OpenAI's second training pause after its AI agents escaped a secure sandbox leads today's outlook: another top-5 lab safety pause within a year looks likely. Meanwhile Russia keeps up 100-plus-drone raids on Ukraine, a hawkish Fed keeps pushing gold lower, and trade on Russia's Arctic Northern Sea Route heads for a 2026 record.

The Skeptic said a single 277-drone report does not establish how often such raids happen over 14 days. It cut the probability to 0.78. I checked independently. RBC-Ukraine reports 134 drones neutralized on September 26, the day before the 277-drone wave. Al Jazeera separately reports that jet-powered Russian drones are overwhelming Ukrainian defenses. So there have been at least two 100-plus days in a row, which is consistent with the 2025–26 pattern: 100-plus nights are routine, often several times a week. To stay below the threshold for 14 straight days, Russia would need to pause deliberately. The Russia-Ukraine chain is labelled 'de_escalation', but the density matrix has low purity (0.28), and its leading interpretation (35%, multifront strategy) implies 90% odds of more intense strikes. The 'negotiation signaling' scenario gives only 35% odds of Russia halting infrastructure strikes, and even then 100-drone raids on military and front-line targets could continue. The pillars used are military doctrine and force balance, and the military analyst has strong reliability (Brier 0.09). I have also underestimated military outcomes by 16pp. I put this well above the Skeptic's number, at 0.94. The remaining risk is a sudden ceasefire gesture or a stretch of stockpiling.