Meta launched an AI app that uses your Facebook and Instagram history to personalize responses from day one. Built on their Llama 4 model, the app transforms years of social media data into an AI that claims to understand your preferences, habits, and interests.
Amazon plans to show customers exactly how much Trump's tariffs inflate their shopping bills. The move sparked immediate backlash from the White House, marking a dramatic shift in the previously warming relationship between Trump and Amazon founder Jeff Bezos.
Alibaba released eight new AI language models Monday that match or beat the performance of top competitors like Google and OpenAI. The new Qwen3 family ranges from compact 0.6-billion-parameter models to a massive 235-billion-parameter version.
Good Morning from San Francisco,
Nvidia flexed hard this week, unveiling chips that promise 50x more punch 🚀.
Just one catch: these beasts guzzle electricity like there's no tomorrow ⚡.
CEO Jensen Huang's new Blackwell Ultra will power everything from chatbots to self-driving cars 🤖, while devouring enough juice to light up 100 homes. Microsoft, Google, and Amazon keep throwing billions at Nvidia anyway 💰.
Speaking of billions, Google just dropped $32B on cybersecurity firm Wiz 🛡️. The plot thickens: Trump's comeback helped seal the deal 🎯. His new antitrust picks sparked hope for smoother regulatory sailing ⚖️. Google didn't play coy - they dangled a massive $3.2B breakup fee and pumped their offer by 39% 📈.
Who needs reasonable electric bills or regulatory worries when you're building the future? 🤷♂️
Stay curious,
Marcus Schuler
💪 Blackwell Ultra: Nvidia's new beast delivers 50x more speed than previous chips. Because apparently, regular fast wasn't fast enough.
⚡ Power Hungry: The GB300 rack system devours 120 kilowatts. Your utility company just got nervous.
🔄 Release Schedule: Annual releases now, not bi-annual. Jensen Huang got tired of waiting.
🎯 Coming Soon:
💼 Business Impact: Nvidia's sales jumped sixfold since ChatGPT's launch. Their accountants are getting carpal tunnel from counting money.
🤖 New Software: "Dynamo" framework promises better performance. Think of it as RedBull for AI chips.
🌡️ Cooling Challenge: These chips run so hot they need car-engine-style liquid cooling. Data centers might qualify as saunas soon.
📊 Market Reality: Microsoft, Google, and Amazon keep buying despite cost. FOMO is expensive in tech.
🔮 Future Tech: Optical connections could replace copper wires. But Huang says they're not reliable enough yet. Copper stays king.
💡 Bottom Line: Nvidia bets bigger computing beats better efficiency. Sometimes brute force wins – if you can pay the electric bill.
Nvidia unveiled its next-generation AI chips Tuesday, promising a staggering 50-fold performance boost. The catch? These chips still guzzle electricity like a data center full of space heaters.
CEO Jensen Huang revealed the Blackwell Ultra chip family, shipping later this year, followed by the Vera Rubin GPU coming in 2026. The new chips will power everything from chatbots to self-driving cars. They'll also consume enough power to make your local utility company break into a cold sweat.
Nvidia dominates AI chip manufacturing, with sales skyrocketing sixfold since ChatGPT's debut. Cloud giants Microsoft, Google, and Amazon continue throwing billions at Nvidia's chips despite their eye-watering energy demands. The latest GB300 rack system draws 120 kilowatts – roughly equivalent to powering 100 homes simultaneously.
The company brushes off concerns about more efficient competitors like DeepSeek. Huang insists that advanced AI requires massive computing power, regardless of software optimization. He championed Nvidia's new "Dynamo" software framework, which promises to squeeze maximum performance from these power-hungry chips.
Industry experts question whether this approach can continue. Mark Wade, CEO of optical computing startup Ayar Labs, warns that traditional copper connections between chips can't sustain the industry's appetite for more power. "Optics is the only technology that gets you off of that train," he notes.
Why this matters:
Read on, my dear:
Prompt:
Full body standing photography of a compressed telluric women full body humanoid android pregnant robot, 8 month pregnancy, profile, beauty cute gorgeous attractive divine, looking down at its transparent belly, and looking at the curled up fetus inside, pregnant, The scene captures their close bond, creating a touching moment between them, Photorealism Photography, Realistic Detail, Depth of field, 8k, Full HD, 3d, Super resolution, octane render, Long exposure, unreal engine, HD
Google snatched up cybersecurity firm Wiz for $32 billion this week. The deal's secret ingredient? Donald Trump's return to the White House.
The tech giant boosted its original $23 billion offer by 39% and dangled a massive $3.2 billion breakup fee. But Wiz executives really perked up when Trump appointed new antitrust leaders. His picks, Andrew Ferguson at the FTC and Gail Slater at Justice, sparked hope for smoother regulatory sailing.
The deal's architects worked furiously behind the scenes. Wiz's new CFO Fazal Merchant joined in January, just as deal talks heated up. Google cloud chief Thomas Kurian helped orchestrate the massive buyout while Wiz weighed going public instead.
Wiz brings serious firepower to Google's security arsenal. The Israeli firm boasts 70% annual revenue growth and rakes in over $700 million yearly. Google thinks that justifies the premium price tag – roughly equal to buying Instagram and WhatsApp combined.
The mammoth breakup fee reveals Google's determination. At over 10% of the deal value, it dwarfs the typical 4-7% range for billion-dollar deals. Google clearly learned from Adobe's failed $20 billion Figma purchase, which crumbled under regulatory pressure.
Why this matters:
Read on, my dear:
OpenAI's first Stargate data center will pack enough computing muscle to make current AI look like a dusty pocket calculator - with room for 400,000 Nvidia chips in a sprawling complex rising from former mesquite fields near Abilene, Texas. The 1.2-gigawatt facility breaks ground this summer and aims to start humming by mid-2026, though OpenAI keeps quiet about exactly how many chips will actually call Texas home.
Stability AI just unveiled "Stable Virtual Camera," a model that transforms flat images into dynamic videos with depth and perspective. The tool generates new camera angles from existing photos, enabling movement effects like spirals, zooms, and pans. Bloomberg reports the struggling company—which previously burned through cash and lost its CEO—hopes this release signals a comeback after appointing "Titanic" director James Cameron to its board. The model stumbles with humans, animals, and water, proving even AI can't handle everything that moves.
Chinese tech giants have flipped the AI playbook, dumping their most advanced models into the open-source wild while U.S. companies clutch their algorithms behind paywalls. DeepSeek R1 kicked off the flood in January, and new models now pour out every few weeks like digital fortune cookies - each one more capable than the last. The strategy brilliantly sidesteps U.S. chip sanctions while transforming China's apparent weakness into a sword aimed at Silicon Valley's business model.
Nvidia, Disney, and Google DeepMind have joined forces to create Newton - a physics engine that promises to make robots move less like C-3PO after an oil change and more like real creatures. The proof waddled onstage at GTC 2025 on Tuesday, when a Star Wars-inspired BDX droid strutted next to Nvidia CEO Jensen Huang, giving theme park fans a glimpse of what's coming to Disney parks next year.
Microsoft partnered with Swiss startup inait to power AI with simulated rodent brains - because apparently regular computers weren't mousey enough. The digital whiskers might just revolutionize everything from Wall Street trading to factory robots.
Tencent posted strong fourth-quarter results, with profits up 90%. The company spent $10.6 billion on AI last year - triple its 2023 investment - while facing tight supply of crucial GPU chips. Gaming revenue soared in both Chinese and international markets.
Federal employees flock to Reddit's anonymous forums as Trump's downsizing hits agencies, turning quirky discussion boards into digital survival networks. What was once a place to gripe about office Zoom calls now hosts half a million monthly visitors seeking leaked memos, job cut warnings, and a splash of dark humor about their uncertain futures.
Trump's commerce chief told federal broadband officials to favor Musk's satellite service over fiber - just months after Musk donated $250 million to Trump's campaign and landed a White House adviser role. Someone's WiFi signal got stronger.
Social media giants now rely on volunteer users instead of professional fact-checkers to combat misinformation. Meta follows X's Community Notes approach despite Bloomberg's analysis showing critical flaws. The system responds too slowly, fails on politically divisive topics, and ironically depends on the same mainstream media sources that platform owners routinely attack. Billionaires get free labor while dodging accountability for their platforms' problems.
A team from Harvard Medical School and MIT has built an AI system that helps doctors make safer drug decisions.
The system, called TxAgent, checks how medications interact and considers patient specifics like age and genetics.
TxAgent uses a database of every FDA-approved drug since 1939, plus clinical insights from trusted medical sources. In tests, it scored 92% accuracy on drug safety tasks, beating GPT-4 by 25%. The system works consistently whether doctors use brand names, generics, or drug descriptions.
Unlike previous AI tools, TxAgent thinks through drug choices step by step. It pulls data from 211 medical resources to check drug interactions, spot potential problems, and suggest safer options. The system bases its recommendations on real medical evidence and established guidelines.
TxAgent excels at complex cases where doctors need to weigh multiple factors. It looks at how drugs interact at the molecular level, checks patient medical history, and flags potential risks. This helps doctors avoid prescribing combinations that might cause side effects or work poorly together.
Researchers tested TxAgent on over 3,000 drug reasoning tasks and 456 treatment scenarios. The system maintained high accuracy across different ways of referring to the same drug - a common source of confusion in healthcare.
Why this matters:
Read on, my dear:
Arxiv.org (pdf) : TxAgent: An AI Agent for Therapeutic Reasoning Across a Universe of Tools
Spinach AI joins your video calls and handles all the busywork you hate. Think of it as that super-organized colleague who never misses a detail, except this one doesn't steal your lunch from the break room.
🗣️ Records and transcribes meetings in 100+ languages - finally putting your high school Spanish to shame
📝 Takes notes while you focus on looking thoughtful and nodding at the right moments
⚡ Auto-updates your project tools and CRM - no more copy-paste marathon after meetings
🤖 Creates reports without asking "can you clarify that last point again?"
🔄 Integrates with Zoom, Slack, Jira, and Google Meet - plays nice with your existing tools
🔒 Keeps everything secure with SOC2, GDPR, and HIPAA compliance - serious stuff for serious business
Backed by Y Combinator, Zoom, and Atlassian. They put their money where their mouth is, so you can put this AI where your meetings are.
Good Morning from San Francisco, The tech world's most intriguing power couple hits turbulent waters. Microsoft and OpenAI,
Good Morning from San Francisco, Meta's AI chatbots slipped into dangerous territory. 🚨 The Wall Street Journal busted them
Good Morning from San Francisco, Baidu unleashed chaos in China's AI market. The search giant slashed prices on
Good Morning from San Francisco, While Apple keeps polishing its promises, Perplexity just crashed the voice AI party 🎭. Their revamped
Fuel your morning with AI insights. Lands in your inbox 6 a.m. PST daily. Grab it free now! 🚀
