Don’t waste
a day.

Small early signals — a search at 2am, a court filing in a small district, a complaint to a federal agency — ripple into massive macro outcomes months or years later: unemployment spikes, housing collapses, civic crises. Most forecasters wait for the storm to confirm itself in lagging statistics. Vinyl listens for the wings.

Every other forecaster reads lagging indicators — GDP, unemployment, housing starts. Those numbers tell you what already happened. Vinyl reads behavioral leading indicators — what people actually search for, file in court, and complain about — months before official statistics confirm what's already in motion.

EQ > (IQ + AI). All the intelligence and AI in the world doesn't matter if you can't read what people are actually feeling and doing right now. Vinyl is the framework for reading that frequency at population scale, with peer-reviewable methodology and replicable code.

dontdiebutterfly.com. Butterflies can die their first day alive. Don't waste a day. Don't miss a signal that was right there the whole time.

Q1 (search leads bankruptcy filings): 4/4 cities, all p < 1×10⁻⁵ — a generalizable law

Q3 (search ↔ unemployment, inverse): 4/4 cities, all p < 0.001 — a generalizable law

Seven (city × target × horizon) cells survive every audit we publish: shuffle-target placebo at n=30 (and n=100 stress), triple gate, training-cutoff invariance at MIN_TRAIN_MONTHS {36, 48, 60}, walk-forward target-leak audit, and model-choice invariance against AutoARIMA / AutoETS / Theta. Six cells are FEATURE_DRIVEN. One — Chicago HPI 3-month — is MODEL_INVARIANT.