The Intelligence Division
The Division operates the GISS service mark: a sovereign-grade analytic practice run on ledger continuity. Its flagship product is the Borea Intelligence Brief & Assessment (BIBA) — a daily edition of record. Its method is simple to state and hard to fake: claim openly, pre-register, grade against the locked instrument, correct on the record.
Anatomy of a BIBA.
| Section | Contents | Discipline |
|---|---|---|
| §— | Chair's 1-Page Summary. The day's banner development, three numbered judgments, the ledger line, standing trackers, and the decision queue. | Readable in five minutes; everything else is evidence. |
| §0 | Watch Floor Banner. Posture phase, watchlist counts by state, standing ratio metrics with streak counts. | Phases change on recorded triggers, not vibes. |
| §1 | The Board. Numbered items in two columns of honesty: what sources claim (cited) versus what the platform assesses, with confidence and source tier. | Claimed ≠ assessed. The distinction is the product. |
| §1A | Source Summary Statement. Lead sources per item; single-source items flagged; conflicting framings logged; vocabulary gaps sent to the deception register. | Framing is tracked as aggressively as fact. |
| §2 | FFIR / Commitments Status. Standing commitments and the day's movement — preregistrations, position notes, outreach tracks with days-silent counters and sunset dates. | Silent tracks close on the record, never quietly. |
| §3 | Watchlist & Annexes. Each active W-nn with its state, today's print, and the explicit next watch trigger. | Every watch names its next trigger in advance. |
| §4 | Judgments — Diagnostic Format. Presenting symptoms; a differential of hypotheses; the discriminator that separates them; and why each losing hypothesis fails it. | One hypothesis is adopted. The rejections are published. |
| §5 | Estimates & Products. Resolutions graded ✅ / ❌ against the locked instrument, new pre-registered estimates, and products issued. | The instrument is locked before the outcome is known. |
Medicine's differential,
applied to statecraft.
Differential: H1 ··· / H2 ··· (ADOPTED) / H3 ···
Discriminator: the observable that separates the hypotheses
Rejected because: why each losing hypothesis fails the discriminator
Multi-service tradecraft is standard practice on the floor: a consequential question is read the way Washington, Moscow, London, Tel Aviv, Beijing, and the unaligned services would each read it — then synthesized. Depth is a discipline, not a clearance level.
Forecasts with their
neck on the block.
LIVE RECORD → THE CALIBRATION LEDGER
How an Estimate Is Born
- Claim. A falsifiable statement about the world, dated absolutely.
- Instrument. The exact resolution criteria — locked before the outcome can be known.
- Resolution date. A fixed window. No rolling horizons.
- Probability. Stated up front, so calibration can be scored.
How an Estimate Dies
- Graded ✅ / ❌ against the locked instrument — even when the outcome is inconvenient.
- Nuance goes in the resolution note. The goalposts never move.
- Tallies are reported product vs. control, marked INDICATIVE with the n-size until the n is meaningful.
- Underperformance has a recovery criterion, stated before it is needed.
Standing orders.
| Order | Text |
|---|---|
| STD-1 | Not fiction. Daily products are real reports built on same-day research. Events, quotes, and figures are never invented. |
| STD-2 | Citations are mandatory. Every factual claim carries a footnote with a source tier — S for primary and official, T1 for top-tier outlets, downward from there. |
| STD-3 | HITL-1. The Chair — a named human confirmer — approves every consequential step. Options are presented lettered with a recommendation; the platform never self-executes. |
| STD-4 | TERM-1. The kill switch stands on every decision record. The word is STAND DOWN. |
| STD-5 | Honest grading. Estimates are scored against locked instruments. Nuance lives in resolution notes, never in moved goalposts. |
| STD-6 | Continuity. Numbering never restarts; open items are carried forward. State is reconstructed from the latest artifacts before anything new is written. |
We publish our misses
in the same typeface.
The Protocol
- Step one. A numbered Register entry describes the miss and its cause.
- Step two. A wrong artifact is re-issued as a corrected edition — suffix C — never edited in place.
- Step three. The next daily edition's banner references the register entry.
Why It Matters
Any shop can publish confidence. Almost none publish a corrections register with the same prominence as the product. The Register is the Division's bond: if the ledger ever moves only forward, you are reading marketing, not intelligence.
Audits are standing: Weekly Platform Audits grade the week's editions and tally the estimate ledger; Red-Cell Reviews attack the platform's own frameworks; Scientific Evaluation Reviews attack its rigor. Findings become numbered Implementation Directives.