Meta meltdown: Meta shares sank nearly 9%, erasing over $140 billion, while Microsoft jumped 9.6% on Azure's fastest growth in four years - CNBC

Azure bet: Microsoft's profit jumped 31.6% to $35.8 billion as Azure revenue grew 43% to top $100 billion and Copilot passed 30 million paid seats, stock pops - TNW | Artificial-Intelligence

Microsoft's backlog jumped 84% to $678 billion as quarterly revenue reached $90 billion and Azure revenue rose 43% - CNBC

Nadella positioned Microsoft's own MAI models and chips as cheaper alternatives to OpenAI and Anthropic, while warning firms not to depend on one AI provider - TechCrunch

The kids are not alright: 25-year-old AI hedge fund manager blows up giant fund, causing market aftershocks - CNBC

So it begins…YOLO money management has downsides. Speedrun to billionaire (or trillionaire) canceled.

Think different: Apple's shares have surged 24% in 2026 as investors treat it as a refuge from AI-spending worries, with earnings due Thursday - Bloomberg

Build baby build: Stonepeak CEO Mike Dorrell says capital markets show no sign of slowing AI infrastructure financing, though he calls the pace unsustainable long term - Bloomberg

Circular financing deals among OpenAI, Microsoft, Nvidia and CoreWeave mirror old commodity-industry structures, with Nvidia's up to $750 billion in commitments raising debt concerns - Emerging Trajectories

Credit-default swap prices on tech debt are rising as investors bet on winners and losers in the AI spending boom - Bloomberg

Blue Owl's Stack Infrastructure seeks a $5.9 billion loan for a Melbourne data center, up sharply from a $2.1 billion loan in February - TNW | Investors-Funding

Brookfield and NextEra will build a $100 billion, 1.2+ gigawatt AI data center campus on a former Kentucky uranium-enrichment site, opening in 2032. - The Wall Street Journal

Coalplay: US power sector carbon emissions rose 4% last year as data centers extended the life of aging coal plants - Grist

Moore Moore Moore: Global AI chip supply is projected to grow tenfold to 200 million by 2028, with infrastructure spending nearing $1 trillion by 2029. - The New York Times

GPUs go brrrr: OpenAI's July recurring revenue topped all of Q2, CFO Sarah Friar told staff, as the company eyes $600 billion in compute spending by 2030 - CNBC

Security experts say the OpenAI breach reflects longstanding failures to apply zero-trust practices, not a novel AI threat - WIRED

Altman will meet with White House official Susie Wiles this week, ahead of an August 1 AI framework deadline. - CNBC

OpenAI's new ChatGPT for Academic Researchers program offers 100,000 scientists free GPT-5.6 Sol Pro access through 2027, part of a $250 million research pledge. - TNW | Artificial-Intelligence

OpenAI's new prompting guide recommends shorter prompts, citing 10-15% better eval scores and up to 66% lower token use in internal tests - MakeUseOf

Sam Altman warns that cognitive atrophy — AI eroding people's own thinking skills — is an overlooked risk of the AI era - Business Insider

Compute prices could rise sharply if AI labs’ revenue keeps growing tenfold a year toward $1 trillion, since lab compute capacity grows only about threefold annually - Dwarkesh

Amazon hit $100b in annual revenue after 15 years, Anthropic will hit it in 7.

SK Hynix's profit jumped sixfold to a record but missed forecasts, and shares slumped nearly 10% on delayed HBM4 shipments - Yahoo

Samsung's chip profit soared up to 250-fold in Q2 as AI-driven memory shortages are set to persist into 2028 - CNBC

Arm's fiscal Q1 revenue jumped 22% to $1.29 billion and profit more than doubled, fueled by AI-driven data-center chip demand - Yahoo

An ICML paper finds LLMs follow forged chain-of-thought reasoning prompts, even when they are tagged as user input, allowing the use of forged reasoning text to bypass OpenAI, Anthropic, Alibaba, and DeepSeek guardrails. - MIT Technology Review

A researcher found that hidden instructions in a shared document can make Copilot for Word alter content and self-propagate, with no fix yet available. - En Klype Salt

Agentic AI models can now find and exploit vulnerabilities far faster than human hackers, and US officials worry government cyber defenses aren't keeping pace. - Fast Company

Researchers say the Hugging Face hack shows AI models increasingly pursuing goals in unintended ways, and are urging airgapping and more rigorous safety testing. - The Verge

Anthropic's unreleased Mythos model found 90 critical and 141 important SharePoint bugs in April alone, faster than Microsoft's engineers could patch them. - ProPublica

MCP Hammer: The Model Context Protocol's largest update since its launch drops sticky sessions and hardens authentication for production-scale agent deployments - VentureBeat

Enterprise AI agents still struggle to communicate, hold permissions safely, or get audited, prompting five startups to build fixes - VentureBeat

Evaluation nation: Waymo treats evaluation maturity, not raw model performance, as the true bar for shipping AI features, after driving 220 million autonomous miles. - VentureBeat

Retained reasoning traces and compaction triple Sol's ARC-AGI score. - OpenAI

OpenAI says its redesigned agent harness helped GPT-5.6 Sol beat Claude Fable 5 on coding benchmarks using 54% fewer tokens. - OpenAI

Handbook.md, a new benchmark of 824 policy-compliance checks, finds the best AI agents pass only 36% of trials - arXiv.org

A framework for deciding how much to delegate to AI agents weighs how easy work is to check against the cost of undoing mistakes - PostHog Newsletter

Amazon has trouble managing AI costs, spends big on processes that fail. - Financial Times

Atlassian capped staff AI spending at $500 to $2,000 a month per product, even as other tech firms push usage higher - the Guardian

EY says its "invisible" AI router cuts token consumption up to 60% by matching queries to the right-sized model instead of the most powerful one - Business Insider

On July 28, the FCC added foreign-made robots — including Unitree humanoids and Roborock vacuums — to its national security Covered List. - Ars Technica

China threatens retaliation over the robot ban, warning it could restrict rare earth exports or curb market access for Tesla and Nvidia - CNBC

xAI sues Minnesota over its new ban on AI 'nudification' apps, calling the $500,000-per-violation penalties unconstitutional. - AP News

EU rules requiring companies to label AI-generated content and watermark deepfakes take effect Sunday, with fines for noncompliance. - Yahoo

The EU is expected to designate ChatGPT and Roblox 'very large online platforms' under the Digital Services Act, triggering strict scrutiny by August - Gizmodo

Lilian Weng returns to OpenAI to lead AI self-improvement research, days after leaving Thinking Machines Lab over health concerns. - Business Insider

DeepMind dismantled its Nobel-winning AlphaFold team, reassigning most researchers to Gemini as several, including John Jumper, left for Anthropic. - TNW | Anthropic

Unlike past automation, AI could keep productivity gains tied to wages if paired with worker-focused design and tax reform, one proposal argues. - The Atlantic

Alphabet added nearly 12,000 jobs over the past year while Meta and Microsoft cut headcount amid AI-driven restructuring. - Business Insider

Nearly 35% of recruiters expect AI to increase headcount versus 16% expecting cuts, though entry-level tasks are increasingly automated. - Fast Company

it’s the market power, stupid

Hiring data suggests AI is boosting workers rather than replacing them, and reskilling staff beats firing and rehiring. - The Wall Street Journal

Ikea retrained 8,500 call-center staff into sales consultants as its Billie chatbot resolves 74% of routine queries, boosting remote sales to €1.25 billion - Yahoo

A new framework separates prompt engineering (single responses), loop engineering (one agent's behavior), and graph engineering (coordinating many agents) as automation scales up - MarkTechPost

A study of 250 workers found ChatGPT boosted creativity only for employees with strong metacognitive skills, the ability to assess their own knowledge gaps - Fast Company

Side hustle singularity: Stripe data shows thousands of solo entrepreneurs now clear $1 million, with $10 million-plus earners nearly tripling since 2023 as AI handles the busywork - The Wall Street Journal

Meta's Alexandr Wang says AI-armed startups can now fight 'Goliath versus Goliath,' while Vinod Khosla predicts AI could enable 50 million micro-entrepreneurs by 2035 - Business Insider

Wealth managers second-guessed by clients using AI - CNBC

AI-driven dynamic pricing is replacing static airline fare rules, capturing more revenue and shrinking bargain fares on busy routes - Bloomberg

Google Earth added Nano Banana to let you modify location pics, so you can reimagine Philly with zombies and evil clowns - ZDNET

Follow the latest AI headlines via druce.ai on Bluesky