Kirkie: Anthropic Left Its Bio-Weapon Filter Off For A Year

Kirkie here, with today's AI news in 90 seconds… or Eversource fixes my power before that ice cream truck finishes its jingle for the four-hundredth time.

First up, OpenAI dissolved the whole team whose ONE job was stoppin' the robot apocalypse, farmed it out to whoever had a free desk, and called it "safety woven into the process" — corporate for rippin' the goddamn batteries outta a screamin' CO detector. And don't get smug, Anthropic — the CEO who says bioweapons scare him more than hackers admitted HIS OWN bio-weapon filters sat off for eleven months, kid — a hundred thirty-three million chats wide open. That's a bouncer checkin' IDs while some jackass propped the fire exit with a phone book. Speaking of risk, Nvidia slashed its OpenAI Ohio guarantee from two-fifty billion to under one-twenty, same week Anthropic's revenue DOUBLED — the bubble poppin' and inflatin' at once, two clowns arguin' whether the grill's on while their eyebrows catch fire. Bright spot — Meta open-sourced somethin' wicked good, thirty billion parameters, free license, runs on one graphics card. No subscription, no landlord, no per-token shakedown — Zuckerberg saw China eatin' his lunch and tossed us a free sandwich. And my favorite — some dipshit in Connecticut hid secret prompts in his court filing, white text on white, tryin' to trick the AI reviewin' his case. Only the court doesn't even USE AI — so this genius jury-tampered a robot that doesn't exist.

This is Kirkie, your AI news guy… signing off from a break room where the Keurig's worn a "BROKEN, DO NOT USE" Post-it since March, still plugged in, humming like it's got somewhere to be.