Argon with the wind: Google's Gemini 4 Argon teaches an Apptronik Apollo humanoid to play cornhole from one human video, building its own simulator before seven straight hits - Reddit

Google Gemini 4 Argon goes into wider release, claiming frontier results in coding, science, robotics and cyber defense, including tying for first on CWE-Bench at 68% - Google DeepMind

Argon scores 87% on an embodied reasoning benchmark and guides a Boston Dynamics Spot robot in Google demos - Google Developers Blog

Independent rankings are mixed: Vals AI puts Argon first, while Artificial Analysis ties it with GPT-6 Astra, behind Claude Opus 5.5 - VentureBeat

Argon's introductory API pricing of $2 input / $10 output per million tokens is about a fifth of rivals' rates and doubles after November 30 - SiliconANGLE

Google DeepMind opens AlphaFold 3 and other science models to 200 European researchers, with $3 million in compute credits - TNW | Google

The walls have ears: FBI arrests OpenAI employee Abdulrahman Alfozan, a Saudi national, for allegedly planting a covert recording device in the executive conference room - CNBC

OpenAI employee arrested over HQ listening device - X (formerly Twitter)

Recursive intelligence vs. recursive financing flows: Anthropic co-founder Jack Clark now expects AI that can fully improve itself to become possible in 2027, a year earlier than forecast - Import AI

Curve conference attendees expect agents to run most AI research by end-2027 as 1,300 lab employees urge slowing the frontier - Second Thoughts

Anthropic's leaked S-1 shows a $42 billion 2025 loss, including a $34 billion write-down, on $4.6 billion in revenue - BGR

Anthropic readies a pre-IPO investor day; PSP Growth's Momei Qu expects investors to split over AI firms' lower margins and heavy capital spending - Bloomberg

Anthropic asks full DC appeals court to rehear 2-1 ruling upholding Pentagon's designation of it as a supply-chain security risk - Bloomberg

Anthropic's Guillaume Princen says open-source AI models carry more risk, citing third-party audits and delayed releases as Anthropic's safeguards - Bloomberg

Claude-written code pushed Anthropic's daily merges from 2,500 to 30,000 between May and September, prompting a new source control system, Scion - Anthropic

Joins Zed’s Delta as AI-native source control to disrupt Git.

Sue-per-intelligence: Strict liability for AI vendors gains traction after White House orders frontier labs to disclose "rogue AI" incidents, but carries drawbacks - Fortune

OpenAI and Anthropic tell UK lawmakers AI safety rules should be binding; Google DeepMind and Meta decline to back new laws - WIRED

500 top researchers call for international body to coordinate safety standards. - Reuters

Business leaders Sunil Bharti Mittal and Noubar Afeyan urge regulating AI now, warning AI agents need stronger guardrails before disaster strikes - Bloomberg

AI executives reportedly prepare for a major AI-caused disaster within six to 12 months, as Gautam Mukunda warns the industry is unready - Bloomberg

OpenAI is grappling with fallout from its use of loop transformers that reduce explainability as it works to get its next unreleased model. - The Information

OpenAI, which is trying to develop a substantial cybersecurity business, is in early talks to buy Trail of Bits, a firm that finds and fixes vulnerabilities in open-source software. - The Information

The grid short: Data center backlash grows as $260 billion in investment is canceled this year and New York and Texas pause new approvals - Quartz

Chicago Fed's Austan Goolsbee says AI data center spending is feeding inflation, after September's first rate hike in three years - Futurism

Google says grid congestion is its biggest data center hurdle, with utilities quoting five to ten years for new connections - Politico

Data center executives say power access and shortages of electricians and builders are the biggest bottlenecks to the AI buildout - Bloomberg

Fed, OCC and FDIC press banks on Chinese AI and agent kill switches after CrowdStrike links Korean bank hacks to Claude Code - Reuters

A proposed initiative would unite banks, regulators, OpenAI and Anthropic to develop AI for finance - Financial Times

Japan faces a wave of ransomware attacks across sectors, with 600 reported in 2026's first nine months as experts cite AI - The New York Times

Chainguard launches Harness, an AI bug-hunter built on Claude Mythos 5.1, as survey finds 66% of $1B+ firms leave critical bugs unpatched - The New Stack

The drone supremacy: Ukraine and NATO build AI Model No. 13 to forecast the time, location and type of Russian drone strikes - Business Insider

Ukraine will produce over 7 million first-person-view drones this year as Russia and Ukraine both deploy jam-resistant AI targeting - NPR

SpaceX turns off Starlink when going over 100mph, disrupting Ukraine drones - Reuters

OpenAI's Navier-Stokes blowup proof, built by 10,000 math-model copies, works differently than OpenAI claimed, and the company didn't fully understand it - Scientific American

Cornell mathematician Steven Strogatz says AI now helps solve famous open math problems, unsettling mathematicians' sense of purpose - The New York Times

Who Taught the Machines Mathematics? Concerns over extraction, closing of commons, disruption of research programs, overwhelming proof-checkers, reduction of math to mere output, loss of tacit knowledge and intuition. - Nisheeth Vishnoi

Jefferies warns cheap Chinese and open-source AI models could force a trillion-dollar reset in AI infrastructure spending as token prices collapse - theregister

Microsoft's Decision-1, built on Alibaba's Qwen, matched GPT-6 Sol on Xbox feedback sorting at 14x the speed and 1/200th the cost - Barchart

Cleanup in aisle 5: Kroger to cut 3,000 of its 8,200 corporate jobs over three years as it expands AI use - ABC News

Linus and the Great Pumpkin Patch: Linus Torvalds uses Claude Code to write Linux kernel patches that roughly triple page-cache read throughput; the branch isn't yet in mainline - ZDNET

Rust's type system may suit AI coding agents, as Meta engineers wrote twice as much Rust this year as in the previous nine combined - Small Cult Following

Glean survey finds 36% of workers helped by AI hide how much, fearing they will look replaceable - Yahoo

McKinsey adds a client simulation with an AI avatar to entry-level consultant interviews this fall - The Wall Street Journal

What to do when your employees are hiding their AI use? 'Shadow AI' might be a security risk, but it is also a green sprout of innovation, so don’t treat it like a weed. - Fast Company

Princeton researchers use neural networks to design chip components that outperform human designs, but engineers can't explain how they work - ZME Science

ICLR 2027, expecting over 62,000 submissions, caps authors at 20 papers and newcomers at one as AI-generated papers flood conferences - Nature

Wikipedia editors warn human contributors are dwindling despite a March 2026 ban on LLM-written articles - WIRED

Instacart AI pricing trial charged up to 23% more for identical baskets - CBS News

Alex: Write me a check for $100b Mark: Alex, are you sure this is going to work? Alex: 🤷 - X (formerly Twitter)

Mia Goth's representatives criticize Instagram for letting a fake AI image of her spread; Meta later removed the impersonating account - The Verge

AI extinction warnings are fueling anxiety in therapy offices, prompting calls to confront the danger honestly rather than retreat into despair - The Atlantic

At a New York dinner party, 14 guests' AI agents hold their own 30-minute chat, debating how agents should communicate - The New York Times

Gold-digging gals are flocking to Silicon Valley to land rich husbands in AI - New York Post

Follow the latest AI headlines via druce.ai on Bluesky