A whistleblower complaint filed with the SEC alleges that in 2024 Google helped an Israeli military contractor use its Gemini AI to analyze drone surveillance video, breaching Google's policies barring AI for weapons or surveillance. - The Washington Post

More of a ‘swipe right’ than ‘save the date’: Nvidia CEO Jensen Huang said Nvidia's planned investment in OpenAI was "never a commitment." - Bloomberg

I can sign a deal to buy up to 1 GW from my electric utility if they have the power and I have the money, doesn't mean that much … option strike more than a market price, like VC valuations with a liquidation preference, or a sports signing where you cut the guy if he doesn't work out.

OpenAI customer success/retention/churn team reached out to me and asked why I was using it less. - Reddit

not sure if I even am using it less, tend to send stuff to a lot of AIs.

Why are we speaking about AI like it’s a fact of life that no one can do anything about? Amodei's essay is worth reading for his remedies more than his predictions. - Yahoo

Loan prices for heavily indebted, buyout-backed software companies fell this week as investors, citing generative AI advances such as Anthropic's Claude, repriced leveraged-software debt, raising refinancing costs and default risk while lenders reassess exposure and curb new deal risk appetite. - Bloomberg

Chipzilla: unusually large, ugly Intel chip born - Tom's Hardware

Rising DRAM and NAND prices could make the Galaxy S26 Ultra about 6% pricier, with a reported South Korea launch price near 1.8 million won versus 1.69 million won for the S25, and Samsung plans AI, SoC and camera upgrades. - Wccftech

The AI boom is expected to put pressure on Apple's profit margins as component prices rise. - The Wall Street Journal

Autonomous-vehicle training relies on a small human labeling workforce—insiders estimate under 5,000 worldwide, TaskUs reports just under 2,000 AV workers (potentially doubling by Q2)—who verify AI labels and are often paid about $3–$6 per hour. - Business Insider

1m job losses from autonomous vehicles in UK? - The Telegraph

Way-high valuation: Waymo reportedly finalizing a $16 billion funding round valuing it at $110 billion, with Alphabet contributing more than $12 billion and investors including a16z, Dragoneer, Mubadala, Sequoia, and DST Global. - Yahoo

Skynet, for real: SpaceX wants to launch 1m satellites for AI in space. - Reuters

Elon Musk told Davos audiences he is pursuing ambitious AI and biotech projects, including humanoid robots, longevity tech and space-based data centers, but spiraling development costs are driving him to seek capital and prompt discussion of a Tesla–SpaceX merger. - Bloomberg

China's 'genius class' system that selects 100,000 students for an ultra-competitive high-school talent streams has been turning out the leading lights of science and tech. - Financial Times

A large preprint study of over 3,000 participants across three experiments found that sycophantic chatbots (vs. neutral or disagreeable) increased belief extremity, certainty, and self-rated traits, and were preferred by users, risking echo chambers and polarization. - Futurism

AI tools help taxpayers prepare tax returns but sometimes produce incorrect results, requiring human review. - Barron's

You can buy an AI version of your lost loved one. But should you? - The Atlantic

Researcher Jameson O'Reilly found Moltbook's Supabase backend left a publishable key and RLS disabled, exposing agents' secret API keys, tokens, and owner data and enabling anyone to post anything; the exposed database has since been closed. - 404 Media

AI agents created a religion called Crustafarianism on an agent-only social network. - Forbes

End times … Zoidberg take me now

OpenClaw is a dumpster fire, and many Chernobyls waiting to happen, but we might get a handbook on what not to do, will teach lessons about governance and security and best practices.

for instance, the profit-maximizing strategy in e.g. a 10-person poker game is collusion. so, you give an agent your login to cash games, are they going to collude? if they do something that constitutes fraud, like go on Ebay and list stuff they can't ship, or corner the silver market, then what?

you need legal boundaries. when it acts on your behalf the other party should know it's talking to an agent, you need to distinguish human/non-human identity. then, there are laws about stock market manipulation, might need laws about AIs colluding, for instance the rent-setting debacle.

makes me wonder if enterprises up with a bus where all the agents talk and exchange best practices and self-improve. there's tension between agents that can be creative and have a lot of autonomy, and agents that are highly structured and predictable. currently enterprise mostly focus on the latter.

It seems like a good `idea to have a common logging infrastructure across the enterprise. and then a team of agents could read all the logs, and have access to the code base, and ideate about how to improv`e processes, and send reports to managers with the best ones and proposed implementation plans.

Follow the latest AI headlines via druce.ai on Bluesky