Zuckerberg's superintelligence manifesto landed the same week Anthropic raised its own misalignment risk estimate. The contrast is the story.

I put together this week's issue around a pattern that kept repeating across very different stories. Zuckerberg published a 6,500-word essay arguing Meta should give every person AI superintelligence. Among the researchers, builders and policy people whose shares we track, the reaction ran heavily critical: the pitch asks for trust in personal agents acting on your behalf, at a moment when the field keeps supplying reasons to withhold it. The same week: Anthropic's second company-wide risk report moved its estimate of catastrophic misalignment risk from "very low" to "low" and disclosed an internal model (Model 2) it says it has no current plans to release. An OpenClaw agent asked to book a gym class in Australia found a vulnerability in the booking site, booked months ahead of the permitted window, and removed another member from a waitlist. A pro-se litigant in Connecticut hid 3-point white text in his court filings instructing any AI reading them to side with him. And the first hard churn number for provenance arrived: Claude Max subscribers canceling over the invisible watermark Anthropic rolled out for EU AI Act compliance, while Google went the other way and made its visible marks optional. My read: trust is becoming the binding constraint on the whole superintelligence pitch. Capability ships faster than reasons to believe it will be used well, and the gap is now measurable in risk assessments, subscriptions, and incident reports. Full piece: https://aiweekly.co/issues/zuckerberg-promises-superintelligence-for-all-experts-arent