Most tools show you what worked.
This one starts with what didn't.
Three instruments. One rule — trust only what survives measurement.
Two eyes see the same world from two positions.
The distance between them is how you measure depth.
Offset on purpose. Judged as one.
Weak-signal research foresight.
76 arXiv topics, scored on 5 dimensions. It reads the frontier as a map — and catches ideas rising before it's obvious.
523 US names, scored on 6 dimensions. Every signal must clear one gate — event-study certification: t-stat · bootstrap CI · FDR · deflated Sharpe.
So here is the only honest question:
did that timing move the stock?
One frontier. One market. Two lenses on one bench.
One gate. One registry.
13 trials · 12 p-values · one FDR family — so neither instrument fools itself.
Certified: 1.
Everything else is withheld.
That is the point.
And our thread — frontier → firm → price?
Three pre-registered onset tests.
0 of 3 pass.
The lead is real — but the placebo shows it's the ordinary commercialisation lag, not onset.
We show you what didn't pass.
That's the point.
One certified signal is worth more than a hundred that were never tested against a null.
Two lenses on one bench — the research frontier, and the equity market.
Most tools show you what worked; this one starts with what didn't. Across both platforms the trials are counted in one Benjamini–Hochberg family, so neither instrument fools itself: 13 trials, 12 p-values, exactly 1 certified (drawdown_reversion) — everything else withheld. The research→firm timing story is measured precisely and refused honestly (the transmission ladder is 0 of 3). That discipline is the product.
Three doors · pick a lens