Deep Dive
Escape the rat race: passion, skills, AI leverage
Harris Poll data shows 60% of US workers face toxic bosses—a historic high. James opens with the uncomfortable truth: you can't just grind through a miserable job anymore. People have options now. The escape hatch requires a specific framework. First, find your passion. It doesn't matter if it's scuba diving, kitesurfing, or kitchen renovation—James met a 25-year-old making $200K monthly renovating kitchens lean and mean. If you love the work, you don't work a day. Second, profile your aptitude. A Swiss company once profiled James and discovered his superpower: analytics. He learned to read charts, build models, and trade from that edge. Third and critical: leverage AI to 10x that skill. The person asking has $50K cash, $20K invested, and 11 months runway. That capital isn't for speculation. It's optionality—runway to build version 2.0 of yourself using your passion plus your aptitude plus AI amplification. Don't waste it on degen moves. Cut your burn rate, stay lean, and focus on that rebuild.
AI is not dead—capex acceleration and the real stack
James thinks about whether easy money in AI has ended every single day. He's not seeing the signs yet. Leopold blew up this week, which spooked some. But that's noise. The AI stack is alive. Capex is the heartbeat. If capex disappears, James is out. He watches it religiously. Right now it's accelerating. Building a 1-gigawatt data center used to cost $50B. It's sprinting to $60B. Nobody can keep up with demand. Micron will double from $800 to $1,600 per Wall Street's conservative targets. But Marvell—connectivity for data centers—has more runway. Jensen Huang said Marvell will be a trillion-dollar company. That's 5.8x from here. Data center velocity isn't slowing by any stretch. The cheapest producer of intelligence wins the AI race. That's coherent compute at scale, at low token cost. SpaceX is building 8 gigawatts of compute across Colossus 1 and 2, plus newer machines. They'll lease old machines, keep new ones. Everyone is starved of compute, chips, memory, power, land, permits. Casualties will happen. Leopold is one. But the thesis hasn't changed. James hasn't lost faith and would not rotate from AI into crypto yet.
Broadcom squeeze vs. Marvell rebound, and the LEAP rule
Broadcom is up 800% in three and a half years. Most juice is squeezed. Analysts see 66% upside to $516 in one year—decent but not explosive. Marvell tanked 45% from highs after a bonkers 350% spike. Huge mean reversion. It's rebounding. Entire AI market got whacked in July when Leopold's firm imploded. Ken Griffin swept in and bought everything at half off—a legendary move. Marvell has far more runway than Broadcom. James would not trade Broadcom to Marvell if you'd pay taxes on short-term gains. Check the tax math first. But here's the non-negotiable rule: never buy short-term calls. Ever. Buy 900-plus day LEAPs instead. December 2028, soon summer 2029. Three years out. Yes, they cost more. But when an asset is beaten down and you buy a leap years away, you're fine. Time decay is brutal on short-term calls. That's when you sell them, not buy them. Honey lost 95% on 35- to 40-day calls and is asking whether to roll them. James's answer: cut losses, protect your future, get out. Don't throw good money after bad. Roll down is a mistake. Take the zero, put remaining capital into high-conviction assets. That's the lesson.
Tesla's structural advantage: 20 lines of business, only 2 are cars
James called Tesla a wild card due to near-term timing and volatility around Cybercab ramp. They're working through something—safety, back-end scale, car cleaning, parking logic. Unknown. But his car gets better every week with updates. Tesla operates in S-curves. Multiple curves hitting simultaneously. We're in the slowly phase. Expected back half of 2026 to be bonkers. We're two months into the last six months and it hasn't kicked off yet. Scrolling X shows Cybercabs stacked everywhere, testing nationwide. James expects launch soon. Tesla won't sit on capital. But they're cautious because mainstream media will crucify them for any accident. Waymo kills someone, crickets. Tesla scratches a car, global news. Elon doesn't advertise. His products threaten mainstream media. He's persona non grata. The exponential growth is coming. Supply chain fears are overblown. Tesla anticipates scarcity years ahead. They begged suppliers for lithium refining. Nobody would. Now they refine their own lithium and build their own LFP batteries in-house, avoiding tariffs. They design and build 4680 cells, AI5 chips, actuators, and Optimus themselves. Elon thanked Micron twice for future memory supply because he knows he'll need millions of AI5 chips TSMC and Samsung can't deliver. His plans are off the charts. But he's building things never built before. Be patient. The structural advantage is unmatched. Twenty lines of business. Only two are vehicles. The rest is compute, silicon, AI, autonomy, five energy modules, services, and ecosystem. If one hardware ramp delays, the portfolio keeps compounding. It's like nothing else on Earth.
Ignore 200-year-old chart patterns—algos trade milliseconds, not cycles
Ron asked about the Benner Cycle. Bob Prechter uses Elliott Wave. James has studied Wyckoff, Elliott Wave, and 18.6-year cycles for years. Benner lost money in the 1800s. Wyckoff comes from centuries ago. You cannot compare modern markets to history. Markets today run in milliseconds, driven by algos. Not static charts that hold for 16, 18, 20 years then panic every 7, 8, 9, 10, 11 years. No causal mechanism survives contact in the modern world. There are 30 to 40 different Wyckoff patterns. When you have 40 patterns, of course you can overlay them onto anything. James made a Wyckoff chart fit COVID infection rates. You can make anything fit if you believe your own fantasy. These methodologies miss all major panics. Simple buy and hold crushes trading around Elliott Wave, 18-year cycles, or Benner cycles. AI triggered a SaaS apocalypse. Money left software, went to AI. Now it's going back to software because it's cheap. Adobe ran. Leopold got toasted partly because of Adobe's move. This hocus pocus from the 1800s is everywhere on YouTube. Put it to bed. Ask AI how stupid and unreliable it is. That might help too.