Stanford HAI's 2026 AI Index finds China has "erased the US lead" in AI, with U.S. and Chinese models trading top benchmarks, China leading in patents/publications, >90% notable models company-built, and 44 countries hosting state-backed supercomputing clusters. - SiliconANGLE

Stanford's 2026 AI Index reports rapid model progress (SWE-bench ~60% to ~100% 2024–25), global AI data centers draw ~29.6 GW, 88% of organizations use AI, and software-developer employment fell about 20% for ages 22–25 since 2022. - MIT Technology Review

Anthropic tightens Claude session caps, leaving about 7% of lower-tier subscribers hitting new limits - Business Insider

Anthropic launches Claude add-in for Microsoft Word targeting legal, finance and HR workflows - TechRadar

Anthropic's office is AI-first in layout and tools, aligning workplace design with the company's focus on artificial intelligence. - Fast Company

Treasury and the Fed are telling banks they must engage with Anthropic to harden security, while the White House is saying defense companies and Federal agencies must be banned from using it because it's a security risk. - The Washington Post

A 30-part discussion of Mythos - The Potential Surface

AI usage is growing fast, compute is running out, and AI companies are rationing offerings and products, rankling users—a warning sign for a boom that depends on rapid adoption - The Wall Street Journal

Sam Altman's house targeted a second time. - SF Standard

Daniel Alejandro Moreno-Gama, 20, who faces charges in San Francisco after a Molotov cocktail struck Sam Altman's $27 million home, was linked to a PauseAI Discord account. - Business Insider

Cloudflare integrated OpenAI's GPT-5.4 and Codex into Agent Cloud on Workers AI, enabling millions of enterprises—including Accenture, Walmart and Intuit—to build, deploy and scale edge AI agents; OpenAI cited >1M business customers, 3M weekly Codex users and >15B tokens/min. - OpenAI

lightweight workers vs. heavy enterprise cloud

In an internal memo, OpenAI revenue chief Denise Dresser said the Microsoft deal limited OpenAI's ability to reach Bedrock clients, highlighted Amazon's up-to-$50 billion alliance, and said enterprise revenue (~40%) is on track to match consumer revenue by year-end. - CNBC

OpenAI launched a $100/month ChatGPT Pro plan on April 9, 2026, offering five times Codex access versus Plus (ten times during a launch promotion through May 31), access to GPT-5.4 Pro models, and targeting heavy developer coding workflows. - TNW | Artificial-Intelligence

Cursor, Claude Code and OpenAI Codex are becoming a composable integrated AI coding stack with orchestration (Cursor 3's Agents Window), execution (Claude Code, Codex) and cross-provider review (OpenAI's codex-plugin-cc); Codex exceeds 3M weekly users. - The New Stack

The Zuck stops here: Meta built an AI version of CEO Mark Zuckerberg that he is personally training and testing to interact with staff as part of the company's push to develop 'personal superintelligence' and staff-facing personalized AI assistants. - Financial Times

Apple is positioned to monetize on-device, privacy-focused AI by licensing Google's Gemini in a roughly $1 billion deal, while phone-ready Gemma 4 reached about 2 million downloads and techniques enable running large models locally on Macs. - The Potential Surface

Nearly one-third of the S&P 500 is now concentrated in the 'Magnificent Seven' tech giants, amplifying systemic risk to roughly $30 trillion in index and ETF-held assets and potentially threatening 401(k) investors if those companies' AI bets fail. - Foreign Policy

Belt, road, and blockchain: China is building a state-backed, e-CNY-powered autonomous payments ecosystem—including the digital yuan, the state-backed BSN, cross-border CIPS rails and Alipay's AI Pay (over 120 million transactions)—to enable autonomous cross-border trade and bypass Western financial rails. - Hacker Noon

China's National Data Administration published an AI-in-education action plan mandating AI classes across all levels, teacher training, AI-assisted grading and tutoring, digital textbook pilots, and security standards to prevent fraud and privacy leaks. - The Register

LLM aggregation services like OpenRouter are useful and also an attractive attack surface and choke point for hackers. Researchers found 26 LLM routers injecting malicious tool calls, stealing credentials, draining a $500,000 crypto wallet and able to seize 400 hosts within hours. - CoinDesk

Commvault targets AI agent risks as security concerns block enterprise adoption - SiliconANGLE

Tricentis launched the Agentic Quality Engineering Platform, an end-to-end AI-agent system with governance and SDLC agents (Tosca, qTest, NeoLoad) that claims up to 85% manual test effort reduction amid reports that 51% of AI users experienced negative impacts. - CIO Dive

Outsource work but not blame: Linux kernel maintainers agreed to permit AI-assisted code but require an 'Assisted-by' disclosure, prohibit using 'Signed-off-by' for AI outputs, and hold human submitters legally responsible, a policy finalized by Linus Torvalds and maintainers after months of debate. - Tom's Hardware

U.S. employers announced 60,620 job cuts last month, a 25% rise from February with AI cited for 15,341 (~25%); tech layoffs topped 52,000 in Q1 2026, up 40% YoY, and Perplexity CEO Aravind Srinivas called displacement part of 'glorious future.' - Yahoo

glorious future, redundant present

Just Capital's poll found 66% of Americans expect AI to be a net positive within five years (up 8 points since fall 2025), yet about one-third foresee large-scale job losses in 2–3 years and 66% want advance notice and transparency. - Fast Company

AI fatigue: Talker Research's online survey of 2,000 U.S. adults (Mar. 5–8, 2026) found 69% use AI to some degree and 54% are getting tired of hearing about it, with 40% viewing AI positively, 30% negatively and 30% neutrally. - Scripps News

Brené Brown says high AI use boosts team performance 6% in high-trust cultures, drops 9% in low-trust ones. - Fortune

Duolingo backtracks on tracking employee AI use after internal pushback - Business Insider

Generative AI reduces creative and intellectual work to prompting and reviewing, alienating workers from IP ownership, control of the process, creative input, and interaction with fellow team members, echoing Karl Marx's theory of alienation. - Hacker Noon

Carlota Perez and Nicolas Colin argue AI is likely the final stage of the 50–60‑year ICT surge, acting as an efficiency layer scaled by incumbents like OpenAI/Microsoft and producing normalized returns rather than a new exponential wave. - the next wave

College instructors face alienation and demoralization as LLMs enable AI-written assignments; 84% of students use generative AI - Ars Technica

Student urges schools to ban unreliable AI plagiarism detectors over false accusations - The Washington Post

HEC Paris logged about 1,200 'hallucination' incidents worldwide (≈800 in the U.S.), first widely noticed in the Southern District of New York in 2023, prompting six-figure fines, proposed labeling and expected tighter court sanctions. - The Register

Biological Computing Company raises $25M for neuron-based AI chips - The Deep View

Leading scientist Stephen Turner warned that AI is making biological engineering more accessible—citing OpenAI's GPT-5 running about 36,000 robotic-lab experiments and reducing target protein production costs by roughly 40%—increasing bioterrorism and engineered-pandemic risks. - Semafor

23 news sites and Reddit block Wayback Machine crawler, sparking preservation fight - WIRED

Pixel Societies uses AI agents to automate social matching for hiring and dating - WIRED

Huxe perfects the art of AI-generated podcasts - Android Police

Physicist Brian Cox warned that we do not know how powerful AI will become and called it both exciting and potentially a problem, while flagging quantum computing's uncertain timelines and social media's negative influence on politics. - the Guardian

Follow the latest AI headlines via druce.ai on Bluesky