Event density and VIDI dispersion define the live research setup
GDELT last-good replay: military-escalation (live pull failed; serving prior window) The live reference set now spans 8 observations across 4 source families, with 7 of 14 monitored symbols outside the clean-state baseline.
Thesis
GDELT last-good replay: military-escalation (live pull failed; serving prior window) The live reference set now spans 8 observations across 4 source families, with 7 of 14 monitored symbols outside the clean-state baseline. The argument is held only while the strongest readings remain corroborated by VIDI state dispersion and fresh source families; if either side decays, the research note should be retired rather than softened.
Confidence 92% · N=8 live observations · 7-day evidence window
Reference Set
N = 8
Source Count
4 sources
Stress States
7 / 14
Avg Conviction
55%
The Argument
The current research question starts with event density rather than a broad market index. GDELT last-good replay: military-escalation (live pull failed; serving prior window) That row is not enough on its own, so the producer keeps the full live reference set in view: 8 real observations, 4 distinct source families, and a VIDI map where 7 symbols sit in state 1 or state 2 while 7 remain clean. The structure is useful because it separates an isolated event from a transmissible setup. If the strongest reading decays and the VIDI map returns to clean-state concentration, the thesis fails; if the evidence set keeps broadening, the note stays live.
§ 02Mechanism and transmission
Why the sequence matters
Transmission is visible when independent evidence rows describe the same pressure from different angles. The second-ranked row is event density on GLOBAL: GDELT narrative density spike: military-escalation — 250 articles in 24h (live verification 2026-05-17) The third-ranked row is form 4 cluster on HAWK: 13 insiders filing Form 4 on HAWK (HawkEye 360, Inc.) in 5 sessions Together, these do not create a forecast by themselves; they create a monitoring frame. Liesel's read is that the desk should watch whether physical, geopolitical, smart-money, and market-state evidence stay synchronized long enough to change risk budget behavior.
§ 03What invalidates the read
The failure test
The invalidation path is explicit. The note should be retired if fresh evidence falls below four rows in the seven-day window, if fewer than eight VIDI snapshots are available, or if the strongest source family stops confirming the same direction. The highest-conviction monitored symbol is AAPL at 98% conviction with state 0; that helps locate where the market state is clearest, but it is not a standalone trade signal. The research conclusion remains conditional on the next independent confirmation.
GDELT narrative density spike: military-escalation — 250 articles in 24h (live verification 2026-05-17)
HAWK
May 18, 18:00
Form 4 cluster
8.67σ
13 insiders filing Form 4 on HAWK (HawkEye 360, Inc.) in 5 sessions
Strait of Hormuz
May 23, 10:38
pwin_chokepoint_stress
85
pwin_chokepoint_stress up at 85
HAWK
May 20, 18:00
Form 4 cluster
8.00σ
12 insiders filing Form 4 on HAWK (HawkEye 360, Inc.) in 5 sessions
Red Sea & Gulf of Aden
May 23, 11:06
pwin_war_risk_corridor
78
Houthi anti-shipping kinetic activity sustained; Container majors continue Cape of Good Hope routing
"The signal is not the loudest row in the table. The signal is that 4 sources are describing the same pressure while 7 monitored symbols sit outside the clean-state baseline."
Strait of Hormuz is the cleanest stress carrier in the current set.
pwin_chokepoint_stress up at 85 The row is included because it is fresh, real, and above the evidence floor; if it decays without confirmation, the research thesis decays with it.
Methodology
N1Reference set uses real signal rows from the last seven days, ranked by absolute z-score or normalized magnitude. Rows without a live reading are excluded.
N2VIDI context uses latest snapshots for 14 monitored symbols. Stress-state count includes state 1 and state 2 symbols only.
N3The note requires at least four evidence rows and eight VIDI snapshots. If either substrate is missing, the producer fails instead of publishing filler.
N4Persona callouts are constrained to named Arcane desk owners and are derived from the same evidence chain, not from separate copy blocks.
N5This research note is a regime interpretation, not an investment recommendation. It describes what would falsify the thesis before it describes positioning.