Nvidia's CEO spent the weekend apologising to the markets, Anthropic shipped two new models and bought a power grid's worth of gas turbines in the same news cycle, and a flagship Linux project rewrote its license because AI keeps eating its code. Here's the day.
Asahi Linux swaps MIT for GPL after AI tools launder its stolen code
Asahi Linux, the project that brings Linux to Apple Silicon, said on Saturday it is moving every one of its repositories from MIT to GPL v2 or later. The reason is grim and very 2026: people were pulling in-development code out of contributors' personal forks before it shipped, then feeding it through AI tools to rewrite it and scrub the origin. MIT made that easy, and the GPL makes it radioactive, because anyone who ships the derived work has to ship it under the same license. Expect more permissively-licensed projects to make the same call if this keeps working. Source: https://itsfoss.com/news/asahi-linux-drops-mit-for-gpl/
ChaCha encryption gets up to 99% faster on AMD Zen 5 in Linux 7.5
Phoronix reports that Martin Willi, a strongSwan developer, has queued a 16-block AVX-512 path for the kernel's ChaCha cipher, targeted at Linux 7.5. The old code stuck to narrower registers because Skylake-era Intel chips throttled hard when you touched the full 512-bit width. Modern chips don't, and on a Ryzen AI Max the new path runs up to 99% faster, with Zen 4 still gaining up to 39%. VPN tunnels and full-disk encryption just got faster for free. Source: https://www.phoronix.com/news/ChaCha-AVX-512-Linux-7.5
Nvidia shares slide as the board investigates Huang over the alleged spy affair
The BBC reports Nvidia closed down about 4% on Monday after the company said its board will run an independent investigation into CEO Jensen Huang. Huang has admitted an affair with a woman US intelligence believes was working for Beijing, tied to Megaspeed, the Singapore firm from last week's espionage reporting, and a memo Marco Rubio reportedly sent to officials last spring. He has apologised, promised to cooperate, and been asked to brief Congressional committees behind closed doors next week. Nvidia says it has no reason to believe its secrets were compromised. The market, clearly, is not so sure. Source: https://www.bbc.co.uk/news/articles/c6zxjx01vyleo
Anthropic ships Fable 5.2 and Mythos 5.2, cuts prices
Per Anthropic's own launch post, the new Fable 5.2 and Mythos 5.2 generate output up to 30% faster than the 5.1 pair and list at $8 per million input tokens and $40 per million output. Fable is open to everyone, while Mythos stays locked behind the company's cyber and bio verification programs. The tester claims are the usual flex: "otherwise infeasible" software migrations, working electronics designed from scratch, Fermat's Last Theorem formalised in 30 hours. Vendor numbers, but the price cut is real, and it lands in the same week the IPO chatter was getting trimmed. Source: https://www.anthropic.com/claude-fable-and-mythos-5-2
Anthropic drops $3.4B on gas turbines
TechCrunch, following the FT, reports Anthropic is buying Caterpillar generators capable of up to 1.7 gigawatts, a $3.4 billion deal for hardware it has historically avoided. The turbines were originally ordered by equipment supplier OnePWR, which couldn't close the financing, so Anthropic went straight to the manufacturer. This is a company that pays xAI $1.25 billion a month to rent compute deciding it would rather own the power plant. When turbine backlogs stretch into the early 2030s, renting watts stops being a strategy and starts being a queue. Source: https://techcrunch.com/2026/11/09/anthropic-joins-ais-natural-gas-frenzy-buys-3-4b-worth-of-turbines/
Trump says a top AI executive was told to "get ready to go to jail"
In a 60 Minutes interview aired Sunday, Trump claimed the Department of Justice told a top AI executive to "get ready to go to jail", and waved off the labs' safety warnings, saying the risk "is not nearly as real as they talk". No name, no charge, no timeline. It reads as strongman theatre for the cameras, but the people running the labs will have heard it, and they plan capital budgets in decades. Source: https://www.theverge.com/ai-artificial-intelligence/1029802/trump-60-minutes-ai
Meta ships Muse Ember, a tiny model built for agent grunt work
Meta introduced Muse Ember on Sunday: its smallest model yet, aimed at the unglamorous chores agents run all day, tool calls, routing, extraction, parsing long documents. Meta claims it matches Claude Haiku 5.5's best score at roughly a third of the cost per task and beats GPT-6 Luna at every setting, citing Artificial Analysis data. It's a bet that the money in agents is in cheap, fast subtasks rather than headline benchmarks. Honestly, a sensible one. Source: https://research.meta.ai/blog/muse-ember