Straighten up and lose the conscience, soldier: The Pentagon is considering severing its relationship with Anthropic over the AI firm's insistence on maintaining some limitations on how the military uses its models. Anthropic says that two areas remain off limits: the mass surveillance of Americans and fully autonomous weaponry. - Axios

Claude of Duty: WSJ reported the U.S. military used Anthropic's AI model Claude via Anthropic's partnership with Palantir in a classified operation to capture Nicolás Maduro in Venezuela, which involved bombing across Caracas and, the defense ministry says, killed 83 people. - the Guardian

based killer robots or bust, silicon valley soy boys

Anthropic recruits Google veterans for data-center push - The Decoder

CC means Claude Code, not community college: Anthropic partnered with CodePath to integrate Claude and Claude Code into programming courses at community colleges, state schools and HBCUs, targeting over 20,000 students, more than 40% from families earning under $50,000. - The Register

Ludicrous speed on Cerebras: How Anthropic serves full Opus 4.6 at about 170 tokens/sec (~2.5× vs 65) via low-batch inference, while OpenAI runs distilled GPT-5.3-Codex-Spark on Cerebras 44GB-SRAM chips for >1,000 tokens/sec (~15×), trading fidelity for much higher speed. - Sean Goedecke

Codex is from Mars, Code Code is from Venus: arly-2026 head-to-head tests found Anthropic's Claude Code excels at architectural reasoning and explanations while OpenAI's GPT-5.3 Codex provides stronger defensive, production-ready fixes and autonomous simulation behavior, indicating complementary strengths for different engineering use cases. - Tom's Guide

Evolution of OpenAI's mission statement - Simon Willison’s Weblog

from “benefiting humanity unconstrained by financial return”, to “creating value for our shareholders”. we’re not smarmy, AI is complicated.

Microsoft AI lead Mustafa Suleyman said Microsoft will reduce reliance on OpenAI and develop in-house foundation models, aiming to launch some in 2026, while still holding 27% of OpenAI’s for-profit arm and IP rights until 2032. - Windows Central

like waiting for your prenup to expire

Developer swyx argues OpenAI should build a Slack-like workplace chat, citing Slack's $27.7B 2021 Salesforce acquisition, rising costs and outages, and potential network effects, embedded agents, and monetization opportunities for OpenAI. - Latent Space

OpenAI rolled out a ChatGPT contact-sync feature that lets users upload address books to find friends using OpenAI products, stores hashed phone numbers (not full contacts), may process numbers of non-opted-in people, and is reportedly unavailable in Europe. - PCMAG

Europe: could you maybe not tho?

Despite more than $150 billion invested over a decade, China is expected to produce about 2% of global AI chips in 2026 and have roughly 70 times less memory storage than foreign chipmakers. - The New York Times

ByteDance launched Doubao 2.0, an 'agent era' upgrade to China's most widely used AI app that can execute multi-step tasks ahead of the Lunar New Year. - Reuters

Google's AI-generated search "Overviews" can contain deliberately injected misinformation that may steer users toward harmful actions, creating reputational and product risk and spurring demand for verification tools and cautious practices like checking original sources. - WIRED

One robot, one vote? Researchers uncovered a network of more than 1,000 AI-driven social bots (the "fox8" botnet) that used realistic interactions and retweets to game recommendation algorithms, evade detectors, and manufacture synthetic consensus to influence public opinion and elections. - Salon.com

Gartner: AI bugs in physical systems could shut down critical services in a G20 economy as soon as 2028, raising outage, repair and supply-chain disruption risks and prompting calls for stronger controls and human intervention. - The Register

Wikipedia, the last good thing left, is experiencing a renaissance as public use rises and AI-driven aggregation both increases its value and exposes it to manipulation, prompting reliance on volunteer editors, citation rules, talk pages and bots to maintain accuracy, experts and community members say. - Vox

News publishers including The Guardian, The New York Times, Financial Times, Gannett, Le Monde and HuffPost are blocking Internet Archive crawlers to prevent AI scraping, with Gannett reporting 75 million blocked AI bots and many sites disallowing IA/OpenAI crawlers. - Nieman Lab

my robots are having an effect lol

NPR host David Greene sued Google after its NotebookLM produced a synthesized voice he says closely matched his NPR voice, prompting legal, consent and licensing questions for AI voice-generation platforms. - The Washington Post

The doctor will see no one ever again: Dr. Mehmet Oz is advancing a proposed $50 billion rural health modernization plan centered on AI avatars, while critics warn the technology could harm underserved communities amid Medicaid cuts and more than 190 rural hospital closures since 2005. - NPR

here, buy some weight loss berries

Vibe-productivity is up: New economic data indicates the United States is moving into a phase of measurable productivity gains from artificial intelligence. - Financial Times

AI is everywhere except in macroeconomic data, says Apollo’s Slok - CNBC

Analysts argue U.S. leadership in artificial intelligence, once presented as proof of American exceptionalism, now undermines that status. - Barron's

Concerns about AI risk are dominating conference calls as earnings grow but investors sell stocks. - Bloomberg

Traders at South Florida finance conferences, including the UBS Financial Services Conference in Key Biscayne, sold financial-services stocks on Tuesday amid fears new AI tools threaten banking and wealth-management business models, prompting concerns about near-term volatility. - Bloomberg

6 feet under in the unicorn graveyard: PitchBook estimates roughly 25% of private unicorn startups—VC-backed firms once valued at $1 billion or more—have fallen below $1 billion, prompting investor write-downs, fundraising challenges and potential acquisition opportunities. - Axios

AI agents powered by LLMs are writing production code and turning engineers into reviewers; Cora says agents produced all company code, enabling a six-person team to match output that formerly required 20–30 engineers, while agent use rose to 31%. - Business Insider

How Generative and Agentic AI Shift Concern from Technical Debt to Cognitive Debt - Margaret-Anne Storey

Stocks like Monday (MNDY) and Asana (ASAN) and related are getting absolutely destroyed but I wonder if you don't need them even more to transform workflows, orgs, culture. what do you think?

A production-grade AI Agent Platform decomposes into seven logical areas of concern and five technical capabilities; it recommends a mixed model hosting strategy with model routing and context caching to control cost and latency, and expects Generative UI adoption by 2026. - Hacker Noon

Associate manager of making sure AI doesn’t do that: IBM is tripling its entry‑level hiring and redesigning roles to require AI fluency, saying software engineers will spend less time on routine coding while HR staff supervise chatbots, the $240 billion company said. - Yahoo

Airbnb says a third of its customer support is now handled by AI in the US and Canada - TechCrunch

hidden cameras and bedbugs no longer the scariest thing at the AirBnb

Zapier's Chief People and AI Transformation Officer says only 26% of leaders report over half of AI pilots scale to production, while 69% of practitioners say most never scale; top blockers include integration complexity, system sprawl and governance reviews. - TechRadar

Aaron Sneed, a solo founder, built 'The Council' of 15 custom GPT agents deployed on OpenAI ChatGPT Business with Nvidia GPUs that handle business functions, save about 20 hours weekly, reduce labor costs and require human experts for final judgment. - Business Insider

The 'onion' prompt, a workflow for ChatGPT, Claude and Gemini, asks users to dump tabs, files and tasks so the AI categorizes items, selects top three priorities for 90 minutes and mandates closing, deleting or deferring at least 20%. - Tom's Guide

this newsletter is like an onion—it has layers and might make you cry.

Reddit user claims ChatGPT turned frustration into corporate-speak higher-ups could understand, got abusive manager fired. - Reddit

Curious Refuge, an AI film academy launched in May 2023, has trained about 10,000 students—95% entertainment or advertising professionals—in 11 languages across 170 countries and was acquired by Promise last February. - CNA

Fixing childhood trauma with AI. False alarm, Old Yeller! - Reddit

I feel bad for anyone born with 6 fingers, man - YouTube

An AI-generated video posted to Donald Trump's Truth Social depicted Barack and Michelle Obama as apes, prompting Obama to call it 'deeply troubling' while Trump blamed White House staffers and said no one was fired. - Deadline

Awful stuff, zero consequences, Europe defenestrates politicos over Epstein, US shrugs, it do be like that. “We Need to Let the Facts Come Out,” Says Lawmaker Who Will Invent Own So-Called ‘Facts’. Urges Party “Let’s Focus on Kitchen Table Issues,” While Arsonist Home Invaders are Torching the Kitchen. “Voters Demand Bipartisanship with the Gang Of Pedos”

Follow the latest AI headlines via druce.ai on Bluesky