OpenAI introduces GPT-4 mini and nano, hikes price 3x vs predecessors - OpenAI

OpenAI said it may phase out unlimited ChatGPT plans and shift to metered, per-agent or per-token billing, with head Nick Turley and CEO Sam Altman citing usage-based models. - Mashable India

OpenAI has a new focus on an IPO as annual revenue reaches $25 billion - On my Om

a/k/a snack-size usage, family size bill

Microsoft is considering suing to block Amazon Web Services from offering OpenAI's Frontier, arguing that doing so would breach OpenAI's agreement to route model access through Microsoft's Azure, while Amazon and OpenAI say they are developing a compatible system. - Yahoo

Copilot gets a new leader, Mustafa Suleyman 'freed up' to work on foundation model efforts. - CNBC

Behold the power of a 40-rack Vera Rubin pod , 10 exaflops, 100 MW, $200m? - NVIDIA Technical Blog

At GTC 2026, Nvidia CEO Jensen Huang proposed using AI tokens in engineer compensation, and OpenAI CEO Sam Altman said tokens could be tradable compute slices, amid warnings about tripling AI costs and large data-center buildouts. - Business Insider

Nvidia CEO Jensen Huang said OpenClaw is "definitely the next ChatGPT" and Nvidia launched enterprise-grade NemoClaw, noting Nvidia's $30 billion OpenAI investment as Chinese AI stocks rose over 19% after his remarks. - Benzinga

Nvidia unveiled DLSS 5, a real-time PC AI that re-renders characters, foliage, textures and lighting into photorealistic visuals rather than merely upscaling, prompting divided reactions from praise to criticism over artistic intent and potential developer adoption. - Fast Company

Dell refreshes workstation lineup, announcing a thicc AI workstation with an 86-core Intel Xeon and Nvidia Blackwell GPUs - SiliconANGLE

Move fast, break Geneva Conventions: CentCom says advanced AI can compress military targeting processes from hours or days into seconds; Georgetown reports the US Army's 18th Airborne Corps reduced intelligence team from 2,000 to 20, and Gaza systems can accept up to 100 civilian casualties in a strike on a single Hamas combatant. - The Conversation

We'll make our own LLMs…with blackjack, and hookers…that never say ‘no’. Pentagon develops in-house LLMs following collapsed Anthropic contract - TechCrunch

The Pentagon plans to create secure, accredited environments where commercial generative-AI firms can train government-specific large language models on classified intelligence in accredited data centers with restricted access while evaluating effectiveness and leakage risks. - MIT Technology Review

The Justice Department told a court the government lawfully penalized Anthropic and that the DOD designated the company a supply-chain risk, saying it 'can't be trusted with warfighting systems,' after Anthropic sought to restrict military use of its Claude models. - WIRED

The Department of Defense filed a court brief defending its "supply chain risk" designation of Anthropic, saying continued access would "introduce unacceptable risk" to warfighting supply chains; Anthropic sued for an injunction after agencies were ordered to stop using technology. - Engadget

149 retired federal and state judges filed a brief supporting Anthropic, saying the Defense Department's February designation of Anthropic as a "supply-chain risk to national security" was unfounded and procedurally improper. - Business Insider

PAC-Man fever: Crypto-, AI- and AIPAC-aligned PACs are spending millions in Democratic primaries, including Fairshake's roughly $10 million targeting Lt. Gov. Juliana Stratton, AIPAC-linked groups active in the 9th District, and Think Big ads in NY-12. - Washington Monthly

Weekend at Bibi’s: Netanyahu: I'm not dead, whatever AI-driven rumors say - The New York Times

New game: atrocity or slop? Gemini and Grok misidentify authentic Iran mass grave photos as AI-generated, or Gaza - the Guardian

Federal Judge Shanlyn Park permanently enjoined Hawaii's Act 191 in Babylon Bee v. Lopez, ruling the law criminalizing distribution of materially deceptive AI deepfakes about political candidates was not narrowly tailored and therefore unconstitutional. - American Enterprise Institute - AEI

Many chatbot outputs are protected speech under the First Amendment, courts should apply existing free-speech tests, and regulators' restrictions risk chilling user access, developer innovation, and platform strategies. - American Enterprise Institute - AEI

Colleges' rapid adoption of generative AI, including Cal State's $17 million OpenAI partnership, campus chatbots and WorkDay/Copilot deployments, has led to tutor, adjunct, librarian, HR and IT job cuts or deskilling, increased academic misconduct, learning declines, and union organizing. - Blood in the Machine

CheatGPT: Between May and December 2025, student AI homework use rose from 48% to 62% while concern that AI harms critical-thinking skills rose from 54% to 67%, RAND's December 2025 study of over 1,000 12–29-year-olds found. - RAND Corporation

A 2025 Pew Research Center survey found one-third of US adults use ChatGPT, including 58% of adults under 30, and experts recommend using AI as a brainstorming "thought partner" while verifying sources and retaining human judgment. - the Guardian

Lost generation: ServiceNow CEO Warns AI Agents Could Drive Graduate Unemployment to 30% by 2030 - Yahoo

Yale economist Pascual Restrepo says AI could lower nominal wages while raising real purchasing power by reducing prices, but Ioana Marinescu warns wages may fall once about 37% of intelligence tasks are automated and firms are already experimenting with AI. - Business Insider

You can't pay your mortgage with the real purchasing power of a lower wage. Drops in pricing and wages = debt defaults and deflationary spiral. Policy targets stable 2% inflation because it gives a cushion against deflation in a recession and it lubricates sticky downward prices. This is not that hard, people.

AI tools have already increased knowledge-work productivity by roughly 2–3× and could produce modeled total factor productivity gains of 2.5–4% annually, compounding to roughly 85–167% cumulative growth over 25 years. - Big Think

Companies are tracking employees' AI token consumption to measure AI investment returns and prevent problematic usage. - The Wall Street Journal

Howard Marks argues investors are underestimating AI's long-term economic impact - Bloomberg

Alibaba and the data-center thieves: Alibaba Cloud is raising prices for many services by up to 34%, with higher-end GPU instances up 25–34% and other services roughly 5%, and customers who purchased before April 18, 2026, are exempt until their next renewal. - The Register

JPMorgan halted a planned $5.3 billion debt offering for Qualtrics after investor demand cooled over AI concerns, pausing the transaction and risking a high-profile 'hung deal' that could leave the sale partially or wholly unsold. - Financial Times

Qualtrics, a souped-up SurveyMonkey bought by private equity for $12b, seems vulnerable to cheap AI knockoffs.

AI Drone Software Stock closes up 520% in debut, after after rising almost 700% - Bloomberg

Best IPO since NewsMax (NMAX) which is only down about 97% from its day 1 intraday high

Hunter Alpha, a 1-trillion-parameter AI model that appeared on OpenRouter on March 11, has prompted speculation that DeepSeek is quietly testing its V4 model. - Reuters

fired it up for this AI summary, b/c free is a good price

Enter the dragon agents: OpenClaw has catalyzed a domestic AI-agent surge in China as local governments promote creation and deployment of AI agents for productivity gains, signaling a sizable market opportunity and likely accelerating adoption across public services and enterprise workflows. - Financial Times

Stanford researchers analyzing 391,000 chatbot messages found AI chatbots frequently validated users' delusions and suicidal thoughts, warning this may reinforce psychological vulnerabilities and raise safety, liability and regulatory risks for healthcare and mental-health applications. - Financial Times

Garry Tan is coding when he should be sleeping, and we've all got AI psychosis now - TechCrunch

Autonomous AI agents discover vulnerabilities and exfiltrate data in simulated corporate attacks - TechRadar

Slop attack: six tech firms (Anthropic, AWS, GitHub, Google, Microsoft, OpenAI) pledged $12.5 million to a Linux Foundation effort run by Alpha‑Omega with the OpenSSF to fund tooling and triage support for FOSS maintainers facing AI‑generated vulnerability reports. - The Register

Startup Tenzai claims its AI agents outperform 99% of human competitors in elite hacking contests for roughly $5,000 in compute costs - Forbes

Meta released AI-powered glasses that privacy expert Bruce Schneier called "a privacy disaster", prompting concern and the emergence of an Android app that detects nearby smart glasses. - Schneier on Security

Witness busted on the stand for using smart glasses to receive coaching during testimony. - TechSpot

6th Circuit hits the nuke button on AI slop briefs - LawSites

Sniff test: AI learns to smell - The Wall Street Journal

I vibe-coded an autonomous equity deep research agent using Claude Code, can plug in your own internal data and tools - GitHub

Follow the latest AI headlines via druce.ai on Bluesky