All Stories

Investigation · Space Weather  ·  We pooled thousands of decaying Starlinks into one drag sensor and aimed it at a moderate geomagnetic storm. The fleet decay rate nudged the right way but cannot separate from its own day-to-day scatter. An honest null from a working instrument.

Investigation · Cross-Match  ·  9,379 US M ≥ 4.5 earthquakes (1910–2014) against 71,008 NUFORC sightings. The folklore predicts clustering near epicenters; we observe 3 coincidences where 9.4 are expected by chance.

Investigation · Cross-Match  ·  On the 10 days when crew tweets pin the filming calendar, no public sensor records a concurrent anomaly.

Near-Earth Objects  ·  AMS reported eyewitness fireball counts running 3.9σ above normal in early 2026. NASA's CNEOS sensor catalog (357 deduplicated events, 1998–2026) says no — and rejects the AMS-equivalent effect at p = 0.015 against the modern-sensor baseline.

Seismology  ·  Japan Trench, May 15, 2026. Zero M4+ events in the 30-day pre-window. Statistically independent of the M7.4 twenty-five days earlier.

Meteorology  ·  9.6M GOES-16 GLM flashes at 4 spatial scales. V3 signal recovers at k_min = −270 min via SPC timestamps but vanishes at every GLM-triggered scale. QRN + cell-local D-layer: disfavored.

Meteorology  ·  Oklahoma 2024 joins Mayfield 2021 as a second clean replication. Outbreak-vs-dose Δd is not Bonferroni-significant. Precursor is general to any tornadic convection.

Seismology  ·  M6.0 + M5.9 south of the Kermadec Islands, April 16-18. Bath's law rejected at p = 0.006. Statistically a doublet, not an aftershock sequence.

Meteorology  ·  Mayfield 2021 replicates the 40m tornado precursor at k_min = −270 min. Pooled Fisher p = 1.12 × 10⁻⁴. 2011-2013 events are null by measurement density, not by signal absence.

Meteorology  ·  A 10-day CONUS WSPR study finds HF signal depression that begins 165 min before confirmed tornadoes, scales with D-layer physics, and survives a forecaster-lag control. N = 106 tornado events.

Fact-Check · Seismology  ·  23 events off Washington — 400 km west of the subduction trench, on a different fault system entirely.

Research · Pre-Registered Null  ·  We locked the methods, pulled 5.3M WSPR spots and 934K aircraft positions, and ran the test. Every bin had an aircraft. The control group was empty. Here's what a null result can still tell you.

Breaking · Seismology  ·  47 events in hours — Walker Lane seismic zone lights up with the strongest quake near Reno in decades.

Research  ·  Two complete solar cycles of WSPR data say the 10 m band gets worse at solar maximum. The textbooks aren't wrong — the metric is. Inside a clean selection-effect bug.

Solar Watch  ·  G2 storm, M7.5 flare, 9 CMEs, and our first cosmic ray Forbush decrease — all in 3 days. We tracked it across 7 data streams. The comet CMEs missed Earth.

Investigation  ·  80,000 UFO reports cross-referenced against fireballs, storms, and 10,000 Starlink satellites.