Open Calls · Indications & Warning
Watch a sealed call resolve in real time — open sealed windows plus telegraphed targets awaiting seal (next: an ISRO GSLV launch).
Open calls — the live indications & warning board
Sealed-OPEN forecasts you can watch resolve in real time, plus telegraphed targets awaiting seal. Each open call is timestamped on the public record before its window closes, so you can grade it yourself as the outcome lands. Open now: the Dalai Lama transition (four dated windows through 5 Feb 2027) and the US markets & economy boom called at the tariff-crash bottom against a ~60% recession consensus (first leg a HIT — no recession, S&P to record highs — thesis ongoing).
Why this board answers “do a forward test”
And the reconciliation, explicitly: none of this resists a real forward test — it prices the moving goalpost. A denominator-fixed forward season with external adjudication is the record’s own stated completion (§7 of the paper, protocol published); any counterparty can convene it today — the buyer-adjudicated Proving Round is the smallest version. What is refused is the rhetorical demand, never the test.
Reality is the referee. Every call is sealed before its window opens, timestamped to a host the operator does not control, and graded by reality — NASA, the FAA, the provider — not by us. “Do it forward” is a circular demand: every call here was a forward test, and the instant it resolves it becomes “past data,” so asking for a new one is a goalpost that moves every time it is cleared — there are already dozens, each cryptographically proven to predate its outcome, which is exactly what lets you treat a resolved call as if you had watched it live. Only against-consensus calls are sealed: the room says go, this desk says no. A base-rate model and an off-the-shelf LLM both follow consensus, so on exactly these calls both are structurally wrong by construction. This desk even drops its own hits when they agree with the room — the correct Mamdani call was excluded from the score because he was the frontrunner. Selecting for against-consensus downside is therefore selecting the hardest, highest-information calls, not the easiest, and selection bias degrades an estimate of an average hit-rate, never a demonstration that a specific against-consensus call can be made at all.