How much confidence should a pattern earn? Try the same small calculation with a handful of numbers and an answer begins to feel inevitable. Try it with millions and that feeling becomes stronger. Yet mathematics asks a question that repetition alone cannot settle: what makes the pattern hold for every possible case, including the ones no computer will ever reach?
In The Simplest Math Problem No One Can Solve – Collatz Conjecture, Veritasium’s Derek Muller explores this gap with mathematician Alex Kontorovich. Their subject needs only multiplication, addition and division, but it opens a much larger conversation about evidence. The video’s value lies in making a difficult question approachable without giving it an easy ending. We can follow the rules ourselves while learning to distinguish an example, a useful intuition and a proof.
Two rules open an unexpectedly large question
Begin with a positive whole number. If it is even, divide it by two. If it is odd, multiply it by three and add one. Repeat with the result. Starting at six gives six, three, ten, five, sixteen, eight, four, two, one. Continue beyond one and the final three values repeat: one leads to four, four to two, and two to one.
The Collatz conjecture proposes that every positive integer eventually reaches this familiar cycle. The word “positive” matters: changing the domain changes the question. So does the word “every.” A successful calculation starting at six establishes what happens to six and to the other numbers along that route. It does not automatically establish the fate of a starting number outside it.
This is a good moment to pause the video and work through a few examples. Use paper rather than trying to anticipate the answer. The exercise reveals something easily missed in a polished explanation: understanding a procedure is different from understanding its long-term behavior.
Nearby numbers can take very different journeys
A small change in the starting number can produce a surprisingly different sequence. Starting with twenty-six reaches one after ten applications of the rules. Starting with twenty-seven takes 111, rising as high as 9,232 along the way. These figures can be checked directly. They describe two finite journeys, without requiring a theory about all possible journeys.
The contrast helps explain the nickname “hailstone sequence.” Values can rise and fall before settling into the familiar cycle. A temporary rise is not evidence of an endless escape. Equally, a long decline in a collection of examples does not, by itself, establish that every possible trajectory must eventually descend.
There is also a simple trap in averaging the two rules. Odd numbers become even after multiplication by three and addition of one, so another division immediately follows. Sometimes several divisions follow. Counting odd and even integers in general does not tell us exactly how those operations will be arranged along a particular sequence. The ordering is part of the problem.
Why a computer search leaves a logical gap
A computer can check a vast range of starting values and record how each behaves. This is meaningful work: it can discover patterns, reject proposed explanations, expose programming errors and rule out counterexamples within the checked range. But a finite range, however impressive, still has a next integer.
Consider a deliberately invented rule: “Every positive integer is smaller than a trillion.” Testing the first million integers would produce a perfect record. Testing a billion would do the same. The mistake would be treating that record as an explanation of why the statement must always hold. Collatz is much subtler, but the distinction between extensive testing and a universal argument remains.
A counterexample would settle matters in the other direction. A cycle among positive integers that never reaches one would be enough. An endlessly growing trajectory would also contradict the conjecture, although observing a computer run for a very long time would not establish that it grows forever. “The program has not finished” and “the program cannot finish” are different claims.
What “almost all” can actually establish
One important result discussed in the video comes from Terence Tao. His 2019 explanation of the research states that almost all starting values eventually reach below any chosen bound that increases without limit as the starting value increases, however slowly that bound grows. Here “almost all” has a precise meaning involving logarithmic density. It is not a casual way of saying every number except a few.
Two distinctions prevent this result from being overstated. The permitted bound still grows; it is not the fixed number one. And the theorem allows an exceptional set. The Institute for Advanced Study’s description of Tao’s 2023 talk makes the separation between this result and full convergence to one explicit.
A simpler counting example gives some intuition for why “almost all” needs care. Powers of two become increasingly sparse among the positive integers, but they never stop appearing. Rarity does not imply nonexistence. This example uses ordinary counting rather than Tao’s technical density, so it illustrates the logical caution rather than reproducing his theorem.
A beautiful picture can guide a better question
The video draws connections between successive numbers and turns them into branching forms. Such a picture makes shared routes visible: once two starting values reach the same number, their subsequent journeys coincide. An abstract calculation becomes something we can inspect rather than merely recite.
We should still ask what has been drawn. A finite picture shows selected numbers under selected visual rules. Rotating branches can create a coral-like shape, but its organic appearance does not show that nature follows Collatz arithmetic. Nor can the edge of an image tell us whether an unseen branch belongs to the same structure.
Try separating three notes while watching: what the rules guarantee, what the displayed examples show, and what remains to be established. For instance, it is certain that an odd integer becomes even after the odd-number operation. It is directly checkable that twenty-seven reaches one. The claim about all starting values demands a further argument. Keeping those notes separate makes the visual explanation more useful.
Let curiosity keep its standards
Collatz offers an unusual kind of educational satisfaction. We can understand the question without possessing its resolution. That leaves room for curiosity without requiring us to turn a striking pattern into a hidden message or an unsupported certainty. A clear limit on what we know can be as valuable as another impressive calculation.
The broader habit travels beyond mathematics. When a claim feels obvious because every example agrees, ask what the examples actually cover. When someone cites a powerful result, look closely at its conditions. And when an illustration makes an idea feel true, return to the argument underneath it. Evidence becomes stronger when we can describe its reach precisely.
Watch Veritasium’s original exploration for the interviews, evolving diagrams and unexpected turns. Then choose a starting number and follow it yourself. The lasting takeaway is the discipline of allowing wonder and careful reasoning to occupy the same page.
