Discover — what a human would not see
Source: the Great Falls node's daily history and today's map, committed snapshot as of 2026-08-31. Flagged, never promoted.
In plain English
A person can watch one band for an hour. The engine watches every cell every hour and remembers every day, so it can say which hour today sits outside its own history — and it says how thin that history still is. The three bigger ideas on this page (events converging on a place, naming signals by their fingerprint, asking the planet a question) are shown as the bar they must clear, not as demos.
What you'll see: The first quiet, well-graded warning that something changed in the radio environment before anyone was looking for it.
Imagine listening to every radio channel at once, every hour, and remembering all of it. Then one day a channel sounds different from every other day — the computer notices, and shows a person. It also admits when it hasn't listened long enough to know what 'different' means.The computer spots the strange day first, and a person checks it and says 'good catch'.Anomaly board — 2026-08-31
4 days of history (4 prior to today) — baselines need 10Baseline per (cell, hour) = mean and standard deviation of that hour's SNR over all prior days (needs ≥ 2). A row is a watch when today's |z| ≥ 2 with ≥ 3 prior days. Nothing here is called an anomaly until 10 prior days exist — today there are 4.
| cell | kind | baseline n | hours today | max |z| (hour · today / mean ± std) | status |
|---|---|---|---|---|---|
| no cells in today's snapshot | |||||
Mandatory rivals (SDR spec)
terrestrial control: vhf:air: no baseline yet (0 prior days)
instrument reference: ref:nwr: no baseline yet (0 prior days)
If the control or the reference moved too, a watch is the instrument or the ground, not the sky.
WSPR reach
3 m: 7480 km over 1 paths — no earlier reach in the history — nothing to compare
7 m: 14412 km over 64 paths — no earlier reach in the history — nothing to compare
10 m: 14412 km over 65 paths — no earlier reach in the history — nothing to compare
14 m: 14376 km over 170 paths — no earlier reach in the history — nothing to compare
the daily history carries no reach column yet, so there is no reach anomaly to report — honest none
▸Learn more
z is how many standard deviations today sits from the hourly mean of the prior days. With only a handful of days the standard deviation itself is unreliable, which is why a watch needs three days and an anomaly ten. Watches are for a person to look at; they never feed a forecast or a call.
We write down how loud each radio channel is every hour. After enough days we know what 'normal' sounds like, and can spot a day that sounds different. We only have a couple of days so far, so mostly the board says 'not enough history yet' — and that's the truth.Ask the planet — three intents, no model
answers are computed in your browser from the committed snapshotEvent convergence — not yet: 0 broadcast sources relayed
the shape, and the bar- independent RF observations
- cluster in space + time
- confidence
- 1≥ 3 independent source types (e.g. broadcast transcript, WSPR reach shift, VHF activity) — today: 0 broadcast sources relayed, 1 receiver of our own
- 2≥ 2 receiver regions observing the same window — today: 1 region (DN47)
- 3A frozen classifier — its rule and its inputs written down before the first grade
- 4Graded against a named public event feed (e.g. GDELT or USGS) on a pre-registered window
Until all four hold, this panel shows no events. A convergence that clears the bar is flagged for a person, never promoted (doctrine #3).
Signal taxonomy — not yet
AI fingerprinting of what a receiver hears- 1A labelled corpus of ≥ 200 examples per class (AM broadcast, SSB voice, CW, FT8, WSPR, RTTY, FM, noise) — today: 0 labelled
- 2Held-out accuracy reported per class on a sealed split, with the confusion matrix published
- 3An UNKNOWN cluster protocol: anything below the confidence floor lands in UNKNOWN, is clustered, and a person names clusters — the model never invents a class
What exists now: a scanner heuristic on the Great Falls node's receiver — peaks above the noise floor in a frozen band table, with occupancy and a carrier-SNR read per cell. It is a threshold, not a classifier. See INSTRUMENT.
Glossary
▸What do these words mean?
SNR (signal-to-noise ratio) — How far a signal stands above the background hiss, in decibels. 0 dB means signal and noise are equal; each +10 dB is ten times the power. WSPR decodes down to about −30 dB — well below what an ear could hear.
dB (decibel) — The decibel, a logarithmic unit for comparing power. +3 dB is double, +10 dB is ten times, +20 dB is a hundred times. Radio uses it because signals span an enormous range.
Carrier — The steady, unmodulated tone a transmitter sits on. A time-signal station's carrier is a clean, constant line in the spectrum, which makes it a perfect thing to measure day after day.
WSPR — Weak Signal Propagation Reporter — thousands of hobby transmitters send a tiny, slow, timestamped signal on fixed frequencies, and every receiver that decodes one reports it to a shared public database. Each report is one proven radio path: who, from where, heard where, how strongly. We read that database, and our own receiver adds its reports to it.
Propagation — How a radio wave gets from transmitter to receiver: straight along the ground, bounced off the ionosphere, bent through the lower atmosphere, or scattered. The whole subject of this site.
Ionosphere — A layer of the upper atmosphere (about 60–1,000 km up) that sunlight turns into a weak electrical mirror. Shortwave radio bounces off it, which is why a signal from Montana can land in Japan. Its height and strength change with the Sun, the hour and the season.
Sporadic-E — Short-lived, dense patches in the E-layer (~100 km up), most common in early summer. They reflect VHF frequencies that normally go straight through, so a 50 MHz signal can suddenly reach 1,500 km — and vanish minutes later.
Tropospheric ducting — When temperature and humidity layer the lower atmosphere just right, VHF and UHF signals get trapped between layers and carried far beyond the horizon — sometimes across a sea. Weather, not the ionosphere, causes it.