Summary The author argues that any serious attempt to model the long-term future mathematically produces extreme outcomes by necessity. When you build a formal model—whether qualitative or quantitative—and actually work through the logic to its conclusion, you can't end up with something mundane. A simple example: if computers improve faster than humans, and you keep applying that rule forward, you eventually get computers vastly smarter than humans. Qualitative models like this can show you the sequence of states the world passes through, but they can't pinpoint timing. Quantitative models do better on timing for the near term, but their detailed predictions break down the further out you go. What matters in the end, though, isn't the timeline—it's what the author calls the "ultimate future," the stable end state where the model stops changing. And here's where it gets stark: there are only two possible stable states. Either technological progress continues unimpeded and you get something transcendent: von Neumann probes spreading through space, uploaded minds running at impossible speeds, Matrioshka brains dimming entire galaxies to infrared, the Omega Point. Or civilization collapses and goes to zero. No middle ground exists mathematically. There's no stable state where things are just... fine. The implication is uncomfortable. You can't think rigorously about the deep future and arrive at business-as-usual. The very act of modeling forces you toward extremes because that's where the math leads. This isn't pessimism or optimism—it's a structural feature of how formal reasoning works when you actually push it to completion rather than stopping at some arbitrary near-term horizon.