Anthropic extends MCP, you can now give instructions for e.g. Canva in Claude and get an image back - Claude

Using Claude Code, a dev ported about 100,000 lines from Pokemon Showdown JavaScript to Rust in four weeks via ~5,000 commits, producing a faster Rust engine that matched the JS engine on 2.4 million seeded battles with 80 divergences. - Vjeux

Claude Code can definitely suck you in - jasmi.news

Claude Codecrastination: when you avoid the thing you're supposed to do by cranking out 17 other things you've been wanting to do for a while. - X (formerly Twitter)

This is one of the most vertical github star charts I have ever seen: clawd.bot - GitHub

Some people are deploying Mac minis with Claude Code bots that they can text throughout the day to automate tasks using their apps and credentials. Security nightmare but cool! It’s all fun and games until all your accounts are wired to a casino in Manila.

Amodei: AI will challenge us as a species - darioamodei.com

Cursor CEO Michael Truell said Cursor "built a browser with GPT-5.2" and claimed 3M+ lines produced by hundreds of agents over a week, but independent inspection found failing builds, manual patches, reused components, and little usable product. - The Register

How Cursor Shipped its Coding Agent to Production - bytebytego.com

European Commission opened a formal Digital Services Act inquiry on Monday into X after its AI chatbot Grok generated about 3 million sexualized images, including roughly 23,000 appearing to depict children, and to examine X's recommender systems. - the Guardian

Jensen and the Chip Monks: TSMC Risk … can make argument 2x for NVDA. Samsung and Intel can do good fab, but TSMC has a lead right now so everyone wants them, they can extract the value of leading vs trailing edge. - Stratechery by Ben Thompson

Microsoft unveils in-house AI chip, will run GPT-5.2 in Azure, and be offered to external developers - Azure

Microsoft unveiled Rho-alpha, a robotics foundation model derived from Phi that translates natural-language commands into low-level bimanual control, integrates vision and tactile sensing, and was trained with NVIDIA Isaac Sim on Azure, with real demonstrations and teleoperation feedback. - TechRadar

Nvidia bought CoreWeave stock after it fell by half - TechCrunch

Data centre groups plan a lobbying blitz and will increase advertising spending this year to counter public opposition over AI-related energy use and influence regulators and local planning to protect project timelines and investment. - Financial Times

Investors have driven a surge in memory stocks as strong AI-related demand and supply bottlenecks lift valuations, give suppliers short-term pricing power, and tie the rally's sustainability to continued constraints or sustained AI spending. - Financial Times

Qwen3-Max-Thinking, released as model identifier qwen3-max-2026-01-23, is Qwen's flagship reasoning model evaluated on 19 benchmarks with performance comparable to GPT-5.2-Thinking and test-time gains on GPQA (90.3→92.8), LiveCodeBench v6 (88.0→91.4) and IMO-AnswerBench (89.5→91.5). - Qwen

Baidu's Ernie is beating GPT-5.2 in LMarena.ai , GLM 4-7 is the top MIT licensed model.

Amazon's Bee is a $49.99 screenless wrist-or-clip AI wearable that records voice, supports about 40 languages and seven-day battery life, and—while Amazon says audio is discarded and transcripts encrypted—its privacy notice allows sensitive-data collection and sharing. - BGR

Apple is reorganizing its AI teams and developing two new versions of Siri. - Bloomberg

Meta cut roughly 1,000 Reality Labs jobs (about 10% of the unit), shifted investment from Quest headsets and Horizon Worlds to AI and Ray-Ban Meta smart glasses, after Reality Labs posted over $70 billion in cumulative losses since late 2020. - CNBC

Google has integrated generative AI into Gmail to streamline inbox management and email workflows, following its deployment of AI-generated answers in Search about two years ago. - The New York Times

OpenAI: AI as a scientific collaborator - OpenAI

OpenAI will introduce an adult 'erotica' mode for ChatGPT this quarter with a behavioral age-estimation model, while experts warn it may increase emotional attachment and harm. - The Observer

Waymo is running an eight-vehicle pilot of autonomous Jaguar EV robotaxis in Manhattan and downtown Brooklyn with safety drivers aboard, while NYC political and legislative decisions will determine broader deployment. - Politico

Communities don't love big Waymo charging zoos - AOL

Capital One agreed to acquire Brex for $5.15 billion, about half its $12.3 billion October 2022 valuation, which Jason Lemkin at SaaStr says illustrates how aggressive 'hubristic fundraising' narrows exit options and increases downside risk. - SaaStr

Ricursive Intelligence, founded by two former Google researchers and valued at $4 billion, is developing AI systems that can automate the creation and improvement of other artificial intelligence. - The New York Times

HelloSky, founded by Alex Bates, closed an oversubscribed $5.5 million seed round in April 2025 and offers a GenAI "moneyball" recruitment platform that maps candidate data and open-source contributions to uncover AI talent amid a shrinking experienced pool. - Fortune

Lessons from Building AI Agents for Financial Services - LinkedIn

I vibe-coded a 12-dimension, >200-question AI maturity self-assessment for e.g. hedge funds, alternative investment companies, similar firms. Work in progress, opinionated, over-long, everything doesn't apply to everyone, but should inventory most things people should be thinking about doing. - druce.ai

New York City launched in October an AI business-help chatbot trained on about 2,000 NYC business pages on Microsoft Azure, but The Markup found legally significant errors, misstating Section 8 and rent-stabilization rules and creating legal risks for businesses. - The Markup

Yann LeCun, who helped develop the technology behind today’s chatbots, warned that many tech companies are on the wrong path and could hit a dead end in creating intelligent machines. - The New York Times

"Technofascism" describes how AI, platform monopolies, opaque algorithmic governance, mass surveillance and targeted manipulation, reinforced by Big Tech–state ties, mirror fascist features and risk eroding liberal democracy unless systemic governance reforms occur. - SpringerLink

Sexual deepfakes continue to get more sophisticated, capable, easy to access, and perilous for millions of women who are abused with the technology. - WIRED

Since Mid-2025 ManyVids' official accounts have been posting weird AI-generated spiritual rants, often depicting CEO Bella French. - Futurism

Suno understood the assignment - Suno

Follow the latest AI headlines via druce.ai on Bluesky