Anthropic Tightens Claude Session Limits as OpenAI Removes Codex Usage Caps - Gizmodo

The decade-long personal Altman-Amodei feud shaping the future of AI - The Wall Street Journal

Anthropic finished training a new LLM, Claude Mythos, and is piloting it with customers; a leaked draft shows a pricier Capybara tier outperformed Claude 4.6 Opus and triggered >5% drops in major cybersecurity vendors' shares. - SiliconANGLE

OpenAI added over 20 plugins to Codex, including Figma, Notion, Gmail and Slack, packaging shareable workflows for planning, research and development and positioning Codex to compete with Anthropic's Claude Code via a marketplace-focused strategy. - ZDNET

OpenAI Codex adds plugins (as Claude Code has had for a whle) - OpenAI

With AI adoption issues, and an increasingly complicated OpenAI situationship, Microsoft is on track for its worst quarter since 2008 with shares down 25% in Q1, while planning AI-related capex to rise to $146 billion in fiscal 2026, up 66% from $88 billion, heightening investor concern about growth and margins. - Yahoo

Microsoft allows enterprise admins to batch-uninstall Copilot from Windows 11 - Windows Central

Meta reorganizes into 'AI-native pods' and adopts Anthropic's Claude Code for training - Business Insider

Meta to fund $11 billion Louisiana power project with 10 gas plants for Hyperion AI campus - Fortune

A publisher built 'Otto OS' in one week using Claude Code to automate editorial and finance workflows. - TechRadar

Doesn’t matter how good your AI is if your data is bad. American forces struck the primary school in Minab, killing 175–180 people—mostly girls aged 7–12—after targeting relied on Palantir's Maven Smart System and an outdated Defense Intelligence Agency database. - the Guardian

Forcepoint chief data strategy officer Ronan Murphy warned manual data labeling is about 70% accurate, leaving roughly 30% misclassified records that, as agentic AI accesses data at rest, in motion and in use, leave firms one prompt from disaster. - SiliconANGLE

Political Deepfakes Since 2025 Surpass Total Volume of Previous Eight Years - the Guardian

PAC to the future: AI Super PACs achieve 91% success rate in Congressional primaries after $7 million spend - Business Insider

License to snoop: AI-Powered License Plate Cameras Feed Searchable Federal Databases - The Conversation

Researchers claim large language models elevate expert consensus and promote more moderate views, contrasting social media's populist, polarising effects, and potentially reshaping platform economics, content moderation, and information markets if model design and governance prioritize consensus. - Financial Times

Dutch Court to Grok: Nudify this: Fines xAI €100,000 Daily Until Grok Stops Non-Consensual AI Images - CNBC

NeurIPS Reverses Ban on Sanctioned Researchers Following China Boycott Threat - Reuters

Qwen Tarantino: Junyang Lin, who led Qwen until he was unceremoniously fired recently, is on X writing (very) long posts about their experiences on Qwen from the inside and how to approach - X (formerly Twitter)

Alibaba.com president Kuo Zhang says 30–40% of the platform's customers are solo entrepreneurs enabled by AI agents, noting Accio Work (launched late 2024) now has 10 million monthly active users. - Business Insider

Analysts Warn AI Infrastructure Boom Could Lead to $9 Trillion Bust - Financial Times

Smells unlike peak ‘99 spirit: Greenspan let the economy run hot in the face of the deflationary impact of globalization and hopes for IT-driven productivity. With tariffs and deglobalization, an immigration crackdown, large fiscal deficits, oil prices/disruptions, Powell has less room for error. - Bloomberg

It’s a bittersweet symphony. Civilization peaked in the 90s.

10x Machina: Ex-Coatue trader wants to use AI to find 10 10-baggers in 10 years - Bloomberg

Senators Warren and Blumenthal target Nvidia's startup investments over antitrust concerns - Yahoo

LangChain of fools - Critical LangChain and LangGraph flaws patched, had potential to expose secrets in millions of AI deployments - The Hacker News

Rapid adoption of AI code-analysis tools is generating exponentially more vulnerability reports in minutes or hours, straining open-source maintainers and pushing OpenSSF and Linux Foundation to create MLSecOps tooling, training and meet EU Cyber Resilience Act compliance. - SiliconANGLE

Developers let AI run wild on unmaintainable 100,000-line apps as experts warn of deskilling risks - Business Insider

AI coding agents have issues that haven't come home to roost yet, skill atrophy, IP ownership, prompt injection. - Stand Up For Me

Cursor deploys new models continuously using RL on user interactions, claims improvements in A/B tests with some reward hacking issues - Cursor

Timeout corner: Jai is a lightweight sandbox for AI agents - Stanford News

The notion that agents are like human employees that should be on org charts is insane, actually. - The Wall Street Journal

Clippy stole my feelings: AI 'social offloading' risks eroding emotional intelligence in the workplace - Fortune

IRS deploys machine learning for audits as 11% of taxpayers adopt AI for 2025 returns - Fast Company

How a Solo Entrepreneur Uses Claude and ChatGPT as an 'AI CFO' for Tax Season - Business Insider

Amazon upgrades Alexa with agentic AI, reporting a 3x increase in user engagement and 50% more smart-home interactions. - TechRadar

Leaks reveal six new Apple products for WWDC 2026 including M5 Ultra Max Studio with 80 GPU cores and 7-inch Home Hub - Geeky Gadgets

Grandmasters adopt unpredictable strategies to counter AI dominance in chess - Bloomberg

Vibe-coding obsessions strain relationships - Business Insider

A robot jogs in Brooklyn - Reddit

Claude Uno, launching April 1, one prompt per day, $19 - Reddit

Follow the latest AI headlines via druce.ai on Bluesky