Weekly Claw

Anchor (Andy opens manually): 00:34–01:50 “every other launch iffy at best… they made it cheaper” · 03:29–04:10 “four major model drops in the last 2 days” · 06:30–07:30 the $2.94-vs-$20 beat · 39:00–40:14 verdict on Sol.

Fallback if the anchor fails: Theo t3.gg, “Elon promised this one would be good…” (2026-09-22, 27:53). Manual open only.

Caveats: all Grid A benchmark figures are vendor-reported (Anthropic, OpenAI, SpaceXAI, Xiaomi), and each vendor's competitor comparisons rely on publicly reported secondhand scores — OpenAI's own footnote says so. Sol/Luna absolute pricing ($2/$10) is per Yahoo Finance; OpenAI's page states the 50% cut, not absolutes. Grok 4.7's “2.1T parameters” appears in press coverage, not the launch post. MiMo's AA Index 46 is Xiaomi's claim; Step 5's weights are a promise dated October 15; Laya's latency figures are Convai-reported with no independent reproduction. The Muse concierge story is a single-source Reuters wire (internal posts; syndication is identical copy); the 95–98% internal success figure and incident detail come from Reuters' full read and are kept to what Reuters states. The Medicare breach: OpenAI statements are company responses, not findings; Transluce itself describes the extent as minor with no evidence of exploitation. CLOSEDQUORUM has no confirmed in-the-wild deployment and its public build is deliberately inert. The ART discovery is Anthropic-reported, preprint not peer-reviewed, function unknown. Theo's $2.94-vs-$20 cost beat is his own math on published prices. All video is manually opened; nothing plays on its own anywhere in this deck (no autoplay).