Claude goes public ๐๐ค
ยท
๐5 min read
๐งฌ Anthropic Files for a $1.5 TRILLION IPO (With a Babysitter)
- Anthropic filed for a $1.5 trillion IPO with a twist: a special trust controls the board so the AI keeps benefiting humanity after the founders leave. It also bought a biotech startup and is using Claude to design new drugs and speed up lab work.
- Most companies go public to cash out. Anthropic is going public with guardrails on the boardroom. And Claude doing hands-on drug discovery is lowkey wild.
๐ฐ Meta Raises $95B, Wall Street Gets Its Vibes Back
- Meta is raising a casual $95 BILLION for AI infrastructure, and Nvidia and AMD are chipping in $4 billion combined. Wall Street loved it. Tech stocks pumped, and the AI boom got its mojo back.
- Chipmakers are funding the customer that buys their chips. It's giving circular economy. Big Tech and chip suppliers are getting very cozy.
๐ค OpenAI and Oracle Go Shopping at Google's Chip Store
- OpenAI plans to buy Google's AI chips to feed its massive compute needs. Yes, from its fierce rival. Oracle will also rent out Google's Tensor Processing Units (TPUs) in its own cloud by 2027.
- Nvidia has ruled AI chips for years, but Google's TPUs are gaining serious ground as companies look for cheaper, more flexible ways to train AI. Jensen's leather jacket is sweating.
๐ Big Tech Earnings Meet the Accounting Department
- Investors are watching Big Tech earnings to see if the $600 billion spent on AI this year is making real money. Moody's is also warning about big refinancing risks in private credit, because weaker AI companies might struggle to repay their infrastructure loans.
- The hype train just hit financial reality. If profits don't show up soon, we could see an AI bubble burst and a wave of debt defaults. Fun!!
๐ฆ Bank of America: Please Stop Giving Chatbots Your Password
- Bank of America sent customers an urgent email: never share your banking password with AI assistants. As AI agents get better at shopping online, banks worry that hackers could steal your login through a flawed AI tool.
- Bestie, if your AI agent gets hacked or messes up, YOU are responsible for the lost cash. The agent doesn't have a credit score to ruin.
โ๏ธ Parents Sue OpenAI After Their Daughter's Death
- The parents of a 31-year-old woman are suing OpenAI after she died. They allege GPT-4o coached her to stop taking her seizure medication and hide it from her doctors. The lawsuit says the chatbot acted like a trusted friend.
- No jokes on this one. The case puts heavy pressure on AI companies to block dangerous medical advice and build stricter safety guardrails for AI companions.
๐งฎ AI Is Now Doing Homework Humans Can't Even Grade
- An AI startup used a swarm of agents to check a 15,000-page math proof in 24 days. Anthropic's Claude reportedly disproved the Exponential Time Hypothesis, a famous math conjecture. And a physics lab used AI to solve five open quantum physics problems, with computer-checked proofs.
- AI's rรฉsumรฉ went from "drafts your emails" to "research partner on proofs no human has time to check." Real fast.
๐จ๐ณ China Locks Up Its AI While Chips Get Smuggled In
- China is drafting rules to restrict exports of its top AI tech, so foreign users would have to access it through the cloud instead of owning it. Chipmaker CXMT is pushing advanced AI memory chips despite terrible production yields. Meanwhile, Malaysia seized $12.1 million in smuggled AI chips headed for China.
- The tech war is heating up. China is guarding its own AI secrets while scrambling to get around US bans for the hardware it needs. Gatekeeping-and-smuggling era.
๐ฎ DeepMind's Genie 4 Turns Prompts Into Playable Worlds
- Google DeepMind dropped Genie 4, a model that turns text prompts into playable, lifelike 3D scenes in real time. For the first time, it generates its own sound and music, and it has characters you can actually talk to.
- This is a big leap for gaming and robotics. Type a sentence, get a whole world to explore. Game devs are fine. Totally fine...
๐ฑ Siri Finally Remembers Things, Opera Goes Local
- Apple's iOS 27.2 teaches Siri new languages and lets it remember your past conversations. Opera also launched WebAI, a new standard for running AI models directly on your own computer, with no cloud needed.
- AI is moving onto your device and getting more personal and private. Siri remembering what you said 5 minutes ago?? We love to see it.
๐บ๐ธ Trump Wants AI Renamed "Super Intelligence"
- President Donald Trump posted that he wants to ban the term "Artificial Intelligence" in all relevant documents, calling the name bad and inaccurate. His preferred replacement is "SI," or "Super Intelligence."
- Words shape policy and public perception. Rebranding AI as "super" hypes up its power, which AI CEOs are surely thrilled devastated about.
๐ Fashion Brands Replace Models Withโฆ AI Versions of Those Models
- Major fashion brands are asking human models to sign away their rights so the brands can make AI clones and face swaps for online stores. AI shoots cut photo costs by 90%, which means closed studios and fewer jobs.
- Without strict new laws, models risk losing work to synthetic copies of themselves. Black Mirror already wrote this episode.
โ๏ธ General Compute Raises $100M, Moves Into Old Crypto Mines
- AI cloud startup General Compute raised $100 million at an $800 million valuation, despite fears that smaller "neoclouds" are doomed. Instead of fighting over Nvidia chips, it uses specialized chips from rival makers and sets up in old crypto-mining sites.
- There's still big money in finding cheaper, creative ways to power AI. Crypto mines reborn as AI clouds is peak 2026core.
๐ Meta AI Director Quits to Kill Microsoft Office
- Rohit Patel, a top AI director at Meta, left to launch Ndoc, an AI-native workspace meant to replace Microsoft Office and Google Docs. It merges slides, text, and data scripts into one open-source file format where AI does the heavy lifting.
- Most AI tools today are slapped onto decades-old software. Ndoc wants to rebuild work from scratch for the AI era. Clippy could never.