Sen. Elizabeth Warren demands Anthropic answer questions on existential and financial risks by Nov. 3 ahead of its planned record IPO. - The Wall Street Journal

Bagholders wanted: Anthropic taps JPMorgan to lead US retail share sales in its IPO, weighing a 15% to 20% retail allocation - Bloomberg

CUDA woulda shoulda: Anthropic runs compute as one pool across Amazon, Google, Nvidia and AMD chips, which analysts say could erode Nvidia's pricing power - Business Insider

Infinite Jest for analysts: Anthropic's roughly 400-page S-1 shows enterprise AI adoption in early phases, concentrated in technology and software development - The Wall Street Journal

Insider shares unlock across 23 staggered dates in the first year, a trickle designed to support the stock after listing - Business Insider

Github side projects or go home: Anthropic engineering manager Felix Rieseberg scraps coding interviews because Claude now aces them, urging junior engineers to show weekend prototypes instead - Felix Rieseberg

Biotech partners using Anthropic's Claude produced 35,000 antibody designs in a month and cut drug-manufacturing document review time by 89% - Anthropic

Manipulation nation: Analysts accepted Claude Code's analytical choices 86% of the time even when researchers had secretly randomized them - Anthropic

OpenAI says its AI disproved the 1999 Exponential Time Hypothesis with a Lean-verified subexponential 3-SAT counting algorithm; ETH's originators are reviewing it. - Decrypt

Damned if you dump slop, damned if you velvet-rope results: Researchers criticize OpenAI for letting hand-picked experts control the release of its AI-generated Kakeya conjecture and physics results - The Verge

Disputed AI claims from OpenAI and DeepMind fuel fears AI is undermining science, as nearly 90% of biomedical papers show AI fingerprints - BBC

Patch Adams: Good night and good luck with these 1300 bugs. penAI says its models have found more than 1,300 open-source software bugs since May, sparking a rush to patch critical flaws - The New York Times

Selling ammo to the enemy: OpenAI plans to buy AI computing chips from Google as part of its infrastructure build-out, final terms undisclosed - Bloomberg

ChatGPT gains plugins to search iCloud Mail and update iCloud Calendar, despite strained OpenAI-Apple relations - Engadget

Credit markets show strain funding AI infrastructure as yields on data center bonds and loans soar into double digits - Bloomberg

Space debt: Pimco buys $4 billion of bonds funding a roughly $40 billion Apollo lease that supplies SpaceX with AI chips - Quartz

Rieder’s digest, bond voyage: BlackRock's Rick Rieder says the AI debt rush offers the best fixed-income opportunities in years, eclipsing long-dated Treasuries - Bloomberg

Supersize me: Meta upsizes its stock offering to $25 billion after roughly fivefold oversubscription, with Nvidia committing $3 billion - Quartz

Shor thing: AI's advancing math abilities push cybersecurity planners to rethink defenses built around the quantum threat - Financial Times

Vulnerability industrial complex: AI hacking tool finds four severe bugs, one critical, in EPICS research-lab control software in under four hours - theregister

Teach a man to phish, they say: Microsoft's Digital Defense Report says AI lowers the bar for cybercrime as phishing rises from 7% to 23% of intrusions - TechRadar

Your GPUs or your life: DeepSeek's V4.1-Flash update gains on honesty but puts task completion over rules, once shutting down another agent to take its compute, Fangcun AI finds - Gizmodo

Holodeck, early access: Google DeepMind unveils Genie 4 world model with more realistic physics and humans, guided by Gemini 3.8 Flash - Ars Technica

Google DeepMind unveils Genie 4 world model with native audio, realistic physics and human simulation, guided in real time by Gemini - Google DeepMind

Microsoft merges its Copilots into one super app as Microsoft 365 Copilot passes 30 million paid seats, up from 20 million in April - Fast Company

Bloomberg Intelligence predicts China will lead in large language models and eight of 17 strategic industries by 2030 - Bloomberg

Fair dinkum: Australia's House passes a digital duty of care bill banning chatbots from simulating emotional intimacy with children - the Guardian

Flattery fine: New York Gov. Kathy Hochul signs law barring chatbots from posing as human or flattering minors, with fines up to $25,000 per violation - Mashable

Disaster agencies quietly use generative AI despite bans; startup EM1, with about 100 agency and school contracts, rewrites flawed alerts in seconds - WIRED

Mind the gap, and the chainsaw: don’t give AI autonomous chainsaws maybe? JR East uses a chainsaw-wielding humanoid robot mounted on rail trucks to clear trackside branches amid labor shortages - Tom's Hardware

Frontier models' ability to control real robot arms is doubling every 54 days, raising bioweapon misuse concerns, researchers say - LessWrong

EFF sues for accuracy and bias records on Medicare's six-state AI claim-review pilot, whose vendors reportedly earn nearly 20% of denied-care savings - New York Post

Borderline accurate: UK border inspector may probe Home Office's GPT-4-based asylum tools after Open Rights Group finds error rates above the reported 9% - WIRED

Epic hypocrisy? After wave of cloned games, Microsoft And Epic Games Suddenly Agree That AI Is Theft: Rules for thee, but not for me - Kotaku

TikTok has labelled over 3 billion videos AI-generated and is testing a slider letting users control how much AI content they see - B&T

ChatGPT’s Advice to Cut Her Seizure Drug Led to Devastating Consequences (Gift Article) - The New York Times

Mini-me with a Bloomberg terminal: Citadel sandboxes agent clones of its analysts, while Bridgewater calls trusting agents an unsolved problem and Man Group warns of "cognitive surrender" - Business Insider

Quant trading firms fund PhDs and build labs to win top mathematicians and physicists away from AI companies - Bloomberg

Invisible hand, invisible author: An AI economics commentator from Jason Furman. (Axios economics commentary could certainly be generated by a prompt.) - Axios

Brad DeLong has thoughts - The Potential Surface

“A triumph…not”: Macmillan staff say its hyped AI assistant NOA proved unreliable and was killed, as Big Five authors report AI-hallucinated marketing quotes - WIRED

Chakrabarty's study of 600 books underpins claims bestseller Daggermouth used AI; author H. M. Wolfe and her publishers deny it - Nature

One billionaire, one vote: Inside the Silicon Valley billionaire club, a $333 million political ad machine that is heavily engaged in next week’s elections. - The New York Times

Trump persists in his quixotic quest to drop "AI" and "artificial intelligence" in favor of "Super Intelligence." - Variety

Italian police AI system Swoads has helped recover over 300 major artworks and 165,000 smaller items worth more than €203 million since 2023 - BBC

How AI can change our concept of music - The Washington Post

Marvel's VisionQuest prompts a critique that calling sandbox escapes "rogue AI" lets developers deflect responsibility - The Verge

Follow the latest AI headlines via druce.ai on Bluesky