🎨 Meta's image glow-up
·
📖5 min read
🎨 Meta Finally Makes Its Own Images (No Middleman Required)
- Meta dropped Muse Image, its brand-new AI model that landed at #2 on the charts, right behind OpenAI. It's free inside Meta AI and can even fix its own mistakes.
- Meta used to rent its image tech from other companies. Now it owns the whole pipeline for Facebook and Instagram. Vertical integration era.
🏢 Microsoft Ghosts OpenAI for Its Own Models
- Microsoft is quietly swapping out OpenAI and Anthropic for its own homegrown models in Excel and Outlook - already running tens of thousands of prompts a week.
- The way that Microsoft funds OpenAI while simultaneously building the exit ramp?? Bestie that's the pettiest breakup in tech. It's giving "we're saving money now."
🇨🇳 China Wants to Gatekeep Its Own AI
- Beijing is cooking up new rules to limit foreign access to its best AI models - including tech from ByteDance and Alibaba - turning them into a national control lever.
- The U.S. restricted its models, and now China's firing back. We're speedrunning a split global AI market into rival zones. Cold War 2: Electric Boogaloo.
📉 Chip Stocks Take an hit on AI Spending Panic
- Samsung, Micron, and SK Hynix all took sharp hits today as investors got the jitters over whether the massive AI data center spending spree actually lasts.
- This is straight-up AI fatigue. Turns out even record profits can't stop investors from spiraling. Vibes: shaky.
🏦 Central Banks Are Officially Freaking Out
- The European Central Bank gave banks four months to build defenses against AI cyberattacks. Meanwhile the Bank of England warned AI threatens financial stability - too many people borrowing to bet on AI.
- When the central banks start sweating about AI hacks AND market crashes, maybe pump the brakes?? Not great!!
🇺🇦 Ukraine Goes Full DIY on AI
- Ukraine is building AI it can run on its own servers, based on Google's open tech, so nobody can flip the off-switch on its military and government.
- In wartime, AI sovereignty is a must-have. You can't have some remote provider casually cutting off your access mid-conflict. Smart move.
⚖️ Open Source Isn't Killing Frontier AI (Yet)
- New theory just dropped: open-source and frontier models aren't fighting - they're in a life cycle. Companies test with expensive models, then switch to cheap open-source for production.
- Top labs still print money because they own the discovery phase. Their premium prices are safe... for now. Emphasis on the "for now."
🍔 DoorDash Made an AI Bug-Catcher
- DoorDash built a test to see how well AI catches bugs in their code. Pairing a free model with Claude Fable 5 caught ~two-thirds of the problems - dirt cheap.
- Turns out you don't always need the most expensive AI. Sometimes the budget option catches the bugs anyway. We love to see it.
👾 Discord's AI Banned 8,000 People for... Chessboards
- A bug in Discord's AI moderation wrongly banned over 8,000 users for posting harmless stuff like chessboards. The AI saw grid patterns and just... lost it.
- Sir, this is a chessboard. This mess shows the danger of pure-AI moderation - humans are still deeply, deeply needed. No because WHY??
⚖️ Lawyers Are Hooked on AI + Norm Bags $120M
- Nearly half of new lawyers use AI daily for research and drafting, per a new survey. And AI law startup Norm just raised $120M at a $1.2 billion unicorn valuation.
- AI is eating the tedious legal grind - faster lawyers, billion-dollar startups. Objection?? Overruled.
🤖 Anthropic Extends Fable 5 + Drops Claude Cowork
- Anthropic is extending Claude Fable 5 access for all paid users through July 12, and launched Claude Cowork - start a task on your computer, finish it on your phone.
- Anthropic's making Claude work anywhere, keeping the pressure ON in the chatbot cage match. The race never sleeps.
🚨 AI Startup CEO Pleads Guilty to Insider Trading
- AppliedAI's founder secretly pleaded guilty to insider trading, using illegal tips from big-shot lawyers about company mergers to bank nearly $1 million in dirty profits.
- Turns out AI founders aren't above the law. The government is cracking down on the trading schemes. The audacity I-
🖌️ Figma Buys the Team Behind Bud
- Design giant Figma acquired the team behind Bud, an AI platform that writes code and automates tasks. Bud shuts down soon; the team moves to help Figma build app-creation tools.
- Figma wants to be more than a design toy - adding AI agents to turn static designs into real, working code. Design-to-deploy era.
🏆 Tufa Labs Wins Big with a Tiny Model
- Tufa Labs grabbed first place and $37,500 in the ARC-AGI-3 reasoning challenge using a small open-source Qwen model - by turning the puzzle into a coding problem in a Python sandbox.
- Small models can solve complex puzzles if you give them the right tools to think with. David beats Goliath, and Goliath cost 100x more.
💰 Chemistry Raises $500M for AI Fund #2
- VC firm Chemistry is raising $500M for its second fund to back early-stage AI startups. Founded by ex-partners from top VC giants who've already backed several AI wins.
- Market fears?? What market fears?? Big money is still gushing into AI. Investors clearly think the boom's got legs.
⚡ Envirotech Merges With Azio AI (For the Power)
- Envirotech Vehicles officially merged with Azio AI, pivoting hard to AI data centers and digital power tools - already building off-grid power for modular data centers in South Texas.
- AI's power hunger is exploding, and companies are pivoting hard to cash in on infrastructure. Everyone's a data center company now.
🚗 Momenta's IPO Debut Was... Fine
- Chinese self-driving startup Momenta had a flat debut in Hong Kong, opening just barely above its IPO price despite raising $751 million for driving software.
- The flat pop shows investors are getting picky - not blindly throwing cash at every AI stock anymore. Discernment era.