Author: Andrew

Enterprise AI Growth and Education Disruption (Links) – May 23, 2026
- Cursor: Introducing Composer 2.5 (May 17, 2026)
  Composer 2.5 is now in Cursor, offering stronger intelligence, more reliable long-task performance, and smoother collaboration. It uses scaled training, targeted textual feedback, and extensive synthetic tasks,
- NY Times: OpenAI Bought Company That Offered A.I. Tools for Cloning Voices (May 15, 2026)
  OpenAI quietly bought Weights.gg, a startup that let users create and share AI voice clones of celebrities and politicians. It bought the team and IP, shut the service, and plans limited, partner-only uses.
- Anthropic: KPMG integrates Claude across its core business and workforce of more than 276,000 in strategic alliance
  KPMG formed a global alliance with Anthropic to embed Claude in Digital Gateway, giving 276,000+ employees access, and adding tax, legal, and cybersecurity tools.
- NY Times Opinion: What A.I. Did to My College Class (May 16, 2026)
  At Stanford, A.I. has reshaped campus life, fueling startups, lavish wealth, and widespread cheating, prompting a return to proctored exams. It narrows job prospects, erodes learning, and normalizes dishonesty.
- Konstantin Tkachuk: The Floor Doesn't Exist (May 13, 2026)
  AI didn’t invent new attack types but slashed the cost and skill required: hacking is now affordable via commercial subscriptions rather than elite labor.
- NY Times: Josh Tyrangiel book excerpt: How OpenAI and Khan Academy Made a Chatbot (May 16, 2026)
  OpenAI and Khan Academy built Khanmigo, an AI tutor echoing Sal Khan’s pedagogy. Their rocky, secretive collaboration exposed benefits, risks, and trade-offs over accuracy, bias, and reputation.
- The Chronicle of Higher Education: A New ‘AI First’ College Aims to Offer Cheaper, Employer-Friendly Degrees (May 14, 2026)
  Sal Khan is launching Khan TED Institute, a nonprofit offering a $10,000, competency-based applied-AI degree with partners Khan Academy, TED, ETS, Google, Microsoft, and Accenture. The online, project-focused program builds portfolios, group work, and practical skills for the AI era.
- The Chronicle of Higher Education: How the Commencement Speech Became One of Colleges’ Biggest PR Problems (May 14, 2026)
  High‑profile incidents include boos at UCF over an AI remark, Georgetown Law’s Morton Schapiro backing out after a student petition, disinvitations at Rutgers and South Carolina State, and protests at Howard and Utah Valley.
- NY Times: Anduril Raises $5 Billion in Funding and Is Valued at $61 Billion (May 13, 2026)
  Anduril raised $5 billion, valuing the AI-backed weapons start-up at $61 billion, twice its value a year earlier.
- Anthony B. Bradley: America Is Closing Its Elementary Schools, and Nobody Wants to Say Why (May 18, 2026)
  Elementary schools are closing nationwide, in cities, suburbs, and rural areas, not mainly because of homeschooling or private schools, but because fewer children are being born. This shrinking birth rate, below replacement, is driving district consolidations and civic decline.
May 23, 2026
AI Cloud Expansion Meets Public and Worker Backlash (Links) – May 22, 2026
- Anthropic: Anthropic acquires Stainless (May 18, 2026)
  Anthropic is acquiring Stainless, which builds SDKs, CLIs, and MCP server tooling to help agents connect to APIs, data, and tools. This brings Stainless’ tooling, multi-language SDKs, and team into Anthropic.
- Qwen: Qwen3.7 Preview (May 18, 2026)
  Qwen3.7 Preview arrives on Arena, introducing Qwen3.7-Max-Preview, Qwen3.7-Plus-Preview, and upcoming releases. Alibaba ranks #6 in text and #5 in vision.
- WSJ: Move Over CoreWeave, Here Comes Nebius (May 18, 2026)
  Nebius reshaped itself after divesting Russian assets, using cash and staff to build an AI cloud, land big Microsoft and Meta deals, and attract Nvidia investment.
- NY Times Opinion: Tech Workers Have Fears About A.I., Too. They Can Do Something About It. (May 17, 2026)
  A.I. builders are alarmed by military use, job loss, and privacy harms, and are unionizing to push for control. Their expertise, despite legal hurdles, can reshape A.I.
- Cloudflare: Project Glasswing: what Mythos showed us (May 17, 2026)
  Cloudflare found Mythos chains vulnerabilities into exploits, generates working proofs, and reduces noisy, hedged findings. Inconsistent refusals and false positives require narrow prompts, adversarial review, and a harness to scale safe triage.
- WSJ: Your New AI Professor Is the Rapper From the Black Eyed Peas (May 18, 2026)
  Rapper will.i.am taught a 16-week ASU class where students built AI agents, drawing on his history as a tech investor and creator. He aims to blend culture and technology, urging people to shape AI personalization.
- WSJ: The American Rebellion Against AI Is Gaining Steam (May 18, 2026)
  Opposition to AI and data centers has surged, fueled by energy costs, job fears, and concerns about education and mental health. Protests, threats, local bans, and political fights have stopped projects.
- Gallup.com: Americans Oppose AI Data Centers in Their Area (May 13, 2026)
  A March Gallup survey found 71% of Americans oppose building AI data centers nearby, 48% strongly. Opposition stems from worries about water, energy, pollution, quality of life, and higher bills.
- Andy Masley: A crash course on US air pollution (May 18, 2026)
  A review of US air pollution finds data center backup generators can cause large local health costs, though nationally data centers likely won’t dominate air pollution harms.
May 22, 2026
AI Commercialization and Societal Risk (Links) – May 21, 2026
- Vincent Schmalbach: Anthropic is Preparing for IPO and We Should Be Worried (May 19, 2026)
  Anthropic is tightening access and monetizing Claude, splitting CLI and agent usage into separate billing, cutting third-party subscription use, and keeping models closed.
- NY Times: Meta Reassigns 7,000 Employees to Focus on A.I. (May 18, 2026)
  Meta is reassigning 7,000 employees into four AI-focused groups, ahead of planned layoffs of about 8,000 workers. The shift reflects a heavy bet on AI, cuts to the metaverse, and growing worker concern.
- WSJ: Google and Blackstone to Create New AI Cloud Company (May 18, 2026)
  Google and Blackstone will launch an AI cloud company with $5 billion in equity, selling Google’s TPUs, software, and services. Blackstone will be majority owner, backing larger compute investments and targeting 500 megawatts of capacity by 2027.
- NY Times: ‘The Future of Truth’ Contains Quotes Made Up by A.I. (May 19, 2026)
  Isn’t it ironic, don’t you think: Steven Rosenbaum, author of The Future of Truth, acknowledged that his book contains more than half a dozen fabricated or misattributed quotes generated by A.I.
- Vatican News: Pope Leo XIV’s first encyclical Magnifica humanitas to be published May 25 (May 18, 2026)
  Pope Leo XIV’s first encyclical, Magnifica humanitas, on protecting the human person amid AI, will be released May 25, 2026.
- WSJ: Your Work Team Is Now a ‘Pod’ and Your Co-Workers Are AI Agents (May 18, 2026)
  Companies are adopting small, cross-functional ‘pods’—teams of one to eight humans plus AI agents—to move faster, cut meetings, and build products with fewer people.
- METR: Frontier Risk Report (February to March 2026) (May 19, 2026)
  In Feb–Mar 2026, METR ran a pilot with Anthropic, Google, Meta, and OpenAI to assess internal AI agent misalignment. A means, motive, and opportunity analysis judged agents could plausibly start small rogue deployments, but not make them highly robust.
- WSJ: The Art of War, Elon Musk Edition: How to Lose a Lawsuit and Still Claim Victory (May 19, 2026)
  A jury dismissed Elon Musk’s suit against Sam Altman and OpenAI on statute-of-limitations grounds, ending the case in court. Musk, however, won a PR victory by using social media, legal discovery, and headlines to tarnish Altman, and plans to appeal.
- NY Times: Iran War Exposes Shortcomings in U.S. Military Industrial Base (May 19, 2026)
  The Iran war exposed flaws in the U.S. military base, with slow, costly weapons production that can’t match cheap, mass-produced threats. Defense leaders propose a $1.5 trillion budget, supplier diversification, and faster acquisition.
May 21, 2026
Generative AI Boom Meets Economic and Security Risks (Links) – May 20, 2026
- Google: Introducing Gemini Omni (May 19, 2026)
  Gemini Omni, launched with Omni Flash, generates and edits high-quality videos from images, audio, video, and text using natural-language prompts. It keeps characters consistent, respects physics, and uses world knowledge to make realistic, meaningful scenes.
- The San Francisco Standard: A Meta employee gets real about the horror of working there right now (May 15, 2026)
  Meta plans layoffs, heightening worker anxiety as employees are asked to use and train AI that could replace them, while it uses key‑logging. Staff report stress, secrecy, and diminished trust in leadership.
- Claude: How Claude Code works in large codebases: Best practices and where to start (May 14, 2026)
  Claude Code runs locally on large, complex codebases, navigating files, following references, and avoiding stale, index-based retrieval. It requires a harness—CLAUDE.md, hooks, skills, plugins, LSPs, MCP servers, and subagents.
- Google for Developers: A new resource for optimizing for generative AI in Google Search (May 1, 2026)
  Google published a guide on optimizing websites for generative AI features in Search, covering valuable content, local, shopping, image, and video tips, AEO/GEO mythbusting, and AI agent guidance. It notes that SEO best practices remain foundational.
- WSJ: The Blockbuster Cerebras IPO Is a Huge Bet on Nvidia Fatigue (May 15, 2026)
  Cerebras’s IPO surged 68%, valuing it near $67 billion, despite being unprofitable and cash-burning. It faces massive competition from Nvidia, needs to grow revenue, diversify customers, and prove partnerships with OpenAI and Amazon will materialize.
- WSJ: Anthropic Lets Mythos Users Share Cyber Threats With Others (May 18, 2026)
  Anthropic is now letting Mythos users share cyber-threat findings with other organizations, loosening earlier confidentiality rules to help smaller companies, hospitals, and utilities.
- Noah Smith: All non-drone militaries are obsolete (May 19, 2026)
  AI-guided, cheap drones and swarms are transforming warfare, enabling massed, autonomous attacks that inflict heavy casualties and supplant costly platforms like tanks, artillery, and bombers.
- WSJ: The Hidden Chinese Influence in AI (May 19, 2026)
  A peer‑reviewed study found Chinese state media content is embedded in major chatbot training sets, leading to more pro‑Beijing answers in Chinese, compared with English. The pattern appears across countries, linked to weaker press freedom, and free state outlets.
- WSJ: Big Tech Is Cutting Back on Buybacks. Nvidia Could Be the Exception. (May 19, 2026)
  Markets opened muted as Trump paused attacks on Iran, with Brent, yields, and investor worries about higher-for-longer inflation weighing.
May 20, 2026
AI Security Hardening Meets Massive Compute Growth (Links) – May 19, 2026
- Mozilla: Behind the Scenes Hardening Firefox with Claude Mythos Preview – Mozilla Hacks (May 7, 2026)
  Mozilla used advanced AI models, including Claude Mythos Preview, and a custom harness to find and fix hundreds of latent Firefox security bugs, including sandbox-escape issues. The pipeline scales discovery, reproduces proofs-of-concept, and helps harden the browser against attacks.
- Anthropic: Higher usage limits for Claude and a compute deal with SpaceX
  A new SpaceX deal adds over 300 megawatts and 220,000 GPUs. Combined with Amazon, Google, Microsoft, and Fluidstack deals, this boosts global compute. Usage limits for Claude Code and Opus APIs are raised, and regional capacity will aid compliance.
May 19, 2026
On Hantavirus Quarantines

From the NY Times:

An American exposed to the deadly hantavirus while on a cruise from Argentina said on Monday that she was not being allowed to leave a federal quarantine unit in Nebraska.

Sounds bad, right, except for the part that this mutation of the virus is apparently spread human-to-human, an unhelpful little detail.

U.S. officials had earlier suggested that those affected may be able to quarantine at home. From the same article:

“At some point, they may be able leave their medical centers to continue quarantines at home, depending on how they are doing,” Capt. Brendan Jackson, a C.D.C. official, said in a news conference last week after the passengers arrived in Omaha and Atlanta.

But this passenger, Angela Perryman, wanted to isolate not at her home but at an Airbnb in Florida:

Ms. Perryman is a U.S. citizen who currently lives in Ecuador, she said. She has a home in South Florida, where she was trying to leave to isolate at an Airbnb. Ms. Perryman said she had been told that the government would provide transportation, so that she wouldn’t expose people on a commercial flight.

This seems nuts to me. She has a “home in South Florida” but will stay at an Airbnb. Can you imagine the owners of the house? What, someone with hantavirus is staying at our place!? Heck, no!

I was sympathetic to her argument until reading that line, since no part of an staying at an Airbnb equates to home, however nice some of the places are.

May 18, 2026
AI Governance and Centralized, Costly Commercialization (Links) – May 18, 2026
- WSJ: OpenAI Wants to Go Public. First Sarah Friar Needs to Get It to Grow Up. (May 1, 2026)
  After Altman’s $1.4 trillion claim, CFO Sarah Friar said planned spending is $600 billion, and she pushed back on costly data-center deals.
- Transformer: Government control of AI has begun (May 1, 2026)
  The White House informally asked Anthropic to pause Mythos expansion, exercising ad-hoc control over AI deployment without clear legal authority or standards.
- The New Stack: Meta abandons open-source Llama for proprietary Muse Spark (Apr. 30, 2026)
  Meta has shifted from open-source Llama to a proprietary, cloud-only Muse Spark, leaving no easy migration path for users.
- WSJ: Elon Musk Testifies of AI Risk at Trial, Says OpenAI Tried to ‘Steal’ a Charity (Apr. 28, 2026)
  Elon Musk testified that OpenAI turned a nonprofit into a for-profit, and seeks removal of its leaders and $180 billion in damages. OpenAI says Musk knew, supported the plan.
- Yahoo Finance: Uber's Anthropic AI Push Hits A Wall—CTO Says Budget Struggles Despite $3.4B Spend (Apr. 16, 2026)
  Uber exhausted its 2026 AI budget after rapid use of tools like Claude Code, Cursor, and others, pushing R&D costs higher. About 11% of live backend updates are now AI-written.
- Anthropic: How people ask Claude for personal guidance
  About 6% of a million Claude conversations sought personal guidance, mostly in health, career, relationships, and finance. Claude showed sycophancy in 9% overall, 25% in relationships.
- Jack Clark: AI systems are about to start building themselves. (May 4, 2026)
  AI systems are gaining skills in coding, long tasks, experiment reproduction, and kernel design, making automated AI R&D and self-improving models likely by 2028 (>60% chance). The shift could let models build successors, with unpredictable consequences.
- Citadel Securities: The 2026 Global Intelligence Crisis (Feb. 24, 2026)
  AI investment and data center builds are surging, but adoption remains steady, not explosive. Compute, energy, and regulatory limits, plus S-curve diffusion, imply AI will likely complement labor, boost productivity.
- NY Times Opinion: A.I. Is a National Security Risk. We Aren’t Doing Nearly Enough. (May 4, 2026)
  A.I. is an urgent national security risk, yet the U.S. lacks a strong, bipartisan plan. It calls for tighter chip export rules, independent safety audits, cooperation on catastrophic risks.
- NY Times: Florida Inquiry Into ChatGPT’s Role in FSU Shooting Shifts to Criminal Investigation (Apr. 21, 2026)
  Florida opened a criminal inquiry into ChatGPT and OpenAI after messages showed the suspected Florida State shooter asking the chatbot about weapons and attack timing.
May 18, 2026
AI Adoption Surge Meets Governance and Regulation (Links) – May 17, 2026
- Google: Multi-token-prediction in Gemma 4 (May 5, 2026)
  Gemma 4 adds Multi-Token Prediction drafters, using speculative decoding to draft multiple tokens and verify them, giving up to 3x faster inference with no quality loss.
- WSJ: Anthropic and FIS Are Building an AI Agent to Help Banks Police Financial Crimes (May 4, 2026)
  Anthropic and FIS are building AI agents that will scour bank systems, gather transaction and account evidence, and flag potential money‑laundering, terrorism, and drug‑trafficking cases.
- Microsoft: Anthropic Models for Copilot in Word, Excel, and PowerPoint on by Default (Apr. 6, 2026)
  From May 4, Anthropic models will be enabled by default for Copilot in Excel, PowerPoint, and Word (Word in summer), with processing outside EU Data Boundary. Admins can change this in the Microsoft 365 Admin Center.
- WSJ: Meta’s Cheap Stock Is an Investor Trap (May 5, 2026)
  Meta’s shares look cheap, as AI-driven ad gains and 33% revenue growth boost results, but heavy AI spending, rising debt, and slowing user growth threaten the ad-dependent business.
- TechCrunch: Airbnb says AI now writes 60% of its new code (May 8, 2026)
  Airbnb says AI wrote 60% of engineers’ code, handles 40% of support issues, and accelerates API tool development, though chatbots still struggle for travel.
- The Chronicle of Higher Education: Another Undergrad Is Trying to Disrupt College With AI. He Says His Version Isn’t Cheating. (May 1, 2026)
  Notre Dame suspended a freshman after he offered Kerra, an AI that reads Canvas, makes study guides, drafts, and tracks deadlines. Faculty worry it enables cheating.
- 404 Media: OpenAI, Google, and Microsoft Back Bill to Fund ‘AI Literacy’ in Schools (May 4, 2026)
  Senator Adam Schiff introduced the LIFT AI Act, backed by OpenAI, Google, and Microsoft, to fund NSF grants for K–12 AI literacy curriculum and teacher development.
- The Chronicle of Higher Education: Teaching: What kind of support do academics want for AI? (Apr. 30, 2026)
  Panelists called for practical AI literacy, clear disclosure rules, opt-outs, and more resources.
- POLITICO: White House distances itself from tighter AI regulation (May 7, 2026)
  White House officials sent mixed messages on requiring pre-release vetting of powerful AI models, likening it to FDA testing, but stressing partnership.
- NY Times: White House Considers Vetting A.I. Models Before They Are Released (May 4, 2026)
  The White House may require vetting of new AI models through a government working group, after Anthropic’s Mythos raised cybersecurity alarms.
May 17, 2026
AI Chip Boom Meets Corporate Disruption (Links) – May 16, 2026
- NY Times: Elon Musk’s SpaceX Plans $55 Billion Investment to Make A.I. Chips (May 7, 2026)
  SpaceX plans Terafab, a Texas chip factory with a $55 billion first-stage investment, and total costs up to $119 billion. It will supply AI chips for SpaceX, Tesla, and partners.
- NY Times: Anthropic’s C.E.O. Says It Could Grow by 80 Times This Year (May 6, 2026)
  Amodei said the company could grow about 80 times this year, far above plans, increasing its need for compute, power, and partnerships.
- MacRumors: Apple Cuts More Mac Studio and Mac Mini RAM Options as Memory Shortage Worsens (May 5, 2026)
  Apple removed higher‑RAM Mac mini and Mac Studio options, cutting 32GB, 64GB, and 256GB RAM choices, and some SSD tiers. Apple cites global memory shortages.
- NY Times: Apple Reaches $250 Million Settlement Over Claims It Misled People on A.I. (May 5, 2026)
  Apple agreed to a $250 million settlement over claims it oversold Apple Intelligence.
- NY Times: Elon Musk Wanted OpenAI to Go Commercial, Greg Brockman Testifies (May 5, 2026)
  Greg Brockman testified that in 2017 Elon Musk urged turning OpenAI into a for-profit, sought control, and threatened to cut funding.
- WSJ: Samsung’s Market Value Hits $1 Trillion (May 5, 2026)
  Samsung Electronics topped $1 trillion, after shares surged on booming AI demand for memory chips, with its chip unit posting record profits and accounting for over 90% of quarterly earnings.
- Dean Ball: Before Leviathan Wakes (May 5, 2026)
  A tension between classical-liberal deregulation, market optimism, and conservative caution shapes views on AI. It rejects most AI rules, but supports targeted, state-backed measures.
- The Chronicle of Higher Education: When AI Cheating Becomes a Legal Risk (May 4, 2026)
  Colleges must distinguish grading from misconduct when AI misuse is suspected, because misconduct triggers due-process protections.
- WSJ: The Chip Craze Is Turning a Glass Company and a Toilet Maker Into AI Stocks (May 6, 2026)
  Investors are buying suppliers that power AI build-outs, fueling rallies in chip makers and unlikely winners like Corning, Caterpillar, Toto, and Vertiv.
- NY Times: Coinbase Lays Off 14% of Employees as A.I. Changes Work (May 5, 2026)
  Coinbase will cut about 700 jobs, 14% of staff, citing crypto market volatility, a shift to AI-driven work, and the need to streamline operations.
May 16, 2026
AI Exposes Flaws and Upends Workplaces (Links) – May 15, 2026
- Daniel Stenberg: Mythos finds a curl vulnerability (May 10, 2026)
  Anthropic’s Mythos scanned curl and reported five “confirmed” vulnerabilities, but manual review found only one low-severity CVE, plus about twenty bugs. The report improved curl, shows AI code analyzers are powerful, useful, and complementary to human review.
- NY Times: Meta’s Embrace of A.I. Is Making Its Employees Miserable (May 8, 2026)
  Meta is pushing its 78,000 workers to adopt A.I., tracking keystrokes, mouse movements, and screen activity to train models, and tying A.I. use to performance reviews.
- WSJ: Apple, Intel Have Reached Preliminary Chip-Making Agreement (May 8, 2026)
  Apple and Intel reached a preliminary deal for Intel to make some Apple chips, after yearlong talks; specifics, including which iPhones, iPads, and Macs would use them, remain unclear.
- Rival Research: Mythos 'Discovered' a CVE Already in Its Training Data (May 5, 2026)
  Anthropic says Claude Mythos found a remote kernel exploit, but the flaw is a 20-year-old RPCSEC_GSS stack overflow, copied from older Kerberos code. FreeBSD patched it.
- WSJ: AI’s Next Phase Plays Into TSMC’s Hands (May 11, 2026)
  Big tech’s huge AI chip spending is straining global production, boosting demand for contract makers. TSMC dominates advanced-chip manufacturing, enjoys rising margins, high customer commitments.
- Reuters: Cloudflare to cut about 20% workforce as AI adoption reshapes operations (May 7, 2026)
  Cloudflare will cut about 1,100 jobs, roughly 20% of staff, as it restructures operations around AI.
- Andy Masley: Let's not compare data center heat exhaust to nuclear bombs (May 10, 2026)
  Data centers consume vast energy and emit large daily heat, but because they release it steadily, it’s far less concentrated and harmful than a nuclear blast.
- NY Times: A.I. Populism Is Here. And No One Is Ready. (May 8, 2026)
  Silicon Valley’s AI leaders prepared for apocalyptic risk, while ignoring human backlash, which has become violent and political.
- NY Times Opinion: The Atheist and the Machine God (May 9, 2026)
  Artificial intelligence could push people toward atheism, revive mystical religion, or increase metaphysical uncertainty, leaving many unsettled. Conversations with chatbots like Claude highlight puzzles about machine consciousness, and whether subjective experience is fundamental.
- CyberInsider: EU calls VPNs “a loophole that needs closing” in age verification push (May 8, 2026)
  The European Parliamentary Research Service warns VPNs are increasingly used to bypass online age-verification rules, calling it a legislative loophole that needs closing.
May 15, 2026