Good morning 👋 Have you been daydreaming about sun-soaked beaches but are locked in 996 at your desk? Don’t sweat it. It’s never been easier to snap envy-inducing casual vacation pics and live your best (fake) life. A new app called Endless Summer uses Nano-Banana to drop you into scenes around the world, no plane tickets needed. For a few bucks, you too can pretend to beat burnout and manifest the soft life you deserve.
🍪 When China dumped Nvidia, it wiped out 25% of their revenue. Jensen Huang spent years comfortably holding 95% market dominance in China – until US export controls crushed that to zero. Now, after losing billions, Huang is suddenly praising Trump’s tariffs and pivoting to American-made chip subsidies. Nvidia and TSMC unveiled their first “Made in the USA” wafer for next-gen Blackwell AI chips, calling it the start of a “$500 billion AI industrial revolution.” Nothing sparks patriotic spin quite like watching China walk away.
💾 This will go down on your permanent record. Stanford researchers found conversations people assume are “private” are being used by default to train AI models. Good luck opting out, since privacy settings are intentionally buried. Policies explaining how your data gets used are scattered, vague, and confusing on purpose. In some cases, companies can also use your data with their other products, creating a profile of you. Hard to say what’s worse: losing privacy or teaching AI our worst habits.
Pro Tip: To stop training OpenAI, go to Settings > Data Controls > Improve the Model for Everyone to turn off
☀️ Whoever wins the energy race wins the AI race. Right now, China is miles ahead. Last year alone, they installed 6x more solar than the US, enough to power six entire UKs. Meanwhile, America’s renewable investment dropped by over a third, tangled up in policy and stalled funding. By 2030, AI workloads will triple energy demands, turning clean power into the ultimate competitive advantage. We could still win the race, if only sunshine had lobbyists.
🤜🤛 Together with
AI for Enterprise
Retool’s new AppGen is the AI every enterprise actually wants. Type a sentence, and it builds a secure, full-stack app directly connected to your data, in your cloud, with your rules.
Why you’ll love it: With Retool you don’t have to sacrifice speed for safety. It’s secure by default and has audit trails baked in. Skip the half-baked toy tools with “demo-only” vibes. Use AppGen to build internal apps you can trust in minutes, not months.
Used by teams at Nvidia, OpenAI, and Brex to move at startup velocity without sacrificing compliance.
🎖️ US generals are now taking orders from chatbots. Commander William Taylor says he’s become “really close” with his AI assistant, relying on it for operational and even personal decisions. The Pentagon admits these bots occasionally leak secrets or give subtly terrible advice, but military leaders insist faster decisions are worth the risk. In one Air Force test, AI drafted battle plans 400x faster than officers. The results were so impressive it hardly mattered that they were somewhat unrealistic. Good to know our military strategy is now guided by an app famous for making things up.
⏳ Andrej Karpathy says AGI is still a decade away. Despite all the hype, he argues artificial general intelligence faces massive, unsolved hurdles. Current AI rewards lucky guesses just as much as actual understanding. It struggles to come up with genuinely new ideas and easily gets tripped up by unfamiliar problems. Ironically, AI’s superhuman memory might actually hold it back, limiting its flexibility to handle new situations. Karpathy’s advice? Think of AI less as our replacement and more as a tool to boost our own thinking.
fun stats
amount Alibaba Cloud cuts Nvidia GPU usage by ✂️ The Chinese tech giant claims it’s new Aegaeon system slashes the number of GPUs needed for AI models from 1,192 to just 213.