CLAIM · DEFENSE & SECURITY
“Total Department of Defense contract obligations under Product Service Code group 13 for federal fiscal year 2026, read from USAspending.gov on 16 November 2026, will be 5.9 billion United States dollars, with an interval of 5.0 billion to 8.5 billion United States dollars.”
This claim comes from Feedstock Before Firepower, published on 10 September 2026, and entered the ledger on 10 September 2026. The firm locked a range of 5 to 8.5 United States dollars, billions, nominal, contract obligations with a central estimate of 5.9; a range item is scored by whether the published figure lands inside the band, outside the Brier rule. It resolves by 16 November 2026, only against a dated public source.
Status
Where the claim stands on the ledger, the probability the firm locked, and the dates that govern it.
lockedStatus
5 to 8.5 United States dollars, billions, nominal, contract obligations (central 5.9)Locked range
2026-11-16Resolves by
Locked means recorded and waiting. Its window is still running. Nothing in a locked claim is edited after the fact.
Extraction and settlement
Where the sentence came from, and the public source that decides it.
Where it was madeAIG-DEF-26-007, forward tranche DEF-2-C2; engine record AIG-FWD-0016
Window as stated2026-11-16
How it will be settledResolved by a single read of the USAspending.gov API v2 spending_over_time endpoint on 16 November 2026, using this request body exactly: {"group": "fiscal_year", "subawards": false, "filters": {"time_period": [{"start_date": "2025-10-01", "end_date": "2026-09-30", "date_type": "action_date"}], "psc_codes": {"require": [["Product", "13"]]}, "agencies": [{"type": "awarding", "tier": "toptier", "name": "Department of Defense"}]}}. Three parameters are stated explicitly because each of them moves the number and neither is the same under a plausible alternative choice: date_type must be action_date, because date_type "date_signed" returned $3.832 billion against action_date's $4.739 billion on the same filters on 10 September 2026, a 19 per cent difference; subawards must be false, because subawards true returned $1.738 billion on the same filters on the same date; and award_type_codes is deliberately omitted, which was verified on 10 September 2026 to return the identical value to an explicit contracts-only list and to a contracts-plus-IDV list, because the product-code filter already confines the result to procurement. The field read is aggregated_amount for the single FY2026 row returned, expressed in United States dollars and converted to billions. On the awarding-agency name: the toptier agency is read as whatever name USAspending then uses for the agency currently published as "Department of Defense" (toptier code 097). On 10 September 2026 the literal string "Department of War" returned $0 on this endpoint, so a zero or near-zero result is to be treated as a query failure and the agency name re-checked against the endpoint's agency autocomplete before the claim is scored, never as a reading of the series. The claim scores as a hit if the value falls at or between 5.0 and 8.5 billion dollars, and the point score is the absolute difference between the read value and 5.9 billion dollars. If the endpoint or the filter syntax has changed by the resolution date, the equivalent query on the successor endpoint is used and the substitution is recorded. The read date is fixed at 16 November 2026 and is not moved for data completeness, because the series continues to settle after fiscal year end and a floating read date would make the claim unscorable.
Where the probability came fromLocked point and interval approved with AIG-DEF-26-007; immutable engine record AIG-FWD-0016; outside the Brier rule
Record classbrier native
Claim familyus-munitions-obligations-fy2026; used to avoid presenting related claims as independent trials.
Learning roleoperational. Holdout claims may score publicly but never train the Learning Loop.
Extracted2026-09-10, against the forward-claim standard
The series it resolves against
The data series, if any, whose values settle the outcome.
This claim names no series the Dashboard carries; it resolves against the public source recorded with it.