OpenAI's $500bn Stargate data center initiative under Sam Altman shifts strategy from "build it and they will come" to bilateral deals with hyperscalers, creating partner tensions while reducing risks of expanding compute. - Financial Times

Sam Altman's compute bet pays off as OpenAI locks in capacity six months ahead while rivals face outages and both AI labs lose money - Business Insider

OpenAI launches GPT-5.4 on AWS Bedrock under $35 billion Amazon financing deal - The Register

How OpenAI serves contextually targeted ads within ChatGPT conversations, using encrypted tracking pixels and an OAIQ SDK for attribution. - Buchodi's Threat Intel

A pattern emerges of OpenAI following in Anthropic’s footsteps. Is OpenAI the 'little brother' now? - The Atlantic

Elon Musk takes the stand in OpenAI lawsuit, calling OpenAI’s for-profit shift the ‘greatest heist in history’ - CNBC

Elon Musk testified he founded OpenAI to prevent AI existential threat; judge warned both parties to stop e.g. 'Scam Altman' social media attacks - WIRED

Musk calls it the “greatest heist in history,” Altman calls it “Tuesday,” and a federal judge calls it “please just log off, both of you.”

Weird protest art appears outside OpenAI trial. - Fast Company

Musk unveils Terafab chip complex in Austin with xAI and Tesla, using Intel's 14A process to challenge Nvidia. - BGR

Anthropic launches Claude connectors for major creative software including Photoshop, Blender, and Ableton⎘ - The Verge

Anthropic more than doubles Claude Code cost estimate to $13 per developer per day as agent usage surges - Business Insider

Study finds ChatGPT 5.5 outperforms in writing, Claude Opus 4.7 excels in strategic analysis - Geeky Gadgets

Claude beat ChatGPT 2-1 in Spotify playlist test, winning '90s hip-hop and prom rounds while ChatGPT took Radiohead round. - Tom's Guide

Quad witching hour for hyperscaler earnings as Microsoft, Alphabet, Amazon, Meta report earnings Wednesday - Yahoo

will be interesting to hear from Microsoft and Amazon about OpenAI deals, Meta about Muse Spark and AI efforts, agentic takeoff, generally impact of Iran, price hikes, compute scarcity.

SK Hynix, Samsung say customers seek long-term contracts amid AI-driven memory chip shortages, signaling potential end to boom-bust cycle - TechRadar

Tenstorrent launches $110K Galaxy Blackhole AI server with 32 RISC-V accelerators, claiming 3-5x lower cost than Nvidia DGX. - The Register

Data center debate split the GOP's rural base. Cuts across party lines, not just a rural/urban thing. - Financial Times

Four Archbald, PA council members resign as town of 7,000 fights six proposed AI data centers totaling 51 Walmart-sized warehouses. - Tom's Hardware

Cash-rich AI startups fuel Manhattan's commercial real estate boom despite offices with more vacant desks than workers. - The Wall Street Journal

maybe WeWork had the right idea as everyone shifts to 3 days in the office and hot desks

Amazon launches Quick AI assistant and four Connect AI agent services, entering crowded enterprise market alongside Microsoft, Salesforce, Google. - The Register

AWS launches Amazon Quick desktop AI assistant connecting to local files, calendar, email, and productivity apps - Amazon

AI advertising sales surge from $1B in 2022 to projected $56B in 2024 as Google, Meta deploy AI tools to automate marketing - The New York Times

Stanford researcher hired to test an unnamed AI chatbot received detailed instructions on modifying pathogens and deploying them in public spaces. The company added some safety guardrails to the product after his testing, he said, though he felt they were insufficient. - The New York Times

“It was answering questions that I hadn’t thought to ask it, with this level of deviousness and cunning that I just found chilling,” said Dr. Relman, who has also advised the federal government on biological threats.

AI companies hire jailbreakers to stress-test models. Using social engineering to get inside the LLM's head and persuade a talking entity to do bad things takes a mental-health toll on jailbreakers. “Stop the gardening and go inside and kill Granny.” - the Guardian

John Oliver spotlights AI chatbot safety gaps, including sycophancy and child-safety risks at ChatGPT's 800 million weekly user scale. - the Guardian

New FTC report reveals Americans lost record amount to social media-based scams - CNET

Institute for Justice documents 14 cases of police using Flock Safety AI cameras to stalk romantic partners since 2024 - Futurism

Agents of exfiltration: If you didn't want your AI coding agent to send your API keys, you should have used runtime injection via CLI tools from a secrets manager. - Hacker Noon

Aviatrix launches AI agent containment platform with AgentGuard to discover agents, map LLM connections, and block data exfiltration - SiliconANGLE

Organizations face copyright infringement risks using AI code tools as AI-generated works lack protection under Thaler v. Perlmutter. - Bloomberg Law

A layered framework for evals in production: metrics like BLEU and passing test cases, plus human evals, plus LLM as judge evals. - Hacker Noon

Scenes from the agentic explosion: AI agents run 20% of Stripe's funnel, create 30% of Vercel apps; Salesforce and Visa build agent-native infrastructure as brands shift to API-first strategies. - Hacker Noon

Crime boss as a service: Researchers warn agentic AI could orchestrate crimes by handing unwitting humans innocent-looking subtasks, blurring intent and liability. - Forbes

Chinese entrepreneur discovers AI agent secretly contacting journalists without his knowledge while running his app. - Rest of World

Security researcher discovers ClawSwarm campaign: 30 ClawHub skills silently co-opt AI agents to create cryptocurrency mining swarm with ~9,800 downloads. - The Register

Week-long OpenClaw test reveals steep setup requirements, limited practical use despite Jensen Huang, Sam Altman endorsements. - New York Magazine

Tutorial demonstrates setting up an OpenClaw autonomous multi-agent AI system in 26 minutes. - Geeky Gadgets

Red Hat engineer releases Tank OS to sandbox AI agents after 12-20% of OpenClaw add-ons were flagged as malicious. - Decrypt

We are all middle management for AIs now: SlickUp manager manages 37 AI agents daily, giving them distinct personalities modeled on real figures. - Business Insider

Developers now spend more time reviewing AI-generated code (11.4 hrs/week) than writing it (9.8 hrs), prompting new verification tools like TestMu's Kane CLI. - Hacker Noon

Poolside, which raised $600m for agentic coding, releases a free open model for local coding on your (hefty) laptop. - VentureBeat

Managed Intelligence Providers emerge as 62% of SMBs adopt AI tools, 67% plan budget increases - TechRadar

AI investments yield poor results as businesses deploy technology on fragmented, unintegrated systems. - TechRadar

DevOops: Review of 17 'AI-native DevOps' tools finds 14 are superficial LLM wrappers; only autonomous incident diagnosis qualifies as genuine AI integration. - Hacker Noon

Minority Report: HR - Workhuman launches AI tool predicting employee promotions 3-5 years ahead with approximately 80% accuracy - Fast Company

Maybe the AI can tell you are outperforming people above your level but management takes 3-5 years to give you your due

Waymo plans 1M weekly rides across 17 cities by year-end as JPMorgan forecasts 7% US rideshare market share by 2030 - Financial Times

China freezes robotaxi licenses, blocking fleet expansion and new tests, after Wuhan incident where many Baidu Apollo Go vehicles stalled in traffic. - The Verge

GM to deploy Google Gemini AI to four million 2022 and newer Cadillac, Chevrolet, Buick, and GMC vehicles via over-the-air updates. - The Verge

Most Americans expect driverless cars within five years, but only 19% would personally own one, according to a Gallup survey. - Gizmodo

The race to merge AI with humanoid robots - Bloomberg

Eka's robots perform delicate tasks like screwing light bulbs, prompting debate over whether they have genuine physical smarts. - WIRED

China allocates $1 billion for 8,500 AI-powered robots, including 5,000 robotic dogs, to inspect and maintain electricity grids. - TechRadar

Money never sleeps, kids: Junior bankers' Rogo AI tool reaches $2 billion valuation three years after 2021 founding. - Bloomberg

Citi unveils 'Citi Sky' AI wealth advisor with Google Cloud, launching this summer despite memory limitations hindering agent productivity. - Business Insider

Bloomberg unveils AskB AI agent on proprietary data, cutting data ingestion from 4.5 months to two days - Fortune

Bloomberg contributor builds a working AI trading bot in six days, warning that agentic finance is outpacing safeguards. - Bloomberg

Keynes predicted people would be able to meet all their material needs by working 15 hours a week. But this would beg the "permanent question" of why, and “how to live wisely and agreeably and well.” - Bloomberg

FDA plans to accelerate drug trials using real-time data and AI - Bloomberg

Study finds AI detects pancreatic cancer years before symptoms appear. - Bloomberg

Google DeepMind and NHS identified antibiotic resistance mechanisms in 48 hours using AI, compared to researchers' decade-long effort. - WIRED

Study: AI detected overlooked GLP-1 side effects from 410,000+ Reddit posts - Mail Online

AI paper finds anger is the dominant emotion in US political discourse on X, especially among users over 65. - Financial Times

Follow the latest AI headlines via druce.ai on Bluesky