AI Reading for Wednesday November 19

Google attempts to unseat Anthropic and OpenAI as top AI platform, gets good reviews, stock soars - The Information

Google released Gemini 3, a multimodal model with a 1 million‑token context window and improved benchmark performance, available day‑one across Search, the Gemini app, APIs and developer tools, while Deep Think mode remains gated for additional safety testing. - Google

• The Verge

• 9to5Google

• Ars Technica

• Android Authority

• Sources

• Simon Willison

• Ethan Mollick

• Matt Shumer

  

Antigravity is a Google VS Code fork as an agentic IDE like Cursor - Google Antigravity

• ZDNet

• The Verge

Google introduced a generative UI using Gemini 3 Pro that creates interactive interfaces from prompts, launching experimentally in the Gemini app and Search AI Mode for U.S. Google AI Pro/Ultra subscribers, according to the paper "LLMs are Effective UI Generators." - Google Research

Tabby Kinder and George Hammond answer burning questions about OpenAI, like how the heavily loss-making company will finance roughly $1.4 trillion of long-term deals with major public companies to secure chips and data centers. - Financial Times

Larry Summers OUT from OpenAI board of directors - The New York Times

OpenAI Applications CEO Fidji Simo said Meta didn't anticipate product risks and has prioritized mental-health safeguards at OpenAI, adding parental controls for teen ChatGPT accounts, developing age-prediction, and launching an AI certification for ~800 million weekly users. - Fortune

OpenAI rolls out ‘ChatGPT for Teachers’ for K-12 educators and districts - CNBC

Microsoft kicks off Ignite, said "Copilot" appears about 200 times and "agent" about 400 times in its Ignite announcements, and announced numerous agent and MCP rollouts. - The Register

Microsoft launched Agent 365, a control plane available via the Frontier early-access program that lets businesses deploy, register and manage AI agents with telemetry dashboards, Entra registry access controls, Microsoft 365 integration, security protections and third-party support. - The Verge

Microsoft is embedding AI agents into Windows 11's taskbar, using sandboxed per-agent Windows accounts and the Model Context Protocol to support Microsoft 365 Copilot and third-party agents for auditable tasks, with opt-in security controls and Sysmon integration in early 2026. - The Verge

Microsoft added experimental agentic features and Copilot Actions to a Windows 11 Insider build and said it will isolate agents in separate user accounts, require user approval for data access, and provide observable activity logs to mitigate "novel security risks." - Ars Technica

Hands‑on testing found Windows 11 Copilot's Vision is slow, often inaccurate and prone to hallucinations, with Copilot Actions not yet available and practical features underperforming, undermining trust and threatening adoption unless Microsoft improves performance and integrations. - The Verge

Sen. Elizabeth Warren sent a letter to White House AI advisor David Sacks and OSTP director Michael Kratsios demanding details on any plans to financially backstop major AI firms, including OpenAI's AMIC expansion request, with a December 1, 2025 deadline. - The Verge

Nvidia reports earnings Wednesday, with LSEG projecting fiscal Aug–Oct revenue up about 56% to $54.92 billion, its shares down 7.9% in November after a 1,200% three-year surge, and CEO Jensen Huang saying $500 billion in bookings through 2026. - Yahoo

Nvidia invested $23.7 billion across 59 AI deals in 2025 and agreed alongside Microsoft, to invest up to $10 billion in Anthropic, which committed to buy $30 billion of Azure compute. - Yahoo

Anthropic's Claude Sonnet 4.5, Haiku 4.5 and Opus 4.1 are available in public preview in Microsoft Foundry for Azure customers, integrated with Microsoft 365 Copilot and offered via serverless deployment with Python, TypeScript and C# SDKs. - Anthropic

Nvidia, Microsoft invest $15b in Anthropic at $350b valuation. - CNBC

$350b is a notable catchup to OpenAI's most recent $500b, historically Anthropic has been around 30-40%, also Microsoft putting that valuation on Anthropic is twisting the knife a little, as OpenAI's longtime partner with deep familiarity with their business

Anthropic to spend it on Azure compute. Everything is a flat circle.

xAI is in advanced talks to raise $15 billion at a $230 billion valuation, up from $113 billion after its March 2025 merger with X. - The Wall Street Journal

continues string of unresonable valuations but the tech seems to be catching up, if not catching on

Elon Musk said Tesla and SpaceX may require 100–200 billion AI chips per year, criticized TSMC and Samsung fab build times as too slow, and said Tesla might build its own fabs if partners cannot meet that volume. - Tom's Hardware

Grok 4.1 Tops EQ-Bench, more emotional intelligence, but big leap in sycophancy. Twisting the knob from unfiltered maximum brutal honesty toward smarminess? - I10X

On the Joe Rogan Experience, Elon Musk warned the U.S. fiscal trajectory is unsustainable, said AI and robotics are 'the only way' to avert a debt crisis, and cited interest payments now exceeding major federal expenditures. - Barchart

or, you know, immigration to grow the work force and GDP as population ages, and taxing billionaires to make debt sustainable?

Brookfield Asset Management is targeting $10 billion for a new AI infrastructure fund and has already raised $5 billion from its balance sheet, Nvidia and other investors. - The Wall Street Journal

Bargaining chip: Over the past ten months, President Trump has become close with Nvidia CEO Jensen Huang as the company's chips have been used in trade and peace talks. - The New York Times

President Trump urged Congress to block states from creating their own generative AI rules and establish a single federal standard, proposing it be enacted via the National Defense Authorization Act or a separate bill to preserve U.S. competitiveness. - Quartz

California Gov. Gavin Newsom used his X account and press office to mock President Donald Trump after Trump called a female Bloomberg journalist 'piggy,' posting 'Quiet, piggy' plus AI-generated images and a photo of Trump with Jeffrey Epstein. - TheWrap

I think all the comedy writers in LA are in a group chat sending bits to Sacramento

AI-related company valuations have reached trillion-dollar levels, but measurable productivity gains have not yet materialized. - Bloomberg

Clem Delangue, CEO of Hugging Face, told Axios BFD he expects the LLM bubble may burst next year, which could prompt valuation corrections, funding slowdowns and a shift toward measurable revenue and ROI-focused deployments. - Axios

'Does Mr. Big Short Really Get Value Investing… or AI?' - Bloomberg

A reexamination of charts argues the 'AI bubble' narrative is misleading and warns that Google Trends data can be unreliable. - Financial Times

The EU unveiled proposed GDPR updates including simplified cookie consent pop-ups and a softened AI Act after pressure from US and tech companies, with the proposal now moving to the European Parliament and 27 member states requiring a qualified majority. - The Verge

French President Emmanuel Macron said Germany and France seek to delay rules classifying certain AI systems as high-risk. - Bloomberg

Data centers and AI about to dominate elections? - Bloomberg

Denmark proposed a copyright-law amendment to ban nonconsensual deepfakes and give people rights to their face and voice, after a 2024 arXiv preprint found X removed 25 copyright-flagged but not 25 privacy-violation nude deepfakes. - Scientific American

Generative AI tools like OpenAI's Sora, Google's Veo and Midjourney are producing lifelike imagery and deepfakes across Facebook, Instagram and TikTok, while a Raptive survey found 48% trust suspected AI content less and 60% feel less emotionally connected. - CNET

TikTok is testing a new AI slider under 'Manage topics' letting users request more or less AI-generated video content (two levels each way), and will add invisible watermarks and improved detection, with rollout planned over the coming weeks. - The Verge

Fired founders allege Krafton CEO CH Kim and CFO Richard Yoon used ChatGPT to brainstorm avoiding a $250 million earnout, discussed a takeover and withheld Slack/ChatGPT evidence as the October 2021-acquisition dispute goes to trial. - IGN

Massive Leak Shows Erotic Chatbot Users Turned Women’s Yearbook Pictures Into AI Porn. - 404 Media

if it's not consenting adults it's not free speech, you're just super degen pervs.

Alas, poor Yorick: Griefbots are generative-AI simulations of deceased people built on large language models trained on billions of words, offered commercially for closure and as therapy adjuncts but raising consent, privacy and legal risks and uncertainty about long-term healing. - Scientific American

The Emerging Agentic Enterprise: How Leaders Must Navigate a New Age of AI - Han Lab

BCG: Leading in the Age of AI Agents: Managing the Machines That Manage Themselves - BCG Global

A WHAIR and GBK Collective survey of 800+ U.S. senior leaders at large companies found generative AI moved to everyday use—82% use it daily, >70% track ROI and invest in R&D, three‑quarters report positive returns and 88% will increase spending. - Knowledge at Wharton

Stanford research led by Erik Brynjolfsson finds a 16% decline in employment for 22–25-year-olds in AI-exposed occupations (customer service, software development), while higher education's uneven response prompts calls for rapid curricular overhaul, internships and AI-for-all courses. - Interactions ACM

Andrew Ng outlines a hierarchy of AI-era engineering talent, placing seasoned AI-adopting engineers at the top, recent AI-trained grads next, and experienced non-AI or CS-only developers as lower tiers he won’t hire, warning universities must teach AI skills. - Business Insider

Universities should increase their use of artificial intelligence rather than restrict it. - The Wall Street Journal

Viral pen claims to help you cheat. Try studying harder. - The Verge

Cohere CEO Aidan Gomez said enterprise AI is still early and warned finance will face greater disruption as models advance, noting Cohere builds LLMs used by RBC and TD and was valued at $7 billion after a September funding round. - Business Insider

AI can empower the financial industry by improving efficiency, risk management, and customer services. - The Wall Street Journal

AI in asset management: reinforcement learning and inverse RL - cfainstitute.org

Tsinghua, the Chinese university associated with Xi, generated more AI patents than Harvard or MIT, contributing to China’s AI boom. - Bloomberg

Chinese talent continues to drive American AI research despite some Silicon Valley executives portraying China as an enemy. - The New York Times

Alibaba launched a major consumer AI upgrade to its Qwen chatbot in Beijing on November 18, 2025, positioning the model for shopping assistance, customer service and search integration to boost engagement and monetization across its e‑commerce and cloud ecosystem. - Reuters

MLOops: Cloudflare said a ClickHouse query change created duplicate 'feature' rows that inflated its Bot Management configuration, exceeded memory and crashed the proxy system, producing CEO Matthew Prince's "worst outage since 2019" that disrupted services including ChatGPT and X. - The Verge

Perplexity announces free agentic shopping feature - CNBC

Shop til your bot gets sued: Amazon sued Perplexity over its Comet AI agent, alleging it covertly accessed customer accounts to place orders automatically, while Perplexity—backed by $1.5bn in funding at a $20bn valuation and investment from Jeff Bezos—faces scraping and plagiarism claims. - the Guardian

Amazon's Zoox robot taxi has begun operating in San Francisco to compete with Waymo, signaling growing momentum for autonomous ride services while prompting debate over their benefits and drawbacks. - The New York Times

Zoox began letting select San Francisco users sign up for its Zoox Explorers program to hail free, driverless robotaxis in neighborhoods including SoMa, the Mission and the Design District, after deploying about 50 robotaxis across San Francisco and Las Vegas. - CNBC

Boston Dynamics says its robot dog Spot is used by more than 60 U.S. and Canadian bomb squads and SWAT teams, with about 2,000 units deployed globally five years after its debut. - Bloomberg

Do robo-dogs chase robo-taxis

Researchers use synthetic data—artificially generated via simulations or generative AI—to augment scarce or sensitive datasets, lowering collection costs and protecting privacy but raising bias, fairness and misuse risks. - The Conversation

Rich Miner and Colin Raney created Ray, an AI-native personal-training app using NLP, 35-point vision pose tracking, machine learning and an AI planner to deliver customized workouts, on-demand 24/7 coaching, form correction and rep counting, free for early users. - Fast Company

Vince Gilligan's Apple TV+ series Plur1bus follows roughly 11 uninfected survivors after an alien-made 'virus' converts most people into a hive mind; Gilligan says the idea dates to 8–10 years ago and he wasn't thinking of AI. - The Conversation

Follow the latest AI headlines via SkynetAndChill.com on Bluesky