The tech industry has become a game of musical chairs played by people who've forgotten the music. I see it everywhere—companies raising billions to build marginally better ad engines while pretending they're saving humanity. It's embarrassing. Here's the thing: innovation isn't a feature update. It isn't a new pricing tier. It isn't another AI wrapper that summarizes your emails slightly faster. Real innovation is the kind that makes the old way of doing things physically impossible to justify. I found out the hard way. When we started SpaceX, the incumbents said rockets cost $500 million per launch because that's what they cost. Period. Nobody questioned it. The white idiot index was off the charts—raw materials were maybe 2% of the price. The industry wasn't optimizing anything; it was just passing along inefficiency as if it were a law of physics. That's the deep problem. The industry has confused "how we've always done it" with "how it must be done." Look at what's happening now. Google's top AI researchers are leaving to start their own thing. X's product chief lasted a year. And what does the industry do? It churns out more of the same—another model, another funding round, another press release about "democratizing AI." I'm not saying AI isn't important. It's existential. But if you're building AI to make your spreadsheet app slightly better, you've missed the point entirely. The real opportunity is vertical integration. Own the entire stack. Don't outsource your core competency and call it "asset-light." That's just a fancy term for "we don't actually know how to make anything." Tesla doesn't outsource its battery chemistry. SpaceX doesn't outsource its engines. And if you're building something that matters, you shouldn't outsource the part that makes it matter. Counterpoint: people say this approach doesn't scale. That you can't do everything yourself. They're right—if you're trying to do everything. But you should be trying to do the one thing that creates disproportionate value. Everything else is a commodity. Treat it that way. The industry's obsession with "asset-light" models has created a generation of companies that are essentially shell games. They're valuable because they've aggregated attention, not because they've created anything durable. That's not a business. That's a rental. I keep coming back to the same question: what does the asymptotic limit look like? If you could solve this problem with zero friction, what would it cost? What would it take? Most companies never ask because they're too busy benchmarking against competitors who are also doing it wrong. The industry needs a reset. Not a pivot. Not a rebrand. A reset back to first principles. Ask why the thing exists. Delete it if you can't justify it. Then optimize what's left. It gets worse. The industry has convinced itself that speed is the same as iteration. It's not. Moving fast without learning is just expensive spinning. Real iteration means you're learning something each cycle that makes the next cycle better. If you're not failing, you're not innovating. But if you're failing the same way twice, you're not paying attention. I'm tired of the performative disruption. I'm tired of "visionary" CEOs who can't explain their own cost structure. And I'm tired of an industry that celebrates fundraising as if it were achievement. The truth is simple: physics doesn't care about your business model. The laws of thermodynamics don't negotiate. And neither should you. Build something that would be impossible for a rational competitor to copy. That's the only moat that matters. FAQ Q1: What is the "idiot index" and why does it matter for tech companies? The idiot index is the ratio of a product's final price to its raw material cost. A high index—say, 50 or more—means the manufacturing and supply chain process is adding massive inefficiency. For tech companies, this concept translates to asking whether your software's cost structure is justified by its actual infrastructure needs, or whether you're paying for layers of middlemen and legacy overhead that could be eliminated through vertical integration. Q2: How does vertical integration apply to a software company? Vertical integration in software means owning your core infrastructure rather than renting it from cloud providers. If your compute costs have an idiot index above 5, you're paying a significant premium for convenience. The asymptotic limit for compute is roughly one-tenth of current cloud pricing, which means companies that build their own optimized infrastructure can gain a 10x cost advantage over competitors who never question their AWS bill. Q3: What's the difference between moving fast and iterating effectively? Moving fast without a learning loop is just expensive chaos. Effective iteration means each cycle produces information that changes your next move. If you're repeating the same failures, you're not iterating—you're just being stubborn. The goal isn't to fail quickly; it's to learn quickly. Failure is only useful if it's informative, and information only matters if you act on it.