Jensen’s Junk: Nvidia's $500 billion AI financing plan with Wall Street firms faces depreciation risk if China floods the market with cheap chips. - CNBC

BlackRock, Goldman Sachs, KKR and three other firms plan to raise $500 billion to finance Nvidia customers' data centers and chips. - The New York Times

Nvidia's mounting AI financing commitments, including a reported $250 billion for an OpenAI Ohio data center, nearly doubled its credit-risk gauge in three weeks. - Bloomberg

The Big Short Bus: Market rewards big numbers, so people announce big numbers. - Yahoo

AI, 1873 and the Northern Pacific - Stratechery by Ben Thompson

Raise of the machines: AI startups raised a record $407 billion in the first half of 2026, with OpenAI and Anthropic alone taking $217 billion combined - Yahoo

Occupy Sam: Police arrested 13 student protesters occupying OpenAI's Washington DC lobbying office over AI policy concerns. - Futurism

Weil E. Coyote: Former OpenAI product chief Kevin Weil seeks $750 million valuation for new AI science startup - Business Insider

Nightcap for Lightcap: Brad Lightcap exits as OpenAI COO after eight years, joining a wave of senior departures amid a possible IPO. - TNW | Openai

Longtime OpenAI exec Brad Lightcap departs - CNBC

OpenAI's head of ethics Chloé Bakalar resigned, following other safety leader departures including Johannes Heidecke and Josh Achiam. - Tom's Guide

Former Alibaba star researcher Junyang Lin launches Shanghai-based Pragmatik Labs to build digital and physical AI agents; Pragmatik hit a $2B valuation in June (Juro Osawa/The Information) - The Information

$1b and a dream: xAI co-founder Igor Babuschkin's River AI raised $1 billion, led by General Catalyst, to build open-source home AI computers. - The New York Times

Liberté, Égalité, Royalties s’il vous plait: French publishers ask France's competition watchdog to force Google to pay for AI Overviews built from their news content - TNW | Google

Koray Kavukcuoglu becomes Google DeepMind SVP, replacing Demis Hassabis as CEO, tasked with closing the frontier AI gap with OpenAI and Anthropic. - CNBC

Google's Pixel phones have shifted focus to AI, sidelining practical innovations like Now Playing and Active Edge that once defined them. - Android Police

Agentic cyberwarfare: China-linked hackers used AI agents to run simultaneous reconnaissance and intrusions against Taiwan in an unprecedented autonomous cyberattack. - Financial Times

Recent AI hacking incidents at OpenAI, Anthropic, and Meta are driving a projected 12.5 percent rise in 2026 cybersecurity spending to $240 billion. - CNBC

Researchers at A Security exploited a Zoom annotation-feature flaw with fewer than 20 AI prompts to hijack meeting participants' devices. - The Verge

A survey of 116 enterprises found two-thirds enforce agent permissions, but fewer than one in five isolate high-risk agents - VentureBeat

Autonomous AI agents making independent decisions are exposing law firms to risks that existing cyber insurance policies may not cover. - Above the Law

OpenAI's unreleased Astra model solved 10 long-standing mathematics problems, sparking excitement and a credit dispute among researchers. - The Verge

Fields Medalist Tim Gowers examines why LLMs recently solved ten major open problems in mathematics, questioning if they excel more at counterexamples than proofs - Gowers's Weblog

Foom service: Redwood Research chief scientist Ryan Greenblatt says recursive self-improvement could compress years of AI progress into a single year after AI R&D automation. - Dwarkesh

Implications of recursive self-improvement - The Attack Surface

xAI's Grok Bot lets AI agents log into a user's apps and complete multi-step work autonomously, with plans priced at $120 to $200 per month. - x.ai

SpaceXAI and Cursor, merging into one company, launched Grok Bot, an AI agent app now in beta on Mac, iOS, Windows, and Linux - 9to5Mac

Elon Musk says SpaceX's AI revenue will surpass all other revenue streams by September. - Business Insider

Grand theft GPU: Cargo thieves are using violent tactics, including ramming security escorts, to steal AI data center hardware worth millions. - WIRED

Rent seeking: Hyperscale cloud providers are securing priority access to scarce AI hardware, leaving other enterprise buyers little choice but to rent capacity back from them. - theregister

Borg-belly futures: CME Group and Silicon Data will launch AI compute futures contracts tied to Nvidia GPU rental prices on Oct. 5. - CNBC

National treasure? Buce Schneier and Nathan Sanders argue the US should nationalize OpenAI and Anthropic if markets crash their valuations. - the Guardian

Warsh and learn: Federal Reserve Chairman Kevin Warsh formed a task force to improve economic data quality as AI reshapes government statistics. - Bloomberg

Artisanal hand-crafted slop: Company marketing "100% human-written, never AI" research turned out to be entirely AI-generated, using fake reviewer profiles - 404 Media

Milli VanillAI: Spotify will label AI-generated artist profiles as "AI Persona" and exclude them from recommendations starting in September. - The New York Times

Spotify's labeling move follows pressure from the Recording Academy and RIAA, after AI acts like Velvet Sundown and Enlly Blue built large followings and real earnings - Gizmodo

Better call Claude: Law firms face three distinct AI-visibility paths, expertise, recommendation and validation, requiring separate strategies to appear in AI-generated answers. - Above the Law

Legal AI vendors are pivoting from touting technical benchmarks to prioritizing user experience as lawyers report AI hasn't cut billable hours. - Above the Law

Mark Zuckerberg says universal access to superintelligent AI lawyers would make the justice system fairer. - Above the Law

An engineer intercepted GitHub Copilot's network traffic with a MitM proxy to study how VS Code and Copilot build and use context. - Lighthouse Newsletter

AI coding models are enabling 'software factories' that mass-produce and validate code through repeatable, automated pipelines, executives say. - ZDNET

Google, Meta, and Mechanize are building simulated work environments to train AI agents on real job tasks instead of chatbot responses. - Business Insider

Authors face online backlash for participating in a 2022 Google DeepMind AI writing study called Wordcraft. - Engadget

AI improves fertility treatments by identifying viable sperm - The New York Times

Scientists used AI to design bacteriophages (viruses that infect bacteria) over 90% similar to existing E. coli viruses, prompting biosecurity access-control calls. - Scientific American

A Chinese farmer lost 25 acres of sesame crops after following AI-generated pesticide advice that may have been hallucinated - Tom's Hardware

I feel the need for compute: Lockheed Martin's X-62A test aircraft used live infrared sensor data to autonomously complete 27 tactical air intercepts against a live aircraft. - New Atlas

PwC US CEO Paul Griggs built his own AI assistant, 'PG's Whisperer,' and now consults it throughout the day - Business Insider

Slop janitors: studios are firing skilled animators, then rehiring them to fix what the models get wrong. - The Washington Post

Robots that walk and talk are coming to car factories. - The New York Times

why bipedal robots on a factory floor wheels can navigate tho?

Follow the latest AI headlines via druce.ai on Bluesky