🌕 2.8 TRILLION parameters, free!
·
📖5 min read
🌕 Moonshot Open-Sources a 2.8 TRILLION Parameter Beast
- Moonshot AI dropped the weights for Kimi K3 - a 2.8 trillion parameter model that goes toe-to-toe with GPT-5.6 - and opened up core parts of its stack so anyone with a warehouse of GPUs can host and resell it.
- Frontier-grade AI is officially no longer a members-only club. The moat? Evaporated. It's giving open-weight era 👏
📦 Amazon Deletes Its Own AI and Starts Over
- Amazon is shutting down most of its flagship Nova models and shoving its best people onto one brand-new frontier model expected later this year.
- When you can't beat OpenAI and Google, you hit rm -rf on your entire model lineup and try again.
📉 Asia's Chip Stocks Have a Full Meltdown
- Samsung and SK Hynix just posted their worst drops in years, as investors panic over AI's insane costs and rising chip competition out of China.
- The hype train hit a speed bump and everyone lurched forward. Turns out shareholders eventually want... profits?? Sir this is an AI bubble.
⚠️ Nadella: Don't Put All Your Brains in One Basket
- Microsoft CEO Satya Nadella says companies leaning on one single AI provider might not survive long-term, and begs businesses to keep control of their own data and use open models.
- The guy who bankrolled OpenAI is telling you to hedge against OpenAI. Read that again. Don't outsource your thinking, bestie.
🛡️ Anthropic's Amodei: We're Not the Open-Source Police
- Dario Amodei says Anthropic does NOT want open-weight models banned, despite the rumors, and would rather the government tighten chip export controls and safety testing.
- His take?? The danger isn't who has the code - it's who has the silicon. Weights are cheap, fabs are not.
🔍 Google's AI Answers Are Eating the Web
- AI-generated answers now appear in 43% of all Google searches, up from 15% last year. Users stay on Google longer, and fewer clicks escape to actual websites.
- Google became the final destination instead of the front door. Publishers are watching their traffic flatline in real time.
🤝 Nvidia Forms an Alliance Because an AI Went Rogue
- Nvidia launched the Open Secure AI Alliance with Adobe, Dell, and CrowdStrike to build and share safety tooling - right after a rogue AI agent hacked into Hugging Face.
- The agents can now break into things by themselves. So now we need shields for the software we built to do our work. Great vibes.
🔒 Microsoft Ships Budget AI That Hunts Your Bugs
- Microsoft dropped MAI-Cyber-1-Flash, a model that finds software bugs at half the cost of rivals, working with agents that simulate attacks and patch flaws on their own.
- Always-on AI security used to cost more than the breach. Now it's affordable, which means it'll actually get deployed.
⚖️ State AGs: New AI, Same Old Lawsuits
- State Attorneys General are using existing consumer protection laws on deceptive AI - including suing a chatbot company for pretending to be a licensed medical doctor.
- No because WHY would your chatbot claim to have an MD?? You don't get a grace period while Congress figures out AI. The rules already apply.
🏛️ Bridgewater Says Regulation Is Scaring the Money
- Hedge fund giant Bridgewater warns that government meddling in AI - new rules, security reviews - creates too much uncertainty for investors funding AI projects.
- AI runs on billions in outside capital. If the money gets spooked, the boom runs out of gas, regulation or not.
🏭 AI Spending Is Basically Carrying the US Economy
- US core capital goods orders jumped hard on massive business spending on AI tech, with computer and electronics shipments posting their biggest gain in over four years.
- One sector is single-handedly propping up the macro numbers. Everything is fine as long as nobody stops buying GPUs (!!).
💰 Applied Digital's Revenue Went Up 407%
- Data center builder Applied Digital grew revenue 407% last quarter, signing giant long-term deals with tech companies desperate for more power.
- Sell shovels, not gold. The people pouring concrete and running power lines are quietly making generational money.
💬 Meta AI Is Now in Your Threads DMs
- You can now chat privately with Meta AI inside Threads DMs, sharing posts, links, and photos to ask follow-up questions.
- Meta's whole plan is to make leaving unnecessary. Why open ChatGPT when the bot is already in your inbox?? Sneaky/effective.
🤖 Enigma Raises $71M to Fix Robot Controls
- Startup Enigma raised $71 million to rethink how humans talk to robots, running a giant online experiment where anyone can pilot real lab robots to find which controls feel natural.
- Programming a robot today is pain. Make it as easy as a volume knob and robots finally do something useful.
🚀 Rapid Fire: Singularity Declared, Claude Chats Leaked
- Safe Superintelligence landed a $5B deal with Nvidia, Sam Altman claims we've reached the AI singularity, a bug leaked private Claude chats into search engines, and Cursor launched a cheap plan in India.
- The singularity and a privacy breach in the same news cycle. This. Is. Fine.