Loop
How it runs
Issuer’s text, not a URL paraphrase. Remixes matched line by line.
Quoted body, one-line source, tag or a reason not to tag.
Does the actor hold the required capability, as of when?
Move History, or hold. Fired-and-held is not an alert.
Only rising / imminent warnings. Developments until 1,000 words.
The loop is files, not a dashboard. An observation is written into a log. The log tags a pre-defined indicator or records why it does not. Capabilities bound whether fired indicators can produce the warned outcome. Likelihood is a judgment from both. Intent is usually the assumption and is stated. A warning moves only when likelihood changes rung. A fired indicator that leaves the rung unchanged is logged, not alerted.
Observation → brief
- Retrieve the artifact. The source’s own text, not a paraphrase of a URL, not an aggregator’s synthesis. A social post that cites a wire is a remix: fetch the cited issuer, match line by line, and flag every addition the post made.
- Log. Heading is the event date plus a concrete slug. Tags, capability, logged date. One-line source (who, how they know, date, URL). Body in the source’s own words, quoted. Verification notes go under the entry, dated, never rewritten.
- Tag or refuse to tag. An indicator ID is applied only when the observation matches the indicator’s wording. Stretching a tag to have one is a miss. Untagged entries name the nearest indicator and why it fails.
- Bind to capability. Does the actor hold what the outcome requires? Cite the capability entry and its as-of date. Unassessed capabilities keep confidence low.
- Move the rung or hold. A move is recorded in the warning’s History with date, what changed, and the source that drove it. No move: log only.
- Brief from the log. The brief describes events and judgments. Record-keeping stays in the files. The brief ends when the word budget is spent.
Warning vs watch
A warning has a warned outcome, a window, pre-defined indicators (including disconfirming), required capabilities, a current likelihood on the ladder, and History.
A watch has no outcome and no likelihood. It has a bar. An item that clears the bar may appear under Developments. Most days most watches produce nothing; that is the expected result. When items under a watch start pointing at a definable outcome with a window, the next step is a warning — not a looser watch.
If the candidate outcome is already true, it is recast as a duration or threshold test so the warning is about persistence or expansion, not the current state. Named short outages stay indicators. “The war exists” is not a warned outcome.
What enters the brief
A brief covers what was logged since the previous brief (or 24 hours if unknown), at most 1,000 words.
Warnings appear only when likelihood rose this period, or while the warning is imminent: standing at likely or higher, or able to occur within about a week. A fall gets one line. Every other warning is omitted. Silence means it stands where it was last reported. If none qualify, the section is exactly: Warnings: none rising or imminent.
Developments are items that clear a watch bar, in order of significance, until the budget is spent. Fired indicators that did not move a rung belong here as ordinary items: name the indicators and state that the rung holds.
Scored forecasts
A time-bounded call at likely or higher with a window of 30 days or less is frozen as a card. The live rung can still move; the card does not. Resolution is binary on the warned outcome inside the window. Hit: the outcome indicator fires; resolve immediately. Miss: the window closes without it.
The first live test is F1 (W7, US diesel export embargo, window through 23 October 2026). A miss is treated as a process failure — after-action and framework changes in the same turn — not “likely misses sometimes.”
Provenance
Every change to a warning or a watch carries date, what changed, and the source that drove it (publication and issue date, URL, or the principal’s instruction). Warnings record it under History; watches under a dated Changes line. This covers new entries, rung moves, and edits to outcomes, indicators, capabilities, and bars.
What is deliberately not in this packet
The live observation log is not reproduced in full. It contains paid-feed text, working notes, and volume that does not help a reviewer. One public-source entry is included so the log shape can be judged. Ingest plumbing (mail folders, job identifiers, local paths) is omitted. The PDF of the brief is the external-share form: no INTERNAL marking, with the live warning/watch catalog as appendix.