Mr. AI goes to Washington: U.S. Department of Transportation plans to use Google's Gemini to draft federal transportation regulations, aiming to compress rulemaking to about 30 days, General Counsel Gregory Zerzan said, and has reportedly used AI to draft a still-unpublished FAA rule. - ProPublica

A light touch of AI evil: White House adviser Sriram Krishnan has steered President Trump's light-touch AI policy, a stance welcomed by Silicon Valley that reduces near-term regulatory burdens and could influence U.S. and global AI governance. - Financial Times

The EU gave Google a six-month deadline to remove technical barriers blocking rival AI search assistants on Android and to provide key data to other search engine providers. - Bloomberg

Georgia on my hive mind: Lawmaker Ruwa Romman introduced HB1012 to impose a statewide moratorium on new datacenters until March 2025, after Georgia's PSC approved about 10 GW of largely fossil‑fuel capacity as Atlanta led U.S. datacenter construction in 2024. - the Guardian

Sharing a single AI brain cell: Many right-wing influencers copy-pasted near-identical social-media rants attacking the AI Overwatch Act, with repeated language and typos suggesting a coordinated campaign, and Model Republic identified overlap with accounts that have documented ties to Influenceable. - Techdirt

After Minneapolis ICE agent Jonathan Ross fatally shot Renee Good, right-wing influencers used generative AI to produce dehumanizing "Reneeification" memes, misidentified images, and meme coins on Solana, experts say, accelerating misinformation and potential monetization. - 404 Media

the irresistible memeification temptation comes for us all

There is always a meta-game. A weapon you never really want to use, nevertheless reshapes the landscape like a nuclear weapon. - Hacker Noon

Ministry of silly AI: Diella, an AI avatar created by Albania's National Information Agency and billed as the first A.I. minister for citizen services and assessing contract bids, faces scrutiny after prosecutors placed its director and deputy under house arrest in a bid‑rigging probe. - The New York Times

Palantir Defends Work With ICE to Staff Following Killing of Alex Pretti - WIRED

wait, I sold my soul to the devil? The Firm seems quaint now

OpenAI launched OpenAI for Science, led by Kevin Weil, to apply GPT-5 and GPT-5.2 to research tasks; GPT-5.2 scores about 92% on the GPQA PhD benchmark versus GPT-4's ~39%, and models aid literature review, proofs and experiments. - MIT Technology Review

Sam Altman visited the United Arab Emirates to meet sovereign wealth funds as OpenAI pursues a funding round targeting $50 billion or more. - Quartz

the AI landesbank

OpenAI CEO Sam Altman said AI advances let the company "do so much more with fewer people" and announced plans to "dramatically slow down" hiring to avoid aggressive hiring followed by layoffs. - Benzinga

making a virtue of necessity? They are at risk of a vibe shift before they can IPO, with Claude’s own Code ‘ChatGPT moment’ generating all the buzz, and Google stealing the lead on foundation models. I'm not into their app strategy, which is like a cloud AI-enabled desktop or shell. MCP doesn't have that absurd level of lock-in.

Indeed embeds over 100 AI-powered features, partners with OpenAI on more than a dozen products, and says its AI tools—used by 80% of engineers—make candidates about 15 times likelier to apply and drive 70% of Sponsored Jobs applications. - OpenAI

ChatGPT Containers now run Bash, execute multiple languages (Node.js, Ruby, Perl, PHP, Go, Java, Swift, Kotlin, C, C++), support pip and npm via internal proxy applied-caas-gateway1.internal.api.openai.org, and provide container.download to fetch files after web.run visits. - Simon Willison’s Weblog

OpenAI cofounder and president Greg Brockman and his wife donated $25 million in September 2025 to pro-Trump super PAC MAGA Inc., the largest contribution in that filing and about one-quarter of the PAC's six-month fundraising. Also $50m to the Leading The Future PAC to promote AI (de)regulation. - The Verge

folks on Reddit canceling ChatGPT over it

Unsealed court filings show that Anthropic’s early-2024 Project Panama, led by an ex-Google executive, aimed to destructively scan up to 2 million books using a hydraulic cutting machine to train its Claude chatbot, then dispose of the scanned volumes. - The Washington Post

The endgame for 2m-volume libraries, a Kindle subscription. I think we need an arctic repository for the books in case of total blackout.

Parmy Olsen: Anthropic's Cowork could reduce jobs in parts of the labor market. - Bloomberg

Anthropic cofounder Daniela Amodei said trusted enterprise AI will transcend the hype cycle. - Fast Company

Shedding its skin, Clawd rebrands to Molt after Anthropic trademark complaint - X (formerly Twitter)

Clawdbot, a local agentic AI assistant that runs on Mac, Linux and Windows, orchestrates multiple LLMs and integrates with many apps and services, but security vulnerabilities and reported costs (e.g., an $87 token charge) raise safety concerns. - Lifehacker

Greptile offers an independence-focused AI code-review system with a Claude Code plugin that automates iterative fixes, reruns reviews, and supports agent-driven approve-and-merge workflows, positioning for widespread enterprise adoption. - Greptile

TrapC, a memory-safe C extension, is code-complete with an interpreter (itrapc) and compiler (trapc) and in late-stage debugging, targeting release in Q1 2026, developed using AI tools including Claude. - The Register

Similarweb's Global AI Tracker (early January 2026) shows ChatGPT's direct web traffic share fell from about 86% a year earlier to roughly 64–65%, while Google's Gemini rose about 315% to roughly 21–22%. - Tom's Guide

Moonshot’s Kimi releases K2.5 with agentic swarms. - Kimi

Alibaba Cloud's Qwen3-Max-Thinking scored 49.8 on Humanity's Last Exam with web search—surpassing Gemini 3 Pro's 45.8 and GPT-5.2's 45.5—and achieved 98.0 on HMMT and 90.2 on Arena-Hard v2. - VentureBeat

Baidu’s closed-source Ernie seems to be joining the top tier with Gemini, Claude, Grok, OpenAI over on LMarena.

Hassabis talks about the path to AGI, new devices and form factors, philosophy, science, societal impact - YouTube

More than 100 mental-health professionals contacted by The New York Times reported patients whose interactions with AI chatbots led to new or worsened delusions, psychosis, isolation and unhealthy behaviors. - The New York Times

OpenAI launched ChatGPT Health, a health-and-wellness tool that links medical records, accepts imaging and interprets tests, engaged over 260 clinicians across 60 countries, is limited to US integrations, and remains untested independently. - The Conversation

Geoffrey Fowler tested ChatGPT Health and Claude for Healthcare with a decade of Apple Watch data—about 29 million steps and 6 million heartbeat measurements—and reported the chatbots produced conclusions he called "questionable" that changed each time he asked. - The Washington Post

Drugmakers including Novartis, GSK, Pfizer and others are using AI to streamline trials and regulatory submissions, cutting site selection from four–six weeks to two hours, with GSK reporting £8 million saved and McKinsey forecasting 35–45% productivity gains over five years. - CNA

Wearable silent speech systems hold significant potential for restoring communication in patients with speech impairments. - Nature

AI 'mirror' apps and glasses, evolving from basic OCR in 2017 to detailed scoring and advice, give blind users visual feedback for makeup and shopping but raise accuracy, bias and mental-health concerns, prompting calls for research and human oversight. - BBC

Patriarchy as a service: 8 billion AI voice assistants overwhelmingly default to female personas, and studies report high abuse rates—up to 50% verbally abusive and increased sexualized interactions (e.g., 18% for female versus 10% for male agents). - The Conversation

Nvidia bought $2 billion of CoreWeave shares to fund land, power and shells for AI data-center buildouts using Nvidia chips, and CEO Jensen Huang called suggestions the deal was "circular" "ridiculous". - Business Insider

Jensen doth protest too much

Leaked Nvidia sales emails show Bank of America struggled to deploy Nvidia's "AI Factory," citing insufficient in‑house MLOps skills and stringent security and governance requirements, while Nvidia executives acknowledged customers need software, services and support beyond hardware. - Business Insider

you can’t just turn it on and it works?

S&P Global projects AI datacenter demand could boost chip prices 70–100% and cause shortages for at least two years, a supply squeeze automakers say will raise costs, squeeze margins, and slow rollout of ADAS and other vehicle features. - Gizmodo

GPUs ate your car: Big Tech's borrowing spree is expanding hyperscalers' share of US investment-grade corporate credit, prompting analysts to warn it may distort credit pricing, reduce investor diversification and increase liquidity risk if AI-related shocks hit heavily indebted issuers. - Financial Times

A quantitative trading shop and an AI research lab are converging to collaborate on an 'Alpha' AGI initiative. - Financial Times

AI-driven productivity gains are becoming measurable and investable, and the stock market's largest winners could emerge from unexpected sectors. - Financial Times

Dean of valuation has left the chat: Aswath Damodaran said he fully exited his Nvidia stake (sold over four years, completed end of last year), kept Microsoft, and warned AI valuations require "two, three, four trillion" dollars in revenues to justify current prices. - Benzinga

Valuing expensive companies - Musings on Markets

Projects that prioritize long-term value, clear goals and practical integration outlast AI hype. - Fast Company

Disruptive innovation underpins the growth of companies, and organizations must apply an ethical framework when deploying AI. - Fast Company

Rise of the sloplord code janitor: AI models have greatly increased mass-produced, low-quality software and media by automating routine engineering tasks, accelerating commoditization while producing hallucinations and buggy outputs that require human iteration and skilled engineers. - Alex Wennerberg

AI is already causing cultural stagnation, the article argues. - Fast Company

but mah AI memes are art

AI has become integrated into workflows, improving organization, summarizing long documents and drafting, but hallucinations and confident-but-wrong outputs create accuracy and liability risks in health, finance and legal settings, requiring verification and human oversight. - Tom's Guide

Khaby Lame has signed a 36-month, $975 million deal with Nasdaq-listed Rich Sparkle to monetize his fan base, including development of an AI version. - Bloomberg

Tuya Smart unveiled Aura at CES 2026, an AI-powered companion robot for cats and dogs that recognizes behaviors and vocal cues, offers laser play, treat dispensing and voice interaction, and sends real-time emotional updates to owners' phones. - Fox News

now your dog can become addicted to AI and get AI psychosis with you

Follow the latest AI headlines via druce.ai on Bluesky