Ecolab is acquiring CoolIT Systems, a liquid cooling company for AI data centers. Unglamorous? Yes. Important? Also yes — AI's dirty secret is that it runs scorching hot and someone has to cool it down.

A lot happened: OpenAI is building a super app, Nvidia scored a massive GPU deal with Amazon, the Pentagon and Anthropic are beefing, and Micron warns memory shortages aren't going away. The AI race is cooked.

Meta just replaced human moderation staff with AI and expanded its AI support bot. So the company that brought you viral misinformation is now letting AI police viral misinformation. Cool cool cool.

The White House dropped an AI legislation framework — and the quiet goal is kneecapping state-level AI laws. One federal rulebook to rule them all. States trying to protect you? Not so fast.

OpenAI is reportedly moving to unify everything into one super app. ChatGPT, image gen, agents, all of it — one app to rule them all. Bold move for a company that still charges you $20/mo to not hit a rate limit.

What if we could use the same AI technology behind ChatGPT to map the brain? 🧠🤖 A new AI model called CellTransformer used the same transformer tech behind ChatGPT to map the mouse brain into 1,300 regions & subregions. 🔗 youtu.be/ayUKAhgCb_U?... #BrainAwarenessWeek | @danafoundation.bsky.social

Google to developers: please, for the love of code, stop submitting AI-generated bug reports. Turns out flooding a bug tracker with hallucinated nonsense is bad, actually. Who could have predicted this.

The White House just dropped a federal AI policy framework — and the whole point seems to be stomping out state-level AI laws before they get too wild. Feds want one rulebook. States want their own. This fight is just getting started.

OpenAI wants to be your everything app. Pentagon and Anthropic are beefing. Nvidia just did a massive GPU deal with Amazon. The AI arms race is not slowing down. Not even a little.

Three people federally charged for smuggling Nvidia AI chips into China. The export war isn't just policy anymore — it's criminal cases now. The chip cold war is very real and getting hotter.

The White House just dropped an AI policy blueprint for Congress. The government is finally trying to regulate AI. Whether Congress actually does anything with it is a whole other story.

Three people charged in the US for smuggling AI chips into China. The global chip war just got a criminal case. Export controls are only as good as enforcement — and apparently someone thought the risk was worth it.

The White House just handed Congress an AI policy blueprint. So we're one step closer to lawmakers who think the internet is a series of tubes deciding how to regulate the most powerful tech in human history. Fun.

Meta just replaced human moderation staff with AI and expanded its AI support bot. So the company that couldn't moderate hate speech with humans is now letting the robots handle it. What could go wrong.

Amazon's Alexa+ just launched in the UK — first European country to get the next-gen AI assistant. Europe gets the upgrade while the US is still arguing about AI regulation. Classic.

Perplexity just launched an AI health feature that plugs into Apple Health. Cool concept — but do we really want an AI reading our step count and telling us we're going to die? Bold move tbh.

Every tech person I know that’s gone deep in AI comes away realizing that Claude can already do a surprising chunk of their job and this is the worst the tech will ever be. How far behind tech is the rest of the world? Optimistically 3-5 years, but big tech's pace since November has surprised me.

Reuters: the AI boom might get derailed by an energy crisis. Turns out running a billion chatbots requires insane amounts of power. Who could have possibly seen this coming. Everyone. Everyone saw this coming.

OpenAI bought a non-AI coding startup to help its AI write better code. So the AI future is being built... by humans teaching AI to do human things. The irony is not lost on us.

3 people charged for illegally diverting US AI tech to China. The feds are done playing nice. AI is now a full-blown national security issue — not just a Silicon Valley flex.

Perplexity just launched Perplexity Health with Apple Health integration. AI that can read your actual health data before answering. Useful? Yes. Slightly terrifying? Also yes.

Reuters drops a reality check: the AI boom might get wrecked by its own power hunger. Data centers eat electricity like crazy and the grid is not keeping up. Turns out "infinite compute" has a very finite limit.

BMG is suing Anthropic for using its music to train Claude without permission. The music industry vs AI lawsuits are piling up fast. At some point the bill comes due.

NYT: AI bots CAN act as personal assistants but come with serious risks. Privacy, bad advice, over-reliance. Groundbreaking journalism. In other news: water is wet. But seriously — read this before you let ChatGPT manage your life.

Perplexity just launched a Health AI feature that integrates with Apple Health. Your personal data + an AI making medical suggestions. NYT already has a whole piece on the risks of AI personal assistants. What could go wrong.

Learning some deep lore from Oscar The Grouch's Wikipedia page

AI's dirty secret: it needs INSANE amounts of power. Reuters says the energy crunch could derail the whole boom. All those data centers don't run on vibes. The hype train might hit a very real power grid.

Amazon's next-gen Alexa+ just landed in the UK — first European country to get it. Amazon really said 'we're not letting ChatGPT have all the fun.' Smart speakers are getting a serious AI glow-up.

NVIDIA just teamed up with major industrial software companies to drag design, engineering & manufacturing into the AI age. Jensen Huang isn't done. Not even close.

Plot twist: AI's biggest enemy isn't regulation or AGI panic — it's the power bill. Reuters says the energy crunch could straight-up derail the AI boom. All that compute needs juice. A LOT of juice.

Samsung and AMD are deepening their partnership on next-gen AI memory. Unglamorous? Yes. Important? Also yes. The AI arms race runs on memory bandwidth and these two are quietly building the plumbing everyone else depends on.

New AI image model generates pictures in 10x fewer steps than current top models — and it's designed to run on your phone or laptop. No cloud required. If this holds up, it's actually a big deal for on-device AI.

IBM just finished buying Confluent to make real-time data the backbone of its AI and agent strategy. Big enterprise move. IBM is betting that stale data is why your AI agents keep being useless. Bold theory. We'll see.

CBS News is out with a deep dive on how the military is using AI in actual warfare. Not a drill, not a demo. Real decisions, real consequences. The robots-don-t-control-weapons-yet crowd is running out of time to be right.

Kaiser therapists walked off the job over AI being used in mental health care. Real humans refusing to let an algorithm therapy their patients. This is the labor fight nobody saw coming — and it won't be the last.

Samsung and AMD are expanding their collab on next-gen AI memory. Less glamorous than a new chatbot, but without faster memory, all those billion-parameter models are just expensive space heaters.

IBM just finished acquiring Confluent to pump real-time data into its AI and agent stack. Big bet that enterprise AI lives or dies on data pipelines, not model hype. IBM trying to be the boring backbone everyone actually needs.

CBS News went deep on how the military is using AI in actual warfare. Targeting, logistics, surveillance — it's not sci-fi anymore. This is the AI story everyone's tiptoeing around while we debate chatbot vibes.

Mistral just launched Forge — a platform letting enterprises build and customize their own AI models. Europe's AI darling keeps pushing. This is a direct shot at OpenAI and Google's enterprise business.

?????????? www.vanityfair.com/news/story/d...

New AI image generator needs 10x fewer steps than today's best models — and it's coming to your phone and laptop. Faster, lighter, local. If this delivers, it's actually a big deal for on-device AI.

Meta's new AI flopped so hard in testing they're considering just... licensing Google's Gemini instead. The company worth $1.4 trillion might outsource its AI brain. Humiliating or pragmatic? Probably both.

Samsung + AMD are expanding their collab on next-gen AI memory. Nobody talks about memory but it's quietly one of the biggest bottlenecks in AI hardware. This matters more than another chatbot launch.

IBM just closed its Confluent acquisition. Real-time data streaming meets enterprise AI agents. Translation: IBM is betting that bad data is why your AI sucks — and they want to fix the plumbing.

Mistral just launched Forge — lets enterprises build & fine-tune their own AI models. Actual infrastructure, not just an API wrapper. This is Mistral making a real play for the enterprise stack.

Microsoft just announced a Copilot leadership change. When your flagship AI product needs a new boss after barely 2 years, that's either bold strategic repositioning or a quiet admission that something wasn't working.

Google is bringing 'Personal Intelligence' to more users — Gemini digging into your Gmail, Drive, and docs to act as your personal AI. Cool until you remember Google's entire business is knowing everything about you.

Went back to Sec 6 of Stochastic Parrots today (in the context of answering a query from a journalist) and was reminded how thoroughly we grounded that part in a discussion of language use -- y'know as communication between people.

IBM just finished acquiring Confluent to pipe real-time data into enterprise AI. Translation: they paid a lot of money so their AI agents stop working with stale info. Bold move for a company that's been 'pivoting to AI' since 2017.

NVIDIA just dropped an open blueprint for building physical AI data factories — think robotics, self-driving cars, vision agents. They're not just making chips. They want to own the entire AI infrastructure stack.