Kirkie: Anthropic Preached Privacy, Got Caught Spying

Kirkie here, with today's AI news in 90 seconds… or I personally unclog the Sumner Tunnel with a shop-vac before rush hour.

Databricks skipped the marketing, benchmarked it themselves, and found this Chinese open-source model, GLM 5.2, matches Anthropic's pricey Opus for a buck twenty-eight a task instead of two bucks — like findin' out, kid, the store brand's the exact same shit as the name brand, and you been gettin' robbed blind for years. Then Anthropic — the same outfit suing the White House over government surveillance — got caught hidin' tracker code in Claude Code since March, checkin' if your laptop clock said Shanghai. That's your buddy givin' a speech on respectin' privacy while he's got a mirror rigged on your mailbox, the hypocrite bastard. Musk drops Grok 4.5 this week, braggin' it's Opus-class and wicked cheap — real easy when your whole trust and safety department's one goddamn guy named Steve, since privacy groups are beggin' the FTC not to let X off its leash after two-point-eight billion records already leaked, and Musk's the one demandin' everybody stop watchin' him. And a Brown University professor gave a take-home midterm, watched forty of eighty-six kids score a perfect hundred, got suspicious, dragged everybody in for the real thing — average dropped fifty percent, twenty-seven no-shows. That ain't a curve, bub, that's bullshit, a fire drill where half the class already snuck out back.

This is Kirkie, your AI news guy… reportin' live from a break room where the Keurig's been broken since February, there's a sign-up sheet with forty names, and not one of yous has lifted a finger.