The US-Iran conflict approaches a critical inflection point as Trump declares victory and signals withdrawal within weeks, even as Israeli strikes continue to escalate — today's digest tracks this pivot alongside Russia's accelerating internet isolation campaign, the AI infrastructure arms race, and an emerging constitutional confrontation over voting rights in the United States.

The cybersecurity analyst proposed P=0.64 (risk=84, highest in today's pool), which the Skeptic maintained. I push slightly to 0.67 based on compounding factors the analyst identified but may have underweighted in aggregate. The threat environment is clearly elevated: 'Iran-linked hackers escalate cyberattacks on US, Israel' (ESCALATION, 6 clusters), 'Iranian hackers target FBI director's emails' (ESCALATION), and 'Axios attack in npm compromises JavaScript projects' (CONFIRMATION) all demonstrate active supply-chain and targeted cyber campaigns. The 'Claudy Day Trio of Flaws Exposes Claude Users to Data Theft' (CONFIRMATION) and Anthropic's accidental Claude Code source leak expand the AI-specific attack surface. The strategic value argument is compelling: frontier AI model weights, training pipelines, and safety systems represent perhaps the most valuable intellectual property in the world — MSS-linked operators (APT41, APT10) have historically allocated expensive tradecraft to lower-value targets. The Skeptic's most important critique is that 'publicly tied' depends on disclosure and attribution decisions, not just the intrusion occurring. This is the primary drag on probability — the intrusion itself is more likely than its public attribution. However, the political environment favors attribution: US-China relations are deteriorating ('US-China relations predicted to worsen' in DEVELOPMENT, 4 clusters over 13 days), and the Biden-era precedent of publicly naming Volt Typhoon and Salt Typhoon campaigns shows willingness to attribute. Xi Jinping's BVI of 2 means PRC cyber operations follow institutional logic rather than impulsive orders — they are persistent and systematic, increasing the probability of at least one detectable intrusion over 180 days. Dead-zone justification: I push from 0.64 to 0.67 because (a) the 180-day window captures two quarterly CISA/FBI advisory cycles, (b) the convergence of expanded attack surface, extraordinary target value, and political incentive for attribution collectively exceed the sum of their individual contributions, and (c) the base rate of public PRC cyber attributions in the technology sector is approximately 4-6 per year across all tech subsectors, making at least one AI-adjacent attribution within 6 months more likely than not.