Impli reveals the exact APEX method professionals use to optimize AI prompts. The article shows the complete system specification—from analyzing requests to executing optimized prompts that work across all platforms.
Web scraping has quietly become the backbone of AI training data. But legal gray areas and sophisticated anti-blocking measures make success tricky. This guide reveals what works in 2025.
Sean Grove from OpenAI says coding is dead. Instead of writing code, developers should write specifications that generate software. AWS just launched Kiro to make this real, while GeneXus claims they've done it for 35 years
Good Morning from San Francisco,
Former OpenAI CTO Mira Murati raised $2 billion for her six-month-old startup. That broke every seed funding record by four times.
She recruited OpenAI co-founder John Schulman and former research chief Barrett Zoph. Top researchers from Google, Meta, and Mistral followed. The "OpenAI Mafia" strikes again—75 former employees started 30 companies and raised $8 billion combined.
The math is brutal. Fewer than 1,000 people worldwide can build cutting-edge AI models. Meta offers $100 million signing bonuses. Microsoft bought Inflection AI for $650 million mainly for its staff.
Meanwhile, Big Tech's AI identity crisis deepens. OpenAI's o3 model hunts information like a bloodhound, scanning dozens of websites per query. No other lab has built anything close.
Google faces a brutal choice: protect its search cash cow or chase AI glory. It picked both, badly. Meta's Zuckerberg spends hundreds of millions poaching talent in panic mode. Apple sits comfortably on the sidelines, letting others burn cash while it focuses on hardware.
The industry's dirty secret: making models bigger stopped working. Success now depends on smart applications, not monster clusters.
Stay curious,
Marcus Schuler
Former OpenAI chief technology officer Mira Murati raised $2 billion for her six-month-old AI startup. The deal values Thinking Machines Lab at $10 billion and breaks every seed funding record in Silicon Valley history.
The previous record was $450 million. Murati's round is more than four times larger.
Andreessen Horowitz led the investment, betting on a team of AI stars. Murati recruited top researchers from OpenAI, Google, Meta, and Mistral AI. OpenAI co-founder John Schulman joined as chief scientist. Barrett Zoph, formerly OpenAI's research chief, became CTO.
This shows the "OpenAI Mafia" in action. About 75 former OpenAI employees have started roughly 30 companies, raising nearly $8 billion combined. Murati's round accounts for a quarter of that total.
The AI industry has a talent problem. Fewer than 1,000 people worldwide can build cutting-edge AI models. This scarcity created a bidding war that makes professional sports look cheap.
Meta offers $100 million signing bonuses to steal top researchers. Microsoft paid $650 million to buy Inflection AI mainly for its staff. Companies buy entire startups just to get their teams.
The size of Murati's round shows a broader shift in AI investing. Venture capitalists make fewer but much larger bets, focusing money on proven teams rather than spreading it across many startups.
AI companies raised $73.1 billion in the first quarter of 2025, but the number of deals fell to a five-year low. Investors learned that most AI startups fail, so they bet big on the few with real talent.
Thinking Machines Lab enters a packed field where performance gaps between top AI models are shrinking. Murati's company aims to stand out through human-AI collaboration and scientific discovery.
Why this matters:
Read on, my dear:
Prompt:
a cozy 3D cartoon-style bedroom scene at night where the same teenage girl with light skin, big expressive eyes, and chestnut brown hair now untied and gently messy from the day, sits on the edge of her bed in soft pajamas. Her skateboard leans against the wall, and a little movie ticket stub peeks from her hoodie pocket. She looks around her softly lit room with a gentle smile, stretching a bit and tucking her feet under a cozy blanket. A sleepy cat curls up at her side. The room glows with fairy lights and pastel-colored details books, photos, and a favorite lamp. Through the window, stars twinkle in a peaceful night sky. The mood is calm, quiet, and full of bedtime warmth and reflection.
The AI revolution has split Big Tech into clear winners and losers. Some companies found their footing. Others are still falling.
OpenAI's o3 model changed everything. It doesn't just answer questions—it hunts for information like a bloodhound. The model can scan dozens of websites for a single query, relentlessly digging until it finds what you need. No other lab has built anything close to it, months after its release.
This puts pressure on everyone else. Google controls the best infrastructure and data but faces a brutal choice: protect its search cash cow or chase AI glory. The company picked both, badly. Its AI search results often feel broken, stuffed with generated nonsense that makes finding real information harder.
Meta's Mark Zuckerberg is spending hundreds of millions to poach AI talent. The hiring spree signals panic. Meta's latest AI models disappointed, and the company knows attention is the scarce resource it competes for. If people spend their time chatting with AI instead of scrolling Instagram, Meta's business model crumbles.
Apple sits comfortably on the sidelines. The iPhone maker doesn't need to win the AI race—it just needs the best AI to run on its devices. Its deepening partnership with OpenAI makes sense. Let others burn cash building models while Apple focuses on hardware that makes AI useful.
Microsoft looked unstoppable two years ago with its OpenAI partnership. Now that relationship is fraying. OpenAI wants independence and profit. Microsoft wants control and access. Both companies need each other, but neither trusts the other.
The industry's dirty secret: making AI models bigger stopped working. GPT-4.5 cost 100 times more to train than GPT-4 but barely improved for regular users. Labs now compete on efficiency, not size. They're making existing model sizes work better rather than building monsters.
This shift hurts companies that bet everything on having the biggest training clusters. It helps those focused on smart applications of existing technology.
Why this matters:
Read on, my dear:
stratechery.com: Checking In on AI and the Big Five
Loveable is a Swedish AI platform that lets anyone build production-ready web apps by chatting in plain language. It already counts roughly 500 k users, earns about $17 M in annual recurring revenue, and is eyeing a funding round near $1.5 B.
Quick Start (< 100 s)
1️⃣ ✍️ Sign up at lovable.dev → hit New Project.
2️⃣ 💡 Type one line: “Expense tracker with login & dashboard.”
3️⃣ ⚙️ Watch Loveable auto-build UI, backend & DB; live preview appears.
4️⃣ 🛠️ Chat tweaks: “Add dark mode 🌙,” “Connect PostgreSQL 🐘,” then regenerate.
5️⃣ 🚀 Press Deploy → host on Vercel/Netlify or 📤 export clean code to GitHub.
Amgen bought Microsoft Copilot for 20,000 employees last spring, then quietly switched to ChatGPT this year. The drugmaker's employees found OpenAI's product more fun to use and better at handling research tasks, leaving Microsoft's sales team scrambling to explain why their $14 billion investment keeps losing customers to its own partner.
Google brought historical Street View imagery to Google Earth, letting users time-travel from a bird's eye view instead of just street level. The feature arrives as Google Earth turns 20, though it feels more like catching up than celebrating—people have been virtually stalking their past selves on Google Maps for years.
Tesla's robotaxis violated traffic laws during their first day of paid service in Austin, prompting federal safety regulators to gather information from the company. Videos posted online show the vehicles speeding past limit signs and driving into oncoming traffic lanes before correcting course.
Uber customers in Atlanta can now hail Waymo's driverless cars through the app, making it the second city after Austin to offer the service. The robotaxis operate in a 65-square-mile area covering downtown and nearby neighborhoods, though they skip highways and airport runs for now.
Abridge raised $300 million at a $5.3 billion valuation, nearly doubling its worth from February's $2.75 billion round. The Pittsburgh startup's AI listens to doctor-patient conversations and writes medical notes automatically, now serving over 150 health systems nationwide and handling 50 million conversations annually.
Amazon transformed 1,200 acres of Indiana farmland into a massive AI facility that will consume enough electricity to power a million homes. The complex, built exclusively for AI startup Anthropic, represents the new scale of computing infrastructure needed for artificial intelligence development.
Sider.ai makes AI useful for everyone through a browser sidebar that never gets in the way. This bootstrapped Boston startup turned a simple idea—put ChatGPT and friends where people actually work—into 6 million weekly users who can't stop clicking. 🚀
The Founders Joel Liu and Rick Fan launched Sider.ai in Boston in 2023 with zero funding and maximum hustle. Liu previously built screen-recording software (Vidline Inc.); Fan brought digital marketing chops from Chongqing University. Their 20-person distributed team spans Boston to Taipei. They asked: why shouldn't everyone have AI help built into their browser? The market answered with millions of downloads.
The Product Sider aggregates 10+ AI models (GPT-4, Claude, Gemini, Llama 3) into one browser extension that works everywhere. It reads PDFs like a research assistant, summarizes YouTube videos in seconds, translates languages on the fly, and writes emails that don't suck. The "Deep Research" mode scans hundreds of sources and spits out analyst-quality reports. Power users love the model flexibility—if GPT-4 chokes, switch to Claude instantly. Free tier hooks users; $5-10/month keeps them.
The Competition Merlin and Luna fight for the same Chrome extension turf, but Sider's multi-model approach gives it an edge. Microsoft pushes Copilot integration in Edge; Google experiments with Bard in search results. The real threat? Big Tech could bake these features directly into browsers tomorrow. Perplexity AI grabs search queries; Grammarly owns writing assistance. Sider plays the Swiss Army knife—decent at everything, great at convenience.
Financing Zero venture capital. Liu and Fan bootstrapped through personal funds and subscription revenue—a rare move in an era of billion-dollar AI rounds. No disclosed valuation, no VC pressure, complete founder control. The freemium model funds growth organically. They hint at future investment but only if it aligns with their affordable, user-first mission.
The Future ⭐⭐⭐⭐☆ Sider built something people actually use daily—the holy grail of productivity tools. But Big Tech circling the space could crush independent players overnight. The company's speed advantage and model-agnostic approach provide some defense, plus regulatory changes might favor nimble startups over platform giants. International expansion and enterprise features could unlock new growth, but they're racing against the clock before Chrome gets its own native AI sidebar.
OpenAI's AI now controls computers while Trump prepares an order forcing AI companies to prove political neutrality for federal contracts. Tech giants must choose: adjust development or lose billions as China advances without limits.
Good Morning from San Francisco, TSMC posted a record $13.5 billion profit. AI demand drove the surge. The chipmaker
Good Morning from San Francisco, China spent $6 million to build what cost OpenAI hundreds of millions. 🤖 DeepSeek's
Trump banned Nvidia's China chip sales in April over security fears. Three months later, he reversed it after meeting CEO Jensen Huang. The $15 billion flip shows how economic pressure can override national security concerns in tech policy.
Get the 5-minute Silicon Valley AI briefing, every weekday morning — free.
