Even if "solving" is impossible in principle, that is not defeat
In Episode 2 we touched on the idea that if the universe's update rule is "computationally irreducible," then to compute the universe is not to find a clever formula but to run it. This time we take it head-on. We look with our own eyes at problems with shortcuts and problems without, actually running cellular automata, and go all the way to why "unsolvable" is not defeat.
A planet's position can be leaped ahead by a formula to \(1000\) years from now — no need to step through \(1000\) steps. This is computationally reducible. It's no accident that most textbook physics is reducible: we pick "solvable" problems to put in textbooks. Reducible systems are really a special, lucky corner.
Many systems, on the other hand, require actually running \(n\) steps to know the state \(n\) steps ahead. You can't jump with a formula. This is computationally irreducible. The system is its own fastest simulator, and no faster predictor exists.
"No shortcut" is not just a rule of thumb; there's a theorem at the root. When a system can do universal computation (as Rule 110 can), you can embed any computation inside it. Then answering "what does this initial state eventually become" in general would amount to solving the halting problem (Turing) — which is impossible.
Universal computation ⟹ the halting problem / Rice's theorem ⟹ general prediction of long-term behavior is undecidable. So a "function that gives the answer before running" is impossible in principle. The dynamic face of Episode 2's Kolmogorov compression floor: if the trajectory is incompressible, then \(K(\text{trajectory})\approx\) its length = it can't be summarized.
Let's shine light again on Episode 2's layer-3 gap. If the universe's update rule is irreducible — and if it's universal, it almost certainly is — then "computing the universe" is not finding a magic closed formula. The universe is running itself, and there's no shortcut that overtakes it. There is, in principle, no way to know the future any faster than real time.
Here's the heart of this episode. Irreducibility sounds like despair, but it's the opposite:
| Consequence of irreducibility | Why it's a positive |
|---|---|
| Science doesn't stop | Even if the whole is irreducible, reducible pockets are everywhere (symmetries, conservation laws, effective theories, statistical laws). Science = the work of finding reducible islands in an irreducible sea. |
| Determinism ≠ predictability | Even if the rule is fully fixed, you don't know the result until you run it. It's precisely because of irreducibility that genuine novelty and complexity arise. |
| Not a gap in knowledge but a theorem | "No shortcut" is not our ignorance but a structural fact about the system. Not a missing piece to be filled, but a verified property. |
| The deep reason for the code | If someone claims "I obtained a closed solution" for an irreducible system, doubt it first — this is the basis for this series' "doubt the solved." |
Computational irreducibility is Wolfram's framework — empirically ubiquitous and intuitive — but "the universe is computationally irreducible" depends on the hypothesis that its update rule is universal, and that's unproven. Reducible / irreducible is not binary but a spectrum, and even irreducible systems have reducible aspects (which is why physics has predicted so well — it captures reducible islands like symmetry and coarse-graining. No contradiction).
The halting problem, Rice's theorem, and Rule 110's universality (Cook) are rigorous mathematics, but applying them literally to physical prediction requires the hypothesis "universe = universal computer." The figure is an actual run of a naive 1D cellular automaton, an illustration of reducible / irreducible, not a proof of which the universe is. Nothing in this document is "solved."
Problems with a shortcut (reducible) are a lucky corner; most are irreducible — to know the state \(n\) steps ahead, you must run \(n\) steps. The root is universality and undecidability (halting problem, Rice), the dynamic face of Kolmogorov's compression floor. In the figure we actually saw Rule 90 (reducible, has a formula) and Rule 30 / 110 (irreducible, can only be run). If the universe's rule is irreducible, there's no shortcut that overtakes the universe, in principle.
But "unsolvable" is not defeat but progress: science = the work of finding reducible islands in an irreducible sea, determinism ≠ predictability, and it's precisely because of irreducibility that novelty emerges. Above all — the basis for this series' very code, "if someone claims a 'closed solution' for an irreducible system, doubt it," was right here. We plant no flag.
Print / PDF: Ctrl+P (⌘+P on Mac). On screen, changing the rule with the slider or buttons runs it on the spot and shows the space-time diagram. "See the answer" opens each solution.