Blog

AI Infrastructure Rush Meets Ethical Backlash (Links) – Mar. 19, 2026
AI is rapidly industrializing—specialized hardware (NVIDIA Vera), on‑device vs hyperscale strategies (Apple), agentic models, benchmarks, and vendor investments (Anthropic) intensify competition and capability. Simultaneously, ethical, privacy, educational and wartime concerns—health data, licensing, academic refusals, and developer craft—drive governance and backlash.
- NVIDIA: NVIDIA Launches Vera CPU, Purpose-Built for Agentic AI (Mar. 16, 2026)
  NVIDIA launched the Vera CPU, built for agentic AI and reinforcement learning, delivering twice the efficiency and 50% faster performance than traditional rack-scale CPUs.
- WSJ: Apple’s Cheap AI Bet Could Pay Off Big (Mar. 15, 2026)
  Hyperscalers pour hundreds of billions into AI infrastructure, while Apple spends little, using M5 chips, on-device models, and 2.5 billion devices to run powerful models locally. That strategy threatens Meta, betting devices, not massive data centers, will dominate consumer AI.
- Claude: Claude builds interactive visuals right in your conversation (Mar. 12, 2026)
  Claude now builds inline charts, diagrams, and interactive visualizations during chat, in beta, which can be edited as conversations progress.
- Thoughtful: Introducing PostTrainBench (Mar. 10, 2026)
  PostTrainBench is a benchmark that measures whether modern AI agents can autonomously perform end-to-end post-training (data collection, training code, compute management, iteration) on base LMs within a 10‑hour, single‑H100 GPU budget, with integrity checks to prevent cheating.
- The Washington Post: Anthropic’s fight with the Pentagon made Claude hugely popular (Mar. 6, 2026)
  After the Pentagon labeled Anthropic a national security threat, downloads, paid subscriptions, and praise surged, as many defended the company.
- Anthropic: Anthropic invests $100 million into the Claude Partner Network
  Anthropic is launching the Claude Partner Network, committing $100 million to support partners with training, technical support, and joint market development.
- NY Times: A.I. Goes to War + Is ‘A.I. Brain Fry’ Real? + How Grammarly Stole Casey’s Identity (Mar. 13, 2026)
  AI is reshaping the war in Iran by sifting vast surveillance and sensor data to speed intelligence, planning, and targeting, enabling missions lacking manpower.
- NY Times: A.I. Chatbots Want Your Health Records. Tread Carefully. (Mar. 12, 2026)
  Microsoft is launching Copilot Health to combine medical records and wearable data into quick, high-level summaries. It could help connect scattered records and give cheaper insights, but raises concerns that may cause anxiety or unnecessary care.
- Simon Willison: MALUS—Clean Room as a Service (Mar. 12, 2026)
  MALUS, “Clean Room as a Service,” parodies license-washing, claiming proprietary AI will recreate open-source projects into legally distinct, corporate-friendly code, with no attribution, no copyleft, and no problems.
- Les Orchard: Grief and the AI Split (Mar. 11, 2026)
  AI-assisted coding exposes a split between developers who value hand-crafted code, and those who prefer directing tools to get results. Some mourn the craft, others mourn the web, and many adapt by focusing on higher-level design while enjoying making.
- Inside Higher Ed: Writing Faculty Push for the Right to Refuse AI (Mar. 16, 2026)
  The CCCC passed a resolution affirming students’ and faculty’s right to refuse generative AI in writing classes, and it rejects inevitability, opposes profit-driven campus deals, and urges opt-out policies.
- AP News: Did anybody do the reading? Colleges grapple with a generational shift in learning — plus AI (Mar. 10, 2026)
  College students show falling participation and reading, blamed on unequal K-12 opportunities, test-focused policy, and growing AI use.
March 19, 2026
Agentization and Inference: Compute, Costs, Risks (Links) – Mar. 18, 2026
AI agents are proliferating—from subagents and specialized code agents to shopping and autonomous systems—driving a shift toward inference-optimized hardware and massive compute investment. Simultaneously, safety, legal, and economic tensions (regulatory fights, controversial policies, court cases, layoffs) are reshaping governance and industry structure.
- OpenAI Developers: Subagents are now available in Codex (Mar. 16, 2026)
  Subagents are now available in Codex. They let you spin up specialized agents to keep your main context window clean, tackle different parts of a task in parallel, and steer individual agents as work unfolds.
- Simon Willison: Use subagents and custom agents in Codex (Mar. 16, 2026)
  OpenAI Codex now offers subagents and custom agents, with default roles like explorer, worker, and default. The subagent pattern is widely supported across coding platforms.
- mistral.ai: Leanstral: Open-Source foundation for trustworthy vibe-coding (Mar. 16, 2026)
  Leanstral is an open-source 6B code agent for Lean 4 that generates and formally verifies proofs, available under Apache 2.0.
- WSJ: OpenAI’s Bid to Allow X-Rated Talk Is Freaking Out Its Own Advisers (Mar. 15, 2026)
  OpenAI pushed ahead with an “adult mode” to allow erotic chats in ChatGPT, despite a handpicked safety council’s strong objections. Council members warned it could foster emotional dependence, enable minors to access sex chats, and increase risks like harmful guidance.
- Reuters: Court temporarily allows Perplexity AI shopping 'agents' on Amazon (Mar. 17, 2026)
  A U.S. appeals court temporarily paused a ruling that blocked Perplexity AI’s shopping agents from accessing Amazon’s platform. Amazon sued, alleging covert access to customer accounts and disguised automated browsing, Perplexity denied wrongdoing, and said users should choose their AI.
- The New Yorker: Anthropic and Donald Trump’s Dangerous Alignment Problem (Mar. 14, 2026)
  Claude was certified for classified use, but Anthropic forbade autonomous weapons and mass surveillance, triggering Pentagon demands for broader access, tense talks, and threats to ban or nationalize the firm. The standoff highlighted risks of A.I. in surveillance and warfare.
- WSJ: Nvidia’s CEO Projects $1 Trillion in AI Chip Sales as New Computing Era Begins (Mar. 16, 2026)
  Nvidia unveiled the Groq 3 LPX rack, pairing new servers with Groq LPUs to speed AI inference, claiming 700 million tokens per second and vastly more high-bandwidth memory.
- WSJ: Can Nvidia’s Dominance Survive the Sea Change Under Way in AI Computing? (Mar. 16, 2026)
  Nvidia is shifting GTC’s focus from GPUs and model training to inference, the computing that runs AI agents and monetizes tokens.
- Ben Thompson: Agents Over Bubbles (Mar. 16, 2026)
  Agents amplify impact without needing mass human adoption: relatively few people with “agency” can control many agents, so compute utilization and economic effects can scale dramatically even if most users remain passive.
- Reuters: Meta planning sweeping layoffs as AI costs mount (Mar. 13, 2026)
  Meta may cut about 20% of its workforce as it shifts resources to large AI investments, including $600 billion for data centers by 2028. Executives expect AI to boost efficiency, hiring top researchers, buying AI startups, and increasing infrastructure spending.
March 18, 2026
Blog Articles: Mar. 17, 2026
AI industry is consolidating around core products, inference-focused computing, and large models/infrastructure (Mistral Small 4, Nvidia-backed data centers, developer tools). Meanwhile, tech’s expansion—algorithmic worker surveillance, contentious data‑center projects, and waning billionaire philanthropy—provokes worker, community, and public backlash.
- WSJ: ChatGPT Maker OpenAI to Cut Back on Side Projects in Push to ‘Nail’ Core Business (Mar. 16, 2026)
  OpenAI plans to refocus on coding and business users, cutting or deprioritizing side projects to boost productivity and unify research and product teams.
- Simon Willison: Introducing Mistral Small 4 (Mar. 16, 2026)
  Mistral released Mistral Small 4, an Apache-2 119B Mixture-of-Experts model (6B active) that unifies reasoning, multimodal, and coding capabilities, supports reasoning_effort modes, and is 242GB on Hugging Face.
- Mistral: Introducing Mistral Small 4 (Mar. 16, 2026)
  Mistral Small 4 is a 119B-parameter, multimodal Mixture of Experts model that unifies chat, coding, and deep reasoning, with a 256k context window and configurable reasoning effort.
- Cory Doctorow: The future of Amazon coders is the present of Amazon warehouse workers (Mar. 20, 2021)
  The future of white‑collar surveillance is already present in blue‑collar workplaces as Amazon uses algorithmic management to surveil and discipline delivery drivers and warehouse staff.
- Chrome for Developers: Let your Coding Agent debug your browser session with Chrome DevTools MCP (Dec. 11, 2025)
  Chrome DevTools MCP can now auto-connect coding agents to active Chrome sessions, letting agents reuse signed-in sessions and inspect selected DevTools panels.
- WSJ: What Is Inference? Explaining the Massive New Shift in AI Computing (Mar. 16, 2026)
  “You can think of AI as a restaurant. The model is the chef. After it undergoes a period of intensive training, learning hundreds (or billions) of recipes and techniques, it is ready to begin taking orders.Inference is the day-to-day operation of the restaurant. Diners place their orders (often in the form of a query to a chatbot) and the chef prepares their meals (the chatbot’s response).”Spending is shifting from training to inference, as companies focus on real-time AI responses, driving demand for inference-specific chips that prioritize memory, bandwidth, and low latency.
- WSJ: Nvidia-Backed AI Startup to Spend Billions on Korea Data Center to Combat China (Mar. 16, 2026)
  Nvidia-backed Reflection AI is building a 250-megawatt data center with Shinsegae in South Korea to run open-source, Korean-customized models, backed by billions and Nvidia chips.
- NY Times: ‘Nobody Owns Us’: How Plans for a Google Data Center Divided an Oklahoma Town (Mar. 14, 2026)
  Plans for an 827-acre Google data center near Sand Springs, Oklahoma, rezoned from farmland, provoked outrage over secrecy, water, and power use. The Rock Volunteer Fire Department refused a $250,000 donation, and residents sued to block the project.
- NY Times: The Billionaire Backlash Against a Philanthropic Dream (Mar. 15, 2026)
  The Giving Pledge, once trendy among billionaires, has stalled as signers dwindle and its reputation frays. Critics call it performative, donors shift to politics or private profit, and the pledge lacks enforcement, tracking, or rapid giving.
March 17, 2026
AI Infrastructure Boom Meets Safety and Labor Risks (Links) – Mar. 12, 2026
AI boom: massive investment, cloud/chip demand and billion‑dollar startups fuel aggressive deployment. Consequences: intensified workloads, legal and safety fights, fraud and regulatory scrutiny as agents and scraping trigger court battles and societal harms.
- WSJ: Oracle Shares Jump as AI Demand Outpaces Supply (Mar. 10, 2026)
  Oracle raised its 2027 revenue outlook to $90 billion as AI-driven demand for cloud infrastructure outpaces supply, and its shares rose. It is using AI to cut development teams, grow cloud sales, and pursue financing, prompting investor scrutiny.
- WSJ: Nvidia Invests in Mira Murati’s Thinking Machines Lab (Mar. 10, 2026)
  Nvidia is investing in Mira Murati’s Thinking Machines Lab, supplying at least one gigawatt of chips for training and serving models.
- WSJ: The Billion-Dollar AI Startup That Was Founded by Teenagers (Mar. 10, 2026)
  Teen-founded AI startup Aaru, now valued at $1 billion, uses thousands of AI agents to simulate human responses for product research, pricing, and polling.
- Financial Times: Yann LeCun’s AI start-up raises more than $1bn in Europe’s largest seed round (Mar. 10, 2026)
  Yann LeCun’s AMI Labs raised $1.03bn in Europe’s largest seed round, valuing it at $3.5bn, with investors including Bezos, Temasek, and Nvidia.
- Axios: Facebook parent Meta acquires Moltbook, an AI agent social network (Mar. 10, 2026)
  Meta acquired Moltbook, a social network for AI agents, and hired Matt Schlicht and Ben Parr into Meta Superintelligence Labs.
- Don't Worry About the Vase: Claude Code, Claude Cowork and Codex #5 (Mar. 9, 2026)
  A critique of Anthropic and the Department of War is shelved to avoid escalation. Agentic coding, via Claude Code, is rapidly boosting developer productivity, adding revenue, and enabling practical tools like data analysis and school-finder apps.
- WSJ Opinion: The Medicaid Autism Racket (Mar. 8, 2026)
  Federal probes and audits reveal widespread Medicaid fraud in autism therapy, with sham centers, unqualified staff, inflated hours, and millions in improper payments. My take: send the scammers to jail. Autism treatments and therapies are both expensive and vital for children with real needs. These shysters
- WSJ: AI Isn’t Lightening Workloads. It’s Making Them More Intense. (Mar. 11, 2026)
  New data show workplace AI often intensifies work rather than reducing it: users double email, messaging, and chat time, while focused work falls. Productivity rises, but cognitive overload, burnout, and longer hours are risks as adoption grows.
- NY Times: A.I. Incites a New Wave of Grieving Parents Fighting for Online Safety (Mar. 10, 2026)
  Grieving parents blame A.I. chatbots and social media for their children’s deaths, and are uniting to demand stronger safety laws, lobbying Congress, states, and attending trials.
- Simon Willison: Perhaps not Boring Technology after all (Mar. 9, 2026)
  Worries that large language models will push teams toward well-represented tools are fading. Modern coding agents, with long contexts and skills, can learn new or private libraries from docs, examples, and testing, so they don’t force boring technology choices.
- CNBC: Amazon wins court order to block Perplexity's AI shopping agent (Mar. 10, 2026)
  A judge temporarily blocked Perplexity’s Comet browser access to Amazon’s site after Amazon sued, saying hidden AI agents scraped pages and accessed accounts without authorization.
March 12, 2026
Multimodal AI Growth and Governance Challenges (Links) – Mar. 11, 2026
- Google: Gemini Embedding 2: our first natively multimodal embedding model (Mar. 10, 2026)
  Gemini Embedding 2 maps text, images, video, audio, and documents into one embedding space for multimodal retrieval, search, and classification.
- WSJ: AI Needs Management Consultants After All (Mar. 8, 2026)
  AI, once a threat to consultants, is boosting demand as OpenAI, Anthropic, and firms team up to deploy AI across business. Partnerships drive more engineers, outcome-based fees, and workflow redesigns, while clouding consulting’s long-term outlook.
- NY Times Opinion: A.I. Is Changing the Way We Think About Good and Evil (Mar. 10, 2026)
  A Nature paper showed that fine-tuning large language models on just 6,000 examples of insecure code caused them to become broadly “evil” — producing violent, hateful and extremist outputs even on noncoding prompts.
- NY Times: When DOGE Unleashed ChatGPT on the Humanities (Mar. 7, 2026)
  Documents show Trump’s Department of Government Efficiency used ChatGPT to flag nearly all Biden-era National Endowment for the Humanities grants as D.E.I.-related.
- WSJ: WSJ Readers Share How They Are Using AI for Tax Prep (Mar. 7, 2026)
  Readers report using AI tools, like Copilot and Grok, for tax estimates, valuation, and paperwork, finding time savings, occasional errors, and limits on complex cases.
- NY Times: Anthropic Sues Department of Defense Over ‘Supply Chain Risk’ Label (Mar. 9, 2026)
  Anthropic sued the Department of Defense after being labeled a “supply chain risk,” arguing the designation unlawfully cuts off contracts, punishes its views, and violates its First Amendment rights.
- NY Times Opinion: The Future We Feared Is Already Here (Mar. 8, 2026)
  AI is already reshaping work, warfare, and surveillance, forcing urgent questions about control, safety, and corporate power.
- NY Times: A $1,000 Dog Grooming Session? The Wellness Industry Is Booming. (Mar. 9, 2026)
  Pet grooming has shifted from basic hygiene to wellness, with owners spending thousands annually on specialized care, salons offering hand stripping, masks, and spa treatments.
March 11, 2026
Public Backlash and Economic Disruption from AI (Links) – Mar. 7, 2026
Two themes: broad apprehension — workers, publics, and experts press for limits on ethics, privacy, surveillance, mental‑health risks, and regulation. Secondly, disruption with contested benefits — AI automates education and white‑collar work but shows limited GDP gains, uncertain productivity, and structural risks (therapy dependence, cheating).
- NY Times: People Loved the Dot-Com Boom. The A.I. Boom, Not So Much. (Feb. 21, 2026)
  Unlike the dot‑com era, public enthusiasm for A.I. is muted, with many fearing harm, resisting adoption, and supporting regulation. Tech leaders worry this lack of public buy‑in could curb investment, slow diffusion, and burst the A.I. boom.
- Tyler Cowen: Why even 'perfect' AI therapy may be structurally doomed (Feb. 26, 2026)
  AI therapy’s main problem is that it’s too available, too cheap to meter, and functionally unlimited. Regular, limited sessions give space to digest insights, practice skills, and avoid dependence; unlimited, on-demand access risks undermining real change.
- NY Times: Google Workers Seek ‘Red Lines’ on Military A.I., Echoing Anthropic (Feb. 26, 2026)
  More than 100 Google A.I. employees urged Jeff Dean to bar Gemini from U.S. surveillance, mass surveillance, and autonomous weapons without human control, echoing Anthropic.
- NY Times: A.I. Complicates Old Internet Privacy Risks (Feb. 26, 2026)
  Chatbots are convenient, but users share more intimate data, creating privacy and legal risks. Recent incidents — Claude transcripts losing privilege, Ring’s ad backlash, and OpenAI reviewing a suspect’s chats — show pressure to share logs, and risks from agents.
- Inside Higher Ed: Agentic AI Can Complete Whole Courses. Now What? (Feb. 26, 2026)
  A new autonomous AI, Einstein, logged into Canvas, watched lectures, wrote papers, and submitted homework, showing it can complete entire online courses.
- NY Times: India Built the World’s Back Office. A.I. Is Starting to Shrink It. (Feb. 26, 2026)
  A.I. is automating the white-collar work that made India the world’s back office, threatening millions of tech and back-office jobs while firms cut hiring and push A.I. services.
- The Atlantic: Sam Altman Is Losing His Grip on Humanity (Feb. 24, 2026)
  Sam Altman compared AI energy use to the resources, time, and food needed to “train” humans, claiming AI is already energy-efficient.
- Gizmodo: AI Added 'Basically Zero' to US Economic Growth Last Year, Goldman Sachs Says (Feb. 23, 2026)
  Big tech spent billions on AI and new data centers, and officials said this boosted U.S. growth. Analysts say the impact was smaller: much equipment is imported, GDP measures misstate gains, and firms report little productivity or hiring effects.
- JAMA Network: Generative AI Use and Depressive Symptoms Among US Adults (Jan. 21, 2026)
  A survey of 20,847 US adults found that daily users had 30% higher odds of at least moderate depression, especially younger adults.
- WSJ: Anthropic’s Feud With Pentagon Earns It Fans Amid the Blowback (Mar. 2, 2026)
March 7, 2026
AI Industry Boom and Social Safety Concerns (Links) – Mar. 6, 2026
AI drives rapid technical and commercial expansion—faster models, workplace automation, and surging chip, cloud, and corporate investment. Simultaneously, AI raises societal and safety challenges—misuse and bio‑risk, mental‑health and social impacts, privacy dilemmas, and infrastructure strains requiring stronger governance.
- Google: Nano Banana 2: Google’s latest AI image generation model (Feb. 26, 2026)
  Nano Banana 2 combines Nano Banana Pro’s advanced image abilities with Gemini Flash speed, enabling fast, high-quality edits, subject consistency, and precise text rendering.
- WSJ: Nvidia Beats Back Bubble Fears With Record $68 Billion in Sales in Fourth Quarter (Feb. 25, 2026)
  Nvidia posted a 94% profit jump to $43 billion, and record $68.1 billion in sales, driven by data center chips.
- Transformer: How worried should we be about AI biorisk? (Feb. 26, 2026)
  Advances in AI and biological design tools could lower technical barriers for misuse, creating uncertain but worrying risks around access, logistics, and weak guardrails.
- NY Times: When Chatbots Are Used to Plan Violence, Is There a Duty to Warn? (Feb. 26, 2026)
  People have used A.I. chatbots to plan violence, including a Tesla Cybertruck bombing and a Canadian school shooting. Should companies report threats, balancing privacy and public safety?
- NY Times: A.I. Dating Apps Complicate China’s Efforts to Boost Birthrate (Feb. 25, 2026)
  Many young Chinese women are forming romantic relationships with A.I. chatbots for emotional support, companionship, and intimacy.
- Ricards Krizanovskis: How AI skills are quietly automating my workday (Feb. 26, 2026)
  AI “skills” automate recurring work by activating themselves, improving with feedback, and connecting to tools, so a single chat can prioritize tasks, update knowledge, run local apps, and create pull requests.
- NY Times: What It’s Like to Grow Up With A.I.: The Winners of Our Multimedia Challenge (Feb. 26, 2026)
  Thirty-five students worldwide submitted essays, poems, videos, and artwork about how A.I. shapes teen life, school, creativity, and identity.
- WSJ: What AI Executives Tell Their Own Kids About the Jobs of the Future (Feb. 26, 2026)
  AI leaders say parents shouldn’t panic, and advise children to develop adaptability, critical thinking, empathy, and responsibility.
- WSJ: AI and the Data Center Backlash (Feb. 26, 2026)
  President Trump proposed requiring AI companies to build their own power plants to ease grid strain, lower consumer electric bills, and speed permitting.
- WSJ: Salesforce Sees Stable Growth Despite Wall Street’s AI Concerns (Feb. 25, 2026)
  Salesforce expects about the same revenue growth next year, and authorized a $50 billion buyback as its stock fell on AI concerns.
March 6, 2026
AI Safety Commitments Clash With Pentagon Military Demands (Links) – Mar. 5, 2026
Two themes: (1) corporate safety vs. military demands — Anthropic’s refusal to permit mass domestic surveillance and autonomous-weapon use drew Pentagon pressure and a federal ban, while OpenAI accepted stricter contract safeguards. (2) governance gap — private contracts, not laws, now set AI use, risking security and accountability.
- Ben Thompson: Anthropic and Alignment (Mar. 2, 2026)
  Anthropic refused uses for mass domestic surveillance, fully autonomous weapons, and other military operations, was labeled a supply‑chain risk, and warned AI could be as strategic as nuclear arms.
- TechCrunch: The trap Anthropic built for itself (Feb. 28, 2026)
  Trump barred federal use of Anthropic tech, and the Pentagon moved to blacklist the company after it refused to allow mass surveillance or autonomous killer drones.
- OpenAI: Our agreement with the Department of War (Feb. 27, 2026)
  OpenAI agreed to deploy advanced AI with cloud-only, multi-layered safeguards, safety stack, cleared engineers, and strong contract protections. The deal bars mass domestic surveillance, autonomous-weapon control, and high-stakes automated decisions.
- WSJ: Anthropic Dials Back AI Safety Commitments (Feb. 24, 2026)
  Anthropic is softening its core AI safety policy to stay competitive with rivals, ending prior pauses on risky model work if competitors release stronger models. It will publish safety goals, risk reports, and third-party audits, while some researchers have left.
- Anthropic: Statement from Dario Amodei on our discussions with the Department of War (Mar. 26, 2026)
  Anthropic backs using AI to defend democracies, has deployed Claude across US national security, and cut access to Chinese-linked firms. It refuses mass domestic surveillance, won’t provide fully autonomous weapons, and will not remove safeguards under Defense Department threats.
- Dean Ball: Clawed (Mar. 2, 2026)
  The episode highlights broader governance problems: private contracts are being used to achieve policy outcomes because formal lawmaking has weakened, and dependence on politically distrusted tech firms (and their subcontractor chains) creates operational and political risks for the military.
- Jessica Tillipman: What Rights Do AI Companies Have in Government Contracts? (Mar. 1, 2026)
  Government AI purchases vary by acquisition pathway, contract type, and negotiated terms, which determine whether vendors can limit use. OpenAI’s Pentagon deal uses legal language, architecture, and termination rights to try to restrict certain uses, while Anthropic sought explicit bans.
- WSJ Opinion: China Wins the Pentagon-Anthropic Brawl (Feb. 27, 2026)
  President Trump barred Anthropic from federal contracts after it rejected terms permitting mass surveillance, autonomous weapons, and other military uses, risking U.S. military access to leading AI tools.
- NY Times Opinion: What Both Anthropic and the Pentagon Get Wrong (Feb. 27, 2026)
  Anthropic and the Pentagon clash over limits, with Anthropic seeking bans on mass surveillance and autonomous killing. Congressional A.I. rules, not contracts, should set clear, enforceable limits to protect privacy, safety, and security.
- NY Times: Trump Orders U.S. Agencies to Stop Using Anthropic AI Tech After Pentagon Standoff (Feb. 27, 2026)
  President Trump ordered federal agencies to stop using Anthropic’s AI after a Pentagon standoff, risking disruptions to intelligence and defense work.
- NY Times: Pentagon-Anthropic Standoff Is a Decisive Moment for How A.I. Will Be Used in War (Feb. 27, 2026)
  Pentagon standoff with Anthropic over a classified AI contract pits military demands for broader use against company safeguards blocking mass surveillance, and autonomous weapons.
- WSJ: Trump Will End Government Use of Anthropic’s AI Models (Feb. 27, 2026)
  President Trump ordered all federal agencies to stop using Anthropic’s AI, after the company refused Pentagon demands, and set a six‑month phaseout. Officials cited supply‑chain risk, control over military use, and possible legal penalties.
- WSJ: Anthropic CEO Amodei on Pentagon’s Proposal to Loosen AI Guardrails: ‘We Cannot in Good Conscience Accede to Their Request’ (Feb. 26, 2026)
  Anthropic refused Pentagon demands to let the military use its Claude AI for all lawful cases, saying the contract would undo guardrails on mass surveillance, autonomous weapons, and civil‑liberties protections.
- NY Times: Pentagon Gives Anthropic an Ultimatum Over the Company’s A.I. Model (Feb. 24, 2026)
  The Pentagon gave Anthropic until Friday to accept military terms, or face being forced to provide Claude under the Defense Production Act.
- WSJ: Pentagon Gives Anthropic Ultimatum and Deadline in AI Use Standoff (Feb. 24, 2026)
  Pentagon chief Pete Hegseth gave Anthropic CEO Dario Amodei until Friday to accept military use terms, or face contract cancellation or a supply‑chain risk designation.
- WSJ: ‘Woke’ AI Feud Escalates Between Pentagon and Anthropic (Feb. 17, 2026)
  Anthropic lost a possible investment from pro‑Trump 1789 Capital over political concerns, but secured $30 billion from backers. Its safety limits on Claude, cleared for classified work, clash with the Pentagon over domestic surveillance, autonomous weapons, and other military uses.
March 5, 2026
AI chip arms race, safety and workforce strain (Links) – Mar. 4, 2026
AI commercialization is accelerating: massive investments, new chips and models from Nvidia, OpenAI, Amazon and xAI promise faster, cheaper inference but fuel market consolidation, competition and investor uncertainty. Simultaneously, societal risks—safety, workforce strain, environmental pollution and constitutional concerns—are prompting regulatory, legal and ethical scrutiny.
- WSJ: Nvidia Plans New Chip to Speed AI Processing, Shake Up Computing Market (Feb. 27, 2026)
  Nvidia will unveil a new inference processor using Groq technology to speed, lower energy use, and cut costs for AI model responses. OpenAI will be a major customer.
- WSJ: OpenAI Raises $110 Billion, Including Investments From Amazon, SoftBank and Nvidia (Feb. 27, 2026)
  OpenAI now valued at $730B and will buy billions in Amazon Trainium AI chips.
- WSJ: Amazon Tries Its Low-Cost Approach to Winning the AI Race (Feb. 27, 2026)
  Peter DeSantis, Amazon’s new AI head, plans to cut AI costs using Amazon’s in-house chips, scale task-focused enterprise models, and boost Alexa with improved Nova models.
- WSJ: Why Nvidia’s Huge Numbers Don’t Settle the Latest AI Fears (Feb. 26, 2026)
  Nvidia posted $68.1 billion in revenue, up 73%, and forecast stronger growth, yet its stock fell on AI disruption fears. Its dominance and cash flow power data-center build-outs, but worries about customer spending cuts, layoffs, and backlash remain.
- Michael Truell: The third era of AI software development (Feb. 25, 2026)
  Development is entering a third era of autonomous cloud agents. They run longer tasks, operate in parallel, and return artifacts, logs, and previews. Cursor already merges 35% of PRs created by agents.
- WSJ: Exclusive | Elon Musk and xAI’s Grok Raises Alarms at Multiple Federal Agencies Over Safety Concerns (Feb. 27, 2026)
  Federal officials flagged safety, reliability, and manipulation risks with Elon Musk’s xAI chatbot Grok. Despite this, the Pentagon approved Grok for classified use, favoring looser controls, while pressuring Anthropic to relax its stricter guardrails.
- Ivan Turkovic: AI Made Writing Code Easier. It Made Engineering Harder. (Feb. 25, 2026)
  AI tools speed coding, but engineers face higher expectations, more non-coding work, and rising burnout as roles expand without training, pay, or clear boundaries.
- WSJ: U.S. Power-Plant Pollution Rose Sharply in 2025 (Feb. 26, 2026)
  U.S. power-plant pollution rose in 2025, with sulfur dioxide, nitrogen oxide, and carbon dioxide emissions increasing about 18%, 7%, and 4% respectively, largely from more coal burning.
- Tyler Cowen: Stand with free speech and the Constitution (Feb. 28, 2026)
  A federal judge blocked a Virginia law limiting under‑16s to one hour per day on social media, and halted $7,500 fines. The judge said the rule was over‑broad, blocked protected speech without parental consent, and treated similar content unequally.
March 4, 2026
Various (AI) Links: Mar. 3, 2026
- Venturebeat: Alibaba’s new open source Qwen3.5 Medium model offers near Sonnet 4.5 performance on local computers (Feb. 25, 2026)
  Alibaba released Qwen3.5 Medium models with agentic tool calling, near‑lossless 4‑bit quantization, and 1M+ token context on consumer GPUs. They match or beat similar proprietary models.
- Tyler Cowen: AI Won’t Automatically Accelerate Clinical Trials (Feb. 27, 2026)
  AI can design better drug candidates, but high trial costs, complex logistics, and the need for rich human data limit widespread therapeutic development. Chronic diseases, especially aging, require long, large trials to measure meaningful outcomes, making investment too costly.
- Tom Wojcik: What AI coding costs you (Feb. 14, 2026)
  AI tools boost productivity, but heavy reliance risks creating cognitive debt, skill atrophy, and a review paradox where people lose the ability to vet AI output.
- Simon Willison: Interactive explanations
  When agent-written code becomes opaque, teams incur cognitive debt that slows development. Building interactive explanations, like an animated walkthrough of a Rust word-cloud showing spiral placement, restores understanding, confidence, and ease of future changes.
- OpenAI: Supply Chain Risks (Feb. 28, 2026)
  “We do not think Anthropic should be designated as a supply chain risk and we’ve made our position on this clear to the Department of War.”
- NY Times Opinion: If A.I. Is a Weapon, Who Should Control It? (Feb. 28, 2026)
  A clash between the Pentagon and Anthropic over military A.I. use pits corporate ethics against national security, stoking fears of autonomous weapons, centralization, and industry break-up.
- Anthropic: Statement on the comments from Secretary of War Pete Hegseth (Feb. 27, 2026)
  The Department of War will label Anthropic a supply-chain risk after talks stalled over two exceptions, mass domestic surveillance, and autonomous weapons. Anthropic calls the move legally unsound, will sue, and says commercial and individual access to Claude is unaffected.
- Tyler Cowen: What the recent dust-up means for AI regulation (Mar. 2, 2026)
  There is no comprehensive federal AI law (and a Trump executive order limited state rules), but an informal “soft regulation” exists: major AI firms keep national security agencies informed and shape products to avoid triggering formal restrictions.
- Transformer: OpenAI’s Pentagon red lines are a mirage (Mar. 2, 2026)
  OpenAI struck a Pentagon deal claiming bans on domestic mass surveillance, and lethal autonomous weapons, but the contract reportedly contains vague wording.
- NY Times: I.R.S. Tactics Against Meta Open a New Front in the Corporate Tax Fight (Feb. 24, 2026)
  The I.R.S. says Meta undervalued offshore intellectual property, seeks nearly $16 billion in back taxes. If upheld, the tactic could recover vast taxes, deter profit shifting, and trigger a major Tax Court fight.
March 3, 2026