The piece argues that raw AI capability isn't the bottleneck anymore—distribution is. Labs are breaking revenue records, but the economy hasn't transformed despite having access to what would've seemed like AGI in 2019. The gap exists because real work is messy. It involves context that doesn't fit in prompts, legacy systems, human coordination, and incentive structures. Electricity took forty years to show up in productivity stats after factories got wired. AI adoption is faster, but institutional change still lags. That gap is the arbitrage opportunity, but the window is closing. The author lays out seven tactics for companies to capture value: build multiplayer networks that coordinate humans and agents (Harvey does this in law), accumulate workflow gravity by owning customer data and domain knowledge (Within captures latent work), let customers control their own transformation through configurable tools and model choice, and tell a credible story about what their industry becomes in five years. You need most of these simultaneously. The key insight is that when models commoditize, the moat shifts from raw intelligence to how deeply you're embedded in customer operations and decision-making. The structural advantage goes to whoever becomes harder to replace through coordination, not just capability. A smaller company that orchestrates better workflows beats a larger one. As model ecosystems fragment and enterprises hedge against single providers, companies that own domain-specific data and learning layers will know more about how their slice of the economy works than anyone else. That gravitational pull—built through product, services, and narrative working together—is what creates defensibility when everyone has access to the same models.