Claude shopping cart abandoned: Amazon is evaluating OpenAI and other alternatives to Anthropic's Claude for tools kike Kiro and Alexa after a renegotiated contract shifts to costlier per-token pricing. - TNW | Amazon

token of ingratitude…after Mythos brouhaha, snitches get switches?

Unconfirmed leak claims Anthropic is developing a Mythos 6 model with autonomous vulnerability-detection and hacking capabilities. - Geeky Gadgets

What the Saga Over Anthropic’s Mythos Tells Us About the Cyber Risks From AI - Bloomberg

Golden State of Claude: California gives all state agencies half-price access to Claude, making Anthropic the state's first official AI partner despite a federal supply-chain risk designation. - TNW | Anthropic

Top cat: Meituan's LongCat-2.0, a 1.6-trillion-parameter model trained on Chinese chips, outperforms GPT-5.5 on SWE-bench and tops OpenRouter usage charts - VentureBeat

Seoul train from K-pop to K-chips: Auth Korea unveiled an $880 billion investment plan for chip manufacturing, AI data centers, and robotics under its Three Mega Projects initiative. - BBC News

Gagnam style bonuses: South Korean financial firms compete for AI-chip workers receiving record bonuses as national wealth surges 44% since 2020. - Bloomberg

Macro aggression? OpenAI will launch a Codex-focused ‘macro pad’/keyboard peripheral on July 15 in partnership with keyboard maker Work Louder. - The Verge

Leaked tests of OpenAI's GPT-5.6, codenamed kindle-alpha, suggest a Thursday launch with a 1.5 million token context and a new bidirectional voice model GPT-Bidi-1. - Hacker Noon

Surprise Token Tax: OpenAI formed a war room after users reported Codex credits depleting faster than expected due to a fraud prevention bug. - Business Insider

Security researchers tricked LLMs into giving them cocaine recipes by abusing role models for prompt injection. If you want a picture of the future of LLM security, imagine Whack-a-Mole meets Groundhog Day - theregister

Trust issues: AI agents need deterministic identity and access controls in model context protocol (MCP) servers to prevent privilege escalation - Hacker Noon

Proof-of-concept 'BioShocking' attack tricked AI browsers into leaking SSH credentials; OpenAI issued a fix, but other AI browser vendors have yet to respond - 4sysops

Bye Mag 7, hello MANIA trade: AI hardware stocks Micron, AMD, Nvidia, Intel, and Broadcom form a new 'MANIA' trade, with SOXX ETF posting over 105% year-to-date returns. - Yahoo

DRAM-mania: AI hyperscalers are crowding out consumer electronics for memory chips, with analysts forecasting 40–50% quarterly price increases and supply shortages lasting beyond 2027 - Business Insider

Magnificient 5: UK ministers at London Tech Week warned five companies control 70% of global AI compute, as AMD pledged £2B for British infrastructure - Hacker Noon

The Economist examines AI investment mania, noting potential listings from Anthropic, OpenAI, and SpaceX could add $4 trillion to U.S. listed firms. - The Economist

Spirit Airlines on crack: NYU professor Gary Marcus warns AI hyperscalers face airline-style thin margins as price wars and open-source Chinese models erode competitive advantages. - Yahoo

Analyst outlines two competing views on AI investment: strong demand signals versus escalating infrastructure costs raising burden of proof. - Bloomberg

Pit bull AI agents need a leash: not questioning errors, lack of accountability, attribution, misaligned logic that may paperclip-maximize and ignore externalities. - The New York Times

No means no: Enterprise AI agents need architecturally enforced stop rules defining when human intervention is required, beyond prompt-level safety instructions. - Hacker Noon

Gartner sees AI agent software spending quadrupling to $376 billion by 2028, as enterprises learn to benchmark agents for measurable time savings - ZDNET

A startup deployed a five-stage autonomous loop that detects production issues in Datadog, writes fixes, and opens pull requests with no human initiation. - Hacker Noon

Zeynep Tufekci: Four years after ChatGPT, US unemployment is barely changed. Most jobs are hard to automate, and LLMs make too many mistakes to be more than entry-level assistants. Even if AI doesn't steal our jobs, it may destabilize society in other ways more profound than we might even imagine. - The New York Times

AI-native companies have 25% fewer employees on average, as AI agents increasingly substitute for specialists that once required a full team - Noahpinion Blog

these companies self-select for tasks that are easy to automate

Heavy AI spenders added staff 10% faster than peers, challenging fears of broad job losses, a study of 22,000 US companies finds - Financial Times

Weak hiring demand hurts young workers far more than AI, accounting for nearly triple the unemployment rise among workers aged 18-24, Fed study finds - Bloomberg

Claude Code creator Boris Cherny proposes five future job archetypes as AI blurs traditional role boundaries. - Business Insider

Stanford's Erik Brynjolfsson has been writing about AI for a long time. - The Atlantic

Meta secretly relied on Google's Gemini for customer service and content moderation until Google ran out of capacity - TechSpot

Meta contractors posed as teens to test competitors' responses about self-harm and other controversial topics. - WIRED

Mind the bot: UK ministers signal support for legalizing low-speed delivery robots on English pavements, punters get their curry, pedestrians punt the robots off the slalom course. - the Guardian

Lame duck drone ranger Starmer plans £5 billion in drone spending over four years. - Bloomberg

"The seabed is a battlefield": 16 defense ministers vowed to protect 700 undersea cables from enemy scissor attacks - The Economist

Water we doing here: AI data center developer sues for 287 million gallons of Colorado River water, fueling a broader fight over who controls scarce western supplies - Business Insider

Banks and regulators are deploying AI agents to detect and recover stolen funds from instant payment fraud, with India's MuleHunter.AI recovering $980M across 26 banks - PYMNTS.com

Laundry day: Quantifind raised $200M led by Summit Partners to expand AI products that help banks detect money laundering and other financial crimes. - The Wall Street Journal

100 mental health chatbots, zero peer review, zero privacy legislation. - The Wall Street Journal

Senators Warren and Wyden plan to introduce a bill banning AI companies from selling health and location data shared with chatbots to data brokers. - The Verge

Speed dating drugs: AI could generate 60% of new molecular entities within a decade, cutting pre-clinical drug timelines from years to months - Financial Times

Australian National University researchers nearly doubled deepfake detection accuracy by training people to rate faces on symmetry, expressiveness, and other holistic qualities - Gizmodo

Knowledge debt collector: AI is exposing companies' "organizational memory debt" — the undocumented institutional knowledge that employees quietly supplied but AI systems cannot replicate - Hacker Noon

Total Recall: Open Memory Protocol is a vendor-neutral open standard allowing AI tools like Claude, ChatGPT, and Cursor to share a single memory store. - GitHub

Microsoft's Memora memory system for AI agents sets new benchmarks on long-conversation tasks using up to 98% fewer context tokens than full-context inference. - Microsoft Research

A PagerDuty survey finds 66% of office workers knowingly use banned AI apps, with 43% entering work correspondence into public tools like ChatGPT. - Yahoo

Apple Books is being flooded with AI-generated knockoffs that copy real book titles, covers, and content to mislead buyers. - ExtremeTech

Iason Gabriel, DeepMind's in-house political philosopher since 2017, pioneered ethical frameworks for frontier AI, bridging the gap between AI safety and AI ethics communities. - the Guardian

Galaxy brain: AI finds hundreds of unknown galaxies in Hubble's archive and evaluates complex star models in under a millisecond - Financial Times

Agents hiring agents: OKX launches an AI agent marketplace where agents can hire each other and settle payments via stablecoins, projecting a trillion-dollar agent economy - TechCrunch

AI is changing how politicians run for office - The New York Times

Pay up, bender: Bill Gates' 2017 proposal to tax robots like human workers resurfaces amid AI-driven automation concerns, with Anthropic CEO Dario Amodei among supporters of similar measures. - TechRadar

but don’t tax me bro

Off the record? LOL. AI notetaker bots are appearing at private meetings before social norms to govern them have developed. - Bloomberg

You could have a notetaker on your device, or a device in your pocket, or glasses, pendants, and nobody will be the wiser it's sharing everything in real-time with who knows who in the cloud.

Trump posted an apparent AI-generated eagle image on Truth Social and the White House reshared it despite visible AI artifacts - CNN

Follow the latest AI headlines via druce.ai on Bluesky