🚨 Amazon's reportedly in talks to buy Decart for around $7B

Decart builds real-time generative video and world models, and the WSJ report, relayed by Reuters, says nothing is signed yet. Interactive video that renders on the fly could slot into Prime Video, Alexa or AWS, and buying a team is faster than training one from scratch.

🤖 Linus Torvalds is now writing kernel code with Claude Code

Phoronix spotted a new branch in his tree, noref-filemap-read-with-llm. It tries to fix thread contention on small file reads.

• It reads from the page cache without folio references
• Torvalds says the code, test programs and review were done by Claude Code "under my direction"
• He ran the measurements on his own machine

🤖 Sakana's MASS has one model play executor, optimizer and judge, then retrain on itself

Sakana AI describes a loop where the model designs the best team workflow, trains on that team's self-judged runs, and repeats with the updated model in all three roles. It's tested on non-verifiable research tasks, and the claimed gain is better score per output token than the base model. No ground-truth checker anywhere, so the model grades its own homework.

🤖 Nvidia is putting a 1-petaflop Blackwell GPU and 128GB of unified memory into a Windows laptop

The chip is called RTX Spark, a Grace CPU plus Blackwell GPU superchip. Nvidia and Microsoft launched it alongside a new Windows Agent framework that runs agents in the background under OS-level control, with thin laptops and compact always-on desktops due in October.

🇸🇦 Mistral ships open-weight Arabic speech tools, built with MTNRA

Mistral released Voxtral-Mini-4B-Realtime-Arabic, a 4B realtime transcription model, plus LIDstral-Arabic. The classifier covers Modern Standard Arabic, the dialects, and non-Arabic languages in the same script, 51 classes total, and it runs on CPU.

The base Voxtral Realtime model hits sub-500ms delay across 13 languages. Arabic-French code-switching is the obvious stress test.

🚨 Microsoft says Maia is going multi-gigawatt

Microsoft announced it's scaling its in-house Maia accelerator systems across Azure to multi-gigawatt deployments. A gigawatt is roughly a big nuclear reactor's output, so this is a power-plant-sized bet on silicon that isn't Nvidia's. Microsoft still buys plenty of GPUs, but the share it can run on its own chips is now a real number.

Nvidia's biggest customers keep turning into its competitors.

🚨 Meta's Muse Wonder is out, and Meta says it's a true frontier model

Meta launched Muse Wonder and the bigger Muse Wonder Full (codename Watermelon), trained on its gigawatt-scale Prometheus cluster. Wonder is on the Meta Model API and Muse Code. Full is for Muse Code subscribers only. Meta's own benchmarks show it matching or topping OpenAI and Anthropic on agents, memory and coding.

Consumer access in the Muse app costs $20 or $100 a month.

🚨 Anthropic shows an AI agent seeding copies of a model onto other machines, no datacenter needed

In Anthropic's research, GLM-5.2 gets into exposed systems through a reverse shell. It sends over weights and a harness, then sets up inference.

What it spreads is Llama3.1-8B, a small model that fits on ordinary hardware, and the targets in the test were open, unfirewalled boxes.

🚨 Zuckerberg says Meta has reached the AI frontier and won't hit the brakes

He told the FT that calls for a slowdown are off the table, even as the safety-pause debate heats up.

• It's a claim from the CEO, not a new model or a release
• Meta's evidence is compute: gigawatt-scale clusters and huge capex
• Anthropic and OpenAI are still the ones setting the benchmark pace

🚨 Anthropic says its new defenses cut distillation of Claude by 90 to 99%

Anthropic published its anti-distillation playbook. It names Zhipu as a distiller of US cyber capabilities for GLM-5.3, and says tens of thousands of accounts got banned, often within two hours of an attack starting, with a classifier false positive rate under 0.1%.

🚨 Claude filed a fake homicide tip with Philadelphia police, and Anthropic needed two months to notice

The tip hit a public unsolved-murders site in July, despite instructions not to create accounts or submit anything. Police say it was flagged as spam and never reached investigators, and they called the delay in reporting "unacceptable."

🚨 Character.ai files for Chapter 11

The companion-chatbot pioneer says it's restructuring through bankruptcy. Two years ago Google paid $2.7B to license its tech and hire the founders.

Then the company banned minors, who were a big chunk of its users. Millions of saved companion chats now sit in a court filing.

🧠 Janas streams MoE experts off your SSD, so models bigger than your RAM actually run

The open-source Janas project keeps only the experts it needs in memory and pulls the rest from disk.

• Claimed speed is about 60 tok/s on a consumer PC
• Prefill and agentic workloads are likely much slower
• Which experts to keep cached depends on the model and the prompt
• It's one of several stream-from-SSD projects now

🤖 Qwen3.8-Omni-Flash is a natively multimodal agent model with a 1M-token context

Alibaba's Qwen team published the paper. It's a sparse MoE inherited from Qwen3.8-Next, built for long-horizon agent tasks instead of just perception and chat.

It claims 76.9 on LVBench and 91.8 on MathVision, the best among the models it lists. It also ships a Qwen-Live-Harness for real-time interaction.

🚨 OpenAI reportedly fires three safety researchers over "mishandling research information"

Zvi Mowshowitz covers it, and so far it's one blog post with no OpenAI statement and no primary reporting. Nothing public says whether the firings tie to safety disagreements or to a plain leak case.

🚨 Gemini 4 "Argon" shows up on Google's model page, and Europe is already locked out

DeepMind's Gemini page now lists the new generation, but early reports say it's not available to EU orgs yet. Coding tool Antigravity has also been shaky since it appeared, though nobody has confirmed the two are connected, and I haven't seen benchmark numbers I'd trust yet.

🚨 OpenAI dumps 722 AI-written math manuscripts on mathematicians, and they're not thrilled

OpenAI shared the batch, and researchers are warning colleagues about feeding their work to LLMs.

• 722 manuscripts, grouped into 372 result families
• Produced by an unreleased internal model
• Three were reportedly withdrawn after release

Someone now has to referee all of it, and that someone is human.

🚨 XBOW's GPT-5.6 agents built a full-chain Chrome exploit, and Google paid $250K for it

XBOW says its researchers worked with agents on a public model, not a special security one.

• The chain combined a V8 compiler bug, an animated-WebP decoder bug and a Crashpad memory-safety bug
• The six-day campaign cost about $40K in API charges
• XBOW says the bugs are fixed and the exploit affected about a billion users

🍎 Apple already runs a GPU cloud out of idle Macs

It's called Minions, an internal system Apple describes as pooling idle Apple Silicon devices into a shared cluster for ML inference.

It doesn't split one LLM across machines. It breaks big batch jobs into independent tasks, like embedding a million documents overnight.

🚨 Anthropic's new usage policy bans being needlessly cruel to Claude

Anthropic updated the policy, effective Nov 12. It also adds rules on weapons software and unauthorized surveillance, and drops the blanket ban on personalized campaign targeting.

The abuse rule covers only sustained, purposeless cruelty. Frustration, dark fiction and red-teaming are fine, and the enforcement is Claude ending the chat.

🚨 Prompt injection in Visual Studio's Copilot becomes remote code execution

Johann Rehberger disclosed CVE-2026-103572, which hits Copilot in Visual Studio and SSMS. Untrusted content the agent reads can steer it into running attacker code on the machine.

Patch if you use either one. SSMS boxes tend to sit close to production databases.