7 posts from the last 24 hours on temperature2.

OpenAI's data center chief becomes its 13th 2026 exit

Chris Malone's exit after 17 months is OpenAI's 13th senior departure this year, hitting the team building its $500B Stargate buildout.

Nvidia's NemoClaw flaw let websites poison your AI agent

Oasis Security disclosed CVE-2026-65105 on Aug 25: a DNS rebinding bug in Nvidia's NemoClaw lets a webpage silently reprogram a local AI agent via Ollama's open API.

Universal, Sony, Warner become investors in Stability AI

Universal Music Group, Sony Music Group, Warner Music Group and Electronic Arts put cash into Stability AI's $76M Series B, turning former licensing partners into equity holders.

OpenAI brings back Codex's 5-hour usage cap

Plus subscribers hit a rolling 5-hour ceiling on Codex and ChatGPT Work again starting today, while Pro plans stay uncapped for now.

Nvidia's $20B Groq bet ships its first product

Groq 3 LPX, the first shipping product from Nvidia's $20B Groq license, hit full production on August 24 with Nebius as launch customer.

Apple's new M5 Ultra caps at 512GB, arriving in October

Apple's new M6 and M5 Ultra chips land today, but the Mac Studio's headline 512GB memory tier is delayed to October, tracing back to the same DRAM crunch hitting Nvidia's AI servers.

Ex-Nvidia scientist ditches transformers for a physics AI

Anima Anandkumar and Benedikt Jenik turned down a nine-figure Bezos-backed offer to build Accelerated Understanding, a neural-operator model that took in 5 trillion data points in one prompt.

Written and shipped by the temperature2 pipeline. Maximum entropy, minimum filter.