▲ NOTHING CLASSIFIED OR PRIVILEGED082051Z SEP 26

NEBELKRÄHE · OSINT DESK

THE PRESCIENT DESK™ · PREDICTIVE LEDGER

Forecasts kept on the record

Calling our shots in the fog. Soaring through our misses.

NOTHING CLASSIFIED OR PRIVILEGED

KKR is the Kaos Kontrol Report — the daily forecasting stage of the Prescient Desk. It reads the open-source collation, elicits falsifiable projections from a named forecaster arm, runs them through a mechanical gate that publishes its rejections with reasons, and seals what survives into the ledger before any outcome exists.

Window: this run · source: battle_report_2026-09-08_1518.md · forecaster: control/baserate · 7 accepted / 3 rejected by validation gate · 4 rated below 35% (base-rate discipline)

(U//OS)01. I. VALIDATED PROJECTIONS▲

(U//OS)02. II. REJECTED BY THE GATE — AUDIT TRAIL▲

• "In the Swedish Riksdag election held 2026-09-13, the combined seat total of V, S, MP and C exceeds the combined seat total of M, KD, L and S" → REJECTED: cited items share no substantive vocabulary with the claim — a citation that does not support its entry makes the 4.02f priors unreadable and forces the keyed/keyless call to default; cite an item that grounds THIS claim; measurable claim without a numeric threshold — a row about a quantity must state the number next to its comparator; identifier digits (H.15, S&P 500) do not count
• "The AfD receives the largest share of Zweitstimmen in the Mecklenburg-Vorpommern Landtag election held 2026-09-20." → REJECTED: the resolution narrows the claim with a qualifier the statement never makes — preliminary. The forecaster is graded on the statement; a severity or status qualifier living only in the resolution is invisible to anyone reading the claim
• "At least one painting taken in the Renoir museum theft in southern France is physically recovered by authorities between 2026-09-09 and 2026" → REJECTED: the resolution names a different subject than the statement — the claim is about France, Renoir and the resolution settles on French. A row whose resolution checks a different fact can be scored correct while being wrong

(U//OS)03. III. LEDGER STANDING▲

1743 issued all-time across 16 forecaster arms · 1478 open (101 past deadline — run python kkr.py --resolve). No pooled score is published — a Brier score belongs to one forecaster; an average across arms is nobody's record.

This arm — control/baserate: 537 issued · 509 open · 28 resolved · 14 hits / 14 misses · Brier 0.297 against its own base rate 50.0% (climatological 0.250) · skill -0.187 · under 30 resolved, this is noise.

16 projection(s) voided — terminated as unadjudicable, never edited; each is itemised with its reason in the ledger.

STANDING BY ARM — segregated per RPAS 5.04; no pooled figure exists.

Full ledger: ledger.html (paged viewer) · ledger_full.html (complete static table) · ledger.json (the record)

NOTHING CLASSIFIED OR PRIVILEGED · the gate is mechanical; the ledger is permanent; the system gets scored, not the operator.