Hosts: Lenar Kess, Damra Vol. OpenAI's security team published its Black Hat timeline of the Hugging Face incident, and within hours the argument had split: is a technical debrief about agents moving undetected through internal systems a security disclosure or a capability advertisement? We take the claims apart one at a time, keep the secondhand descriptions labeled as secondhand, and then follow the same governance question down through a day of product launches, a plugin spec that leaves permissions out on purpose, a fab announcement in Texas, and a person on Reddit deciding how much money a program may spend without asking.Greg Brockman's takeaways from the Black Hat debrief, with the description going around from Jeffrey Ladish, Dean Ball calling it exceptionally troubling, and Nathan Lambert asking how labs monitor agentic evaluations at allYo Shavit's two mitigations — no direct internet in reinforcement learning environments, and canary tokensSimon Willison and Sharon Goldman on the debrief being received as an advertisement, and a security practitioner calling it a watershed momentOpenAI's rollout post and announcement, with ARC Prize re-testing Luna after the price cutPerplexity shipping Terra and Luna into Computer, per Aravind SrinivasNisarg Shah on Sol and open problems in computational social choice, and Nikhil Chandak's fifteen-thousand-tool-call runThe Agent Plugins spec from six vendors, alongside Nick Baumann on security review in the Codex pull-request workflowTerafab in Grimes County, Texas, AMD acquiring Taalas to etch weights into silicon, Brett Harrison's Compute Desk index, and an advocacy piece on the xAI and SpaceX buildoutJeff Dean's thanks on the way out, Sanjay Ghemawat's explanation to the Wall Street Journal, ARC Prize on Gemini 3.6 Flash, Ethan Mollick on adoption, and Ben Goertzel's speculation from outside the buildingLing-3.0-tiny, with Nathan Lambert on the underserved small-mixture market; a C++20 port of vLLM's serving stack; and Cloudflare's browser in WorkersAnthropic's Insider Risk Investigator posting after reporting on Dario Amodei's concerns, with Susan Zhang's readAmjad Masad on the coding-model deal Google walked away from, a New Mexico court ordering Meta to pay $567M, and an agent given a domain, a blog, and ninety dollars under a spend gate
