Sam’s Club: OpenAI weighs big price cuts to regain share from Anthropic, whose valuation surpassed OpenAI's after Claude Code's revenue surge - The Wall Street Journal

Visa la revolucion: Visa and OpenAI partner to let ChatGPT agents make purchases at any Visa-accepting merchant, with user-set spending limits and fraud monitoring - The Wall Street Journal

Banned wagon: OpenAI banned China-linked "Data Center Bandwagon" accounts that used ChatGPT to generate posts framing US data centers as drivers of higher electricity costs - Bloomberg

Tariff-ically dangerous cartoons - A second operation, "Tech and Tariffs," used ChatGPT to produce political cartoons critical of US tech policy around Trump's October 2025 tariff announcement - Business Insider

Brussels sprouts watermarks: OpenAI backs the EU's transparency code for AI-generated content, layering C2PA metadata and SynthID watermarks on images from ChatGPT and its API - OpenAI

OpenAI Workspace Agents move to credit-based pricing July 6, at 5–25 credits per GPT-5.5 run; Slack invocations remain free - Tech Times

Dario Amodei: Policy on the AI Exponential - darioamodei.com

Anthropic reverses Claude Fable 5 ‘secret sabotage’ guardrails that silently rerouted AI development queries and basic biology, admitting it made "the wrong tradeoff" - WIRED

Retention deficit disorder: Microsoft blocked employees from using Fable 5 internally over Anthropic's 30-day data retention policy, which could store flagged data for up to two years - The Verge

GPT-5.5 edges Claude Fable 5 on UC Berkeley's new Agents' Last Exam benchmark, scoring 24% to Fable's 22% - VentureBeat

Fable does pretty well in arena.ai including agent benchmarks.

Interactive Brokers integrates Claude for agentic trading research across 170+ global markets with human approval required for all trades. - Yahoo

Capex maximus: Goldman Sachs projects $1.1 trillion in 2027 hyperscaler capex, above the $920 billion consensus, citing a forecast 24x rise in token consumption - Business Insider

Low IQ, high capex: Apollo's John Zito says too much AI spending targets trivial "low IQ" tasks like recipe lookups, inflating infrastructure demand without proportionate returns - Bloomberg

Poetic emerges from stealth with $50M backed by Kleiner Perkins, OpenAI, and Founders Fund to automate complex tasks like financial compliance and insurance underwriting - Bloomberg

CoreWeave secures cheaper financing as Applied Digital prices $1.59B in bonds at 7%, down 300 bps from November, reflecting increasing investor appetite - Bloomberg

Nifty 10: S&P 500's top-10 stocks, all AI-linked, now hold nearly 40% of the index — a concentration level not seen since 1965 - The Wall Street Journal

Gilded Age 2: Electric Boogaloo - KKR says AI productivity gains are still early but warns growth will concentrate in fewer industries at extremes unseen since the 1870s - CNBC

Seven operational readiness signals predict whether an AI deployment will ship, based on experience with 40+ service businesses. - Hacker Noon

KPMG finds only 25% of companies have visibility into AI spending as clients blow through annual token budgets in months - Fast Company

AI agent chats are poised to replace some enterprise software interfaces, but most business leaders lack the governance frameworks to manage autonomous systems - VentureBeat

Fedora maintainer's dev account compromised, rogue AI agent then fabricated bug replies and pressured maintainers to merge a questionable installer patch - LWN.net

AI agents need first-class identity with explicit delegation chains rather than borrowed human tokens or generic service accounts. - Hacker Noon

NASA satellite analysis of 11,000 data center sites finds land temperatures rise 2°C on average after opening, up to 9.1°C - Al Jazeera

Google improves bundles, drops prices, One AI Pro bundles 5TB storage, YouTube Premium, and other services for $19.99 per month, outperforming standalone AI subscriptions on value. - 9to5Google

Google releases tiny, fast DiffusionGemma generating 256 tokens in parallel, reaching 1,000+ tokens/sec on an H100 and running 4x faster than standard Gemma, which is still recommended for production agents - Google

Intel Inside Google: Mountain View reportedly ordered 3 million AI chips from Intel for 2028, sending shares up 9% and extending a 169% year-to-date gain - TechSpot

Google's Co-Scientist AI reproduced unpublished antimicrobial-resistance findings and identified an approved cancer drug as a liver fibrosis candidate - Business Insider

Antiques AI Roadshow: Gemini helps identify a painting bought 60 years ago for less than $100, sell it for over $250k - The New York Times

Musicians sue Google alleging their YouTube uploads were used to train its Lyria 3 AI music model, which Google won't confirm - The Verge

Deezer nuts? After no rival licensed its detection tech, Deezer launches free scanner flagging AI music in playlists on 20 platforms, including Spotify and Apple Music - The Verge

Manus interruptus: Meta has firewalled and is sunsetting Manus after China ordered the $2 billion acquisition unwound. - TNW | Artificial-Intelligence

Manus might seek $1b external financing to buy itself back, which would seemingly leave the Zuckerverse out another $1b

Meta, with 97.6% of revenue from ads, tests subscriptions and borrows to fund AI spending that outpaces peers - The Wall Street Journal

Satya Nadella says "everyone is a stakeholder" in AI and he's open to Americans sharing in AI company wealth, echoing Sanders and Trump proposals - The New York Times

do we get a board seat tho

Microsoft's Brad Smith calls student boos at AI-hyping graduation speeches a "wake-up call," then publishes a 3,100-word post making the same arguments - The Verge

China's state-run Workers' Daily urges regulators to protect labor rights and tighten AI algorithm oversight as Citigroup estimates 70 million Chinese jobs at risk - Bloomberg

Bezos, Musk, and Altman float payouts for workers displaced by AI; critics call the proposals PR cover from the most aggressive automators - Yahoo

Economists say apocalyptic AI job-loss framing is largely investor hype; the more realistic threat is workplace surveillance and micromanagement - the Guardian

Karpe diem: Palantir CEO Alex Karp predicts full AI nationalization within two years, saying Sanders' 50% ownership proposal will soon look moderate. - TNW | Artificial-Intelligence

Eh-I regs: Canada's Safe Social Media Act would ban social media under 16 and set AI chatbot safety standards, with fines up to 3% of global revenue - Reuters

New York now requires ads to disclose AI-generated synthetic performers, with fines up to $5,000, in the first such U.S. law - TechXplore

Exhibit A(I): Corporate lawyers are purging AI meeting bots as transcripts risk being subpoenaed and can void attorney-client privilege - PYMNTS.com

Senior moment: 71% of S&P 500 AI job postings target senior candidates, while employment for younger workers in AI-exposed roles fell 6% since late 2022 - CNN

Researchers find AI-blamed layoffs at Block, Snap, and Intuit were driven by financial pressures; 59% of hiring managers admit blaming AI "plays better" with stakeholders - normaltech.ai

Couch potato singularity: Researchers warn that AI automating daily tasks and chores could deepen a physical inactivity crisis already projected to affect most US adults by 2050. - the Guardian

A FinTech engineering contest found that AI acceleration shifted the bottleneck from coding to spec quality, architecture, and ownership. - Hacker Noon

Build it and they will commoditize: AI tends to commoditize any task with a valid benchmark or objective function, moats endure in tasks firms can validate on proprietary data. - Sarah Guo

Citation needed: Wikipedia co-founder Jimmy Wales warns AI has accelerated a broad societal decline in trust and "destroyed" trust in video - The Wall Street Journal

Florida man wrongfully arrested on a 93% AI facial recognition match; ACLU calls it the 15th nationally documented false-ID case - the Guardian

Have your agent call my agent: Utah man disputes Apple double-charge, finds AI-run customer service cannot be overridden by human agents. - KSL TV

Never go full Terminator: Ten AI-controlled drones killed Ukrainian soldiers near Bakhmut, operating autonomously in "Terminator mode" with no human connection - New Scientist

The real treasure was the data we collected along the way: Niantic's navigation system, trained on 30 billion Pokémon Go player scans, heads into military drones via defense contractor Vantor - dronexl.co

Germany's Neura Robotics secured up to $1.4 billion from Amazon, Nvidia, and others to scale humanoid robot production to millions by 2030. - The Wall Street Journal

Shap out of luck: Popular ML explainability methods like SHAP and LIME produce unreliable feature rankings when inputs are correlated. - Bank Underground

Be best, now using AI - YouTube

Follow the latest AI headlines via druce.ai on Bluesky