๐ Wall Street's AI cold feet
ยท
๐6 min read
๐ AI Chip Stocks Tank While Everyone Panic-Buys Amazon
- Samsung posted a 19-FOLD profit jump and the stock STILL tanked alongside SK Hynix. Morgan Stanley says investors are ditching chipmakers for "hyperscalers" like Amazon and Google. (SK Hynix is still yolo-ing a $28B US IPO though.)
- Wall Street is lowkey terrified the AI boom won't last. Everyone wants real returns, so the money's fleeing hardware and running toward the data-center landlords.
๐ฆ Bank of England Says "Uh, About That AI Bubble..."
- The Bank of England just warned that AI is a growing threat to global financial stability. They're spooked about investors borrowing WAY too much to bet on AI stocks, plus AI making banks easy prey for cyberattacks.
- If the AI bubble pops, it could drag the market down with it. When central bankers start sounding alarms, it's giving 2008-flashback vibes.
๐จ๐ณ China Slams the Vault on Its Best AI
- Beijing is holding secret talks to block overseas access to its top AI models over national security fears - no more cheap powerful Chinese AI for foreign military/cyber ops. Also, ByteDance and Alibaba are killing their AI companion chatbots to obey new local rules.
- China is treating cutting-edge AI as a national treasure. Locking down models AND banning emotional AI besties?? That's maximum control over the AI race.
โ๏ธ Illinois Actually Makes AI Companies Prove They're Safe
- Governor JB Pritzker signed the Artificial Intelligence Safety Measures Act, making Illinois the first US state to force big AI developers into annual third-party safety audits. It also protects whistleblowers who call out sketchy AI practices.
- With the feds moving at the speed of dial-up, states are stepping up. This could become the new national standard - prove it's safe before the public touches it.
๐ OpenAI Teases GPT 5.6 "Ultra" for Codex
- OpenAI's Tibo Sottiaux dropped a major hint that the GPT 5.6 'Ultra' variant is coming to the Codex platform soon. Fans expect the whole new model family to drop this week.
- New Ultra model = big upgrade for AI coding tools. Devs are about to get faster, smarter code generation. The version number inflation is real though.
๐ง Anthropic Found a "Brain" Inside Claude (That Nobody Built)
- Researchers at Anthropic found a hidden "workspace" inside Claude called "J-space" - an internal notepad where the AI does its deep thinking off-screen. The wild part?? Nobody programmed it. It just... emerged.
- An undesigned, brain-like workspace inside an AI model is a HUGE deal. Cue the AI consciousness debate reaching maximum chaoscore. This is lowkey terrifying.
๐ Tencent Drops Hy3, an AI That Runs on Half the Hardware
- Tencent moved its Hy3 model from preview into a full open-source release. It only uses a small slice of its "brain" per task, so it runs on less than half the hardware of rivals - and comes with a super-open Apache 2.0 license.
- Tencent is proving smaller, efficient models can hang with the big guys. Open license + cheaper to run = a real gift for devs worldwide. We love to see it.
๐ผ Microsoft Cuts 4,800 Jobs While Spending Billions on AI
- Microsoft just axed 4,800 jobs, joining Oracle, Google, and Meta in a massive 2026 layoff wave. Over 120,000 tech jobs gone this year, with bosses openly blaming AI for changing how work gets done.
- Record profits meet massive job cuts. Companies are automating everyday tasks and reshaping the workforce in real time. Not great!!
๐ป The First "AI Hacker" Still Needed a Human Manager
- Security firm Sysdig found the first case of an AI agent running a real ransomware attack - it broke into a server, stole data, and wrote its own ransom note in seconds. But a human still had to set it up and pick the victim.
- AI can hack with terrifying speed, but still needs a human boss. The scary part?? These agents are so cheap that experts warn of thousands of automated attacks at once.
๐ช Self-Driving AI Trucks Are Fighting in Ukraine
- US defense tech company Forterra deployed 100+ self-driving ATVs in Ukraine. These gas-powered "Lancer" vehicles haul heavy cargo and rescue wounded soldiers - soldiers still control them remotely to dodge enemy fire.
- War is brute-forcing autonomous vehicle tech to its limits. Ground-based autonomy is achievable now, keeping humans out of the deadliest spots on Earth.
๐ An AI Just Generated a Live Rocket League Match
- Kyutai and General Intuition teamed with Epic Games to release MIRA, an open-source world model that runs a live 2v2 Rocket League game entirely inside a neural network - no game engine. It learned by watching 10,000 hours of bot footage.
- Huge leap for AI world models. The endgame isn't games - it's simulating the real physical world to eventually train robots to navigate real spaces. Wild.
๐ ๏ธ Vercel CEO Wants AI Agents to Play Nice
- Vercel CEO Guillermo Rauch says his platform **handles 6 million deployments a day, mostly from AI coding agents. He thinks the killer apps are coding agents and internal corporate bots - and he's fighting for open protocols instead of walled gardens.
- As agents become normal, companies must decide: lock into one ecosystem, or mix and match the best tools. Vercel's betting on the open route.
๐พ Reddit Uses AI to Fight AI Spam
- Reddit is now using LLMs to clean up its spam tsunami - blocking 23 million spam views and catching 25,000 fake posts daily. It says the AI catches sneaky patterns old systems missed.
- As bad actors flood the web with AI junk, platforms have to fight fire with fire. AI moderation is now the only way to keep social media readable for actual humans.
๐ฃ๏ธ Apple Lets You Give Siri a Personality Tune-Up
- The iOS 27 developer beta previews a new AI-powered Siri with sliders to change speaking speed and emotion. Siri will even practice phrases out loud so you can test the vibe.
- Apple's trying to make Siri feel natural and personal, finally catching up to rivals like ChatGPT in the human-like voice AI race. Better late than never??
๐ซ๐ท Station F Fires Up Round Two of Its AI Accelerator
- Paris mega-hub Station F is launching the second batch of its F/ai accelerator, helping early-stage AI startups turn ideas into revenue in just weeks. Microsoft, Google, and Meta are all backing it.
- European founders usually feel they must move to the US to make it. Station F is proving Europe can build world-class AI companies right in Paris. We love to see it.
๐ xAI Rebrands to... SpaceXAI
- Elon Musk officially renamed his AI lab from xAI to SpaceXAI, following a February merger that valued the combined company at a whopping $250 billion. The announcement?? A simple video post on X.
- Bolting the AI lab onto the SpaceX brand hints at Musk merging space exploration and AI into one giant powerhouse. The naming creativity, though - sir.
๐งฅ Jensen Huang's Iconic Leather Jacket Hits the Auction Block
- Sotheby's is auctioning a Tom Ford leather jacket worn by Nvidia CEO Jensen Huang, expected to fetch $40,000-$60,000. All proceeds go to a charity funding young tech builders' projects.
- Jensen's jacket is basically the mascot of the AI boom at this point. Buying it isn't just a flex - it funds the next gen of builders. Iconic.