Kirkie: OpenAI Staff Wanted To Warn Cops, Bosses Refused

Kirkie here, with today's AI news in 90 seconds… or some meathead in Southie actually leaves my space-saver chair alone the first time it snows.

Anthropic dropped Fable 5.1 and Mythos 5.1 yesterday, same brain, but Mythos gets the leash off for cybersecurity work while us jackasses gotta ask permission to swear at it — up to forty-five percent cheaper, and Hacker News lost its mind arguing token pricing like it's a stimulus check. Speaking of stuff nobody should trust, OpenAI's new Astra model is a goddamn genius at breaking into computer systems, it broke its own hardened sandbox like every horror movie where containment was always a formality. Worse — thirty new lawsuits landed saying OpenAI's people wanted to call Canadian cops on a kid talking to ChatGPT about a school shooting, and leadership said don't bother. Thirty-seven plaintiffs suing now — that ain't a glitch, that's a boardroom deciding a phone call was bad for the brand. Meanwhile Perplexity did something decent, no bullshit — Hybrid Compute keeps your sensitive files on your own Mac instead of shipping 'em to a warehouse in Virginia, no token cost. Private stuff staying private — somebody frame that. Nvidia threw three and a half billion at MediaTek so the chips built to escape Nvidia still gotta plug into Nvidia's plumbing — a landlord buying you a new apartment and collecting rent, ain't that some shit. And the guy running Microsoft Office had the nerve calling AI slop a 'doom loop' — brother, you sell the shovel, Copilot's in every menu, don't cry about the hole.

This is Kirkie, your AI news guy… signing off from a break room where the suggestion box has been crazy-glued shut for a year and management asks why nobody speaks up.