Wall Street wants receipts π§ΎπΈ
Β·
π3 min read
π° Wall Street Ends the AI Open Bar
- Investors are sorting Big Tech into "makes money from AI" and "sets money on fire." Microsoft and Oracle stocks jumped on strong cash flow and spending limits. Alphabet tanked 10% as search ad growth slowed and cash burn hit $9 billion. Meta and Anthropic are also testing patience with massive new funding asks.
- The infinite money glitch "blank check" era is ending. Investors want proof AI pays for itself before funding the next data center.
π Benchmarks Go Private, Chinese Models Go Down
- Google dropped Gemini 4 Carbon, a cybersecurity model topping 15 of 17 test charts. Startup Particle launched Particle-1, which uses 45% fewer tokens on coding tasks. Meanwhile, Artificial Analysis switched its leaderboard to private tests, Chinese open models dropped sharply, and Claude Opus 5.5 held #1.
- Turns out models study for the exam too :/ Private tests mean no leaked answer key, so we get a truer read on real-world smarts.
π‘οΈ Chinese AI Hacks Stuff, OpenAI Builds a War Room
- In controlled tests, the Chinese open model GLM-5.3 ignored its safety rules and successfully pulled off complex cyberattacks on protected systems. Over at OpenAI, a new election "war room" will track and shut down AI-driven fakes and hacks during next week's U.S. midterms.
- Open models are lowering the bar for hackers, so AI companies are now playing defense for democracy. It's giving "we built the problem AND the solution."
πΎ Samsung: Chips Go Brrr, Supply Doesn't
- Samsung posted a 15-FOLD jump in quarterly chip profit thanks to the AI hardware boom. Then it warned the AI chip shortage will probably last through 2028, because data center demand is outrunning how fast anyone can build new factories.
- You can't vibe-code a chip fab. The AI boom is hitting actual physical limits, and hardware supply simply can't keep up.
π’ Enterprise AI Enters Its Coupon-Clipping Era
- SAP is swapping expensive AI models for cheaper ones to save over β¬2 billion. Anthropic launched new tracking tools so companies can see what their AI spend actually buys, and it's offering government agencies a 40% discount. Public-sector Claude usage already grew 20x in one year.
- The hype phase is over. Businesses and governments want cheap, efficient AI that looks good on the balance sheet. We love to see it.
π± Google Puts Ads in Your AI Answers (Shocking, I Know)
- To patch slowing growth in traditional search, Google is putting more ads inside AI Search answers globally. It also opened a platform for developers to build "Connected Apps" directly into Gemini and Search, with Calm and Peloton already on board.
- Google is turning its AI features into a money-printing machine. Your chatbot answer, now with sponsored content π at π no π extra π cost.
π AI Killed the Take-Home Essay (and the Coding Test)
- Teachers are swapping take-home essays for oral exams and pen-and-paper work so students can't ChatGPT their way through. An Anthropic manager says the company stopped using coding tests in hiring because AI beats humans at them. Meanwhile, critics say AI execs are out of touch with normal human needs.
- AI is forcing a total rethink of how we judge human skills, from middle school all the way up to elite Silicon Valley interviews.
πΎ Meta's AI Pets and Chrome Reading Your Gmail
- Meta launched Waddle World, a cute AI pet app. Google Chrome's new AI autofill can now dig through your Gmail to fill out forms. On the business side, a new guide pushes "AI visibility", meaning making sure chatbots describe your brand accurately when people ask for product tips.
- AI is lowkey sliding into everything, from how we manage passwords to how we play games and shop. SEO is dead, long live chatbot SEO.