEs through the local break-out for the AI workloads.