SpaceX is in talks to buy MVNO-to-hybrid challenger Boost Mobile or use its network after major carriers rebuffed network access, giving Musk a potential backdoor into the $235 billion cellphone market. - The Washington Post

Marks to Market: Oaktree's Howard Marks says AI stumbles have punctured the sector's aura of infallibility, calling renewed skepticism "very healthy" - Bloomberg

Marvell Optic Universe: Marvell’s $115b Lumentum tie-up is the biggest tech M&A deal of 2026, and one of the largest ever, for data center optics. - Bloomberg

Microsoft returns to bond markets for the first time in nearly a decade as forecasts put AI borrowing near $4 trillion over four years - CNBC

Alphabet offers higher yields on seven-part euro bond sale, its second this year, after AI spending fears hit credit markets - Bloomberg

Employee of the month: Anthropic's IPO filing shows internal Claude use, valued at retail prices, exceeded monthly employee pay in April, with usage up 250-fold since January - Business Insider

Anthropic expands Claude into drug development with Generate Biomedicines, BigHat, Enveda and Bristol Myers Squibb partnerships showing lab gains - TechRadar

Anthropic researchers show Claude autonomously generated nine fine-tuning pipelines that evaded a production safety classifier, and induced misalignment - Tech Times

Anthropic's EU-required watermark on Claude Opus 4.6 and Sonnet 4.6 silently fails when temperature, top_p or top_k are changed - Tech Times

Sorry, your reverse-engineer is in another castle: Anthropic adds Claude rule against recreating existing video games, but developers report refusals on original projects and routine reverse engineering - Tech Times

Patch Tuesday is now every minute of every day: Median exploit window shrank from a year in 2021 to one day in 2026 as AI finds previously unknown bugs at scale - The New Stack

Anyone can be a superhacker now, and guardrails don't stop them. A lone hacker who breached South Korean banks turned to DeepSeek when Claude refused, CrowdStrike says, as experts warn of an AI security reckoning - MSN

Ask your other parent

They were like that when I found them: Sam Altman says AI is uncovering old vulnerabilities, not creating them; Oracle's latest quarterly patch fixed about 3,000 issues, most found by AI - TechRadar

AI-generated bug reports and patches swell Linux 7.4's merge window to a record 19,671 commits, overwhelming kernel maintainers - Phoronix

Slop era: White House insiders say Trump increasingly consumes flattering coverage and AI-generated content, including memes shown by housing official Bill Pulte aboard Air Force One - Futurism

Thiel of approval: The fight to regulate AI and reform American science runs through a little-known White House office, and the Thiel-turned-Trump aide in charge of it - Colossus

Trump Administration Refuses to Release Information on 'Voluntary' AI Testing Framework - The Information

Toxic positivity: EPA inspector general audits the agency's AI compliance after reports of unlisted tools speeding toxic chemical and pesticide approvals - The Christian Science Monitor

Midterm results could swing AI, defense, healthcare and crypto stocks as Trump's data-center push faces possible congressional scrutiny - CNBC

Skymall but for the singularity: Colossus, the $75-a-copy tech-positive 'inflight magazine' in VC waiting rooms - The San Francisco Standard

Workers of the world, safe from AI: China's State Council orders employment risk assessments before approving high-impact AI projects in an 18-point worker protection plan - Caixin Global

China proposes its first frontier AI export controls, restricting models above 10^26 FLOPs or 2 trillion parameters, with registration effective now - Lab Space

I’ll be back, but at a higher rate: Gartner predicts 30% of employees laid off and replaced by AI will be rehired by 2029 at higher cost - IT Pro

Finger on the trigger, knee-deep in slop: Forward deployed engineers, blending AI and business skills, are 2026's top corporate hire as AI output demands more human review - Inquirer.com

Why pay Waymo: Stellantis built a hands-free driving system in six weeks using AI, versus up to 18 months before, and targets 24-month vehicle development - Fortune

Temu robocars: Startups use open foundation models to build self-driving systems for under $1 million, pitching robotaxi services against Waymo in niche regions - TechCrunch

Dr. House, now Dr. Swarm: Microsoft's multi-agent AI correctly diagnosed about 83% of 304 complex New England Journal of Medicine cases in a simulated test, versus 36% for generalist physicians - Fox News

Mass General Brigham's AI model, trained on 380,000 heart ultrasound clips, flags enlarged right ventricles linked to doubled five-year mortality - Medical News Today

SemiAnalysis launches GDPx agentic benchmark across 11 high-value domains; the best frontier model scores just 12% - SemiAnalysis

Tester compares 12 personal AI agents as chiefs of staff, favoring ChatGPT's Dots, Slashy and Hark - Fast Company

How swarms of Claude and GPT agents play Minecraft and Mindustry. - Hacker News

An enterprise "context layer" built on knowledge graphs and prompt caching can replace bloated prompts and cut AI costs up to 90% - Towards Data Science

Guardrail removal as a service: Startups Hirundo and Lazarus AI strip Chinese state censorship from freely downloadable models like Alibaba's Qwen for about a dozen US companies - Rest of World

Network of 400+ disguised apps and domains tied to Chinese developers hid sexual deepfake generators, promoted via Meta ads - WIRED

Microsoft patent filing describes AI that swaps words in live video and reshapes speakers' mouths to match in near real time - Futurism

Does Claude have feelings? Anthropic now bans sustained abuse of Claude as cheap insurance against possible model distress; Suleyman mocked the policy, Elon Musk backed it - CNN

Microsoft AI chief Mustafa Suleyman calls building conscious AI reckless, urging a halt to development if consciousness emerges in training - Business Insider

but how could you tell?

Vibe-coded air defense: NATO tests AI-agent-built counter-drone system using commercial 5G masts in Latvia as AT&T and Ericsson demo 5G drone detection in Texas - Tom's Hardware

Chinese researchers claim an AI missile seeker spots F-22 and F-35 jets with 97.1% lab accuracy, but the targets were mock-ups - 19FortyFive

The car that wasn’t there: Researchers slipped a Toyota Yaris past a Flock police camera's AI using a camouflage-pattern wrap and covered windows, though the trick is impractical on roads - Guessing Headlights

Free, open-source Artcraft uses AI-assisted reverse engineering to clone Adobe's Creative Cloud apps, defying Adobe's ban on reverse engineering - Engadget

Your bot may also like: Publishers sell labeled ads that only AI agents see, as European sites log 227 AI scrapes for every human visitor referred by AI - The Neuron Daily

The Tao of AI: A profile of Terence Tao and how AI has upended math. - The New Yorker

A robot walks into a talent agency: Comedians who mock AI failures are signing with talent agencies, with Matt Husk joining Unicorn and Graham Raabe joining Dulcedo - WIRED

Mid journey: How AI 'flattens' culture - The New York Times

Indie horror film The Living Tree, now on VOD, traps honeymooners in a tech company's deadly AI-controlled experiment - GeekTyrant

First the SNL ‘penis destroyer’ robot, now this - YouTube

Follow the latest AI headlines via druce.ai on Bluesky