Collatz Conjecture Explained: Why 3n + 1 Is So Hard

Quick answer: The Collatz conjecture says: start with any positive whole number; if it is even, halve it; if it is odd, triple it and add 1; repeat, and you will always reach 1. Named after Lothar Collatz, it is still unproved. Computers have checked every starting number up to 2^71 (David Barina, 2025), and Terence Tao proved in 2019 that almost all starting numbers eventually fall below any slowly growing threshold. A July 2026 “disproof” turned out to exploit a bug in the Lean proof checker.

This is the easiest unsolved problem in mathematics to state. A ten-year-old can play with it, and the world’s best mathematicians cannot prove it. Paul Erdős reportedly said mathematics “may not be ready” for problems like this.

The Rule

Pick a positive whole number \(n\). Then repeat:

\[ n \to \begin{cases} n/2 & \text{if } n \text{ is even} \\ 3n + 1 & \text{if } n \text{ is odd} \end{cases} \]

The conjecture says you always end at 1 (after which you loop 1, 4, 2, 1).

Try It Yourself

StartSequenceSteps to 1Highest value
66, 3, 10, 5, 16, 8, 4, 2, 1816
77, 22, 11, 34, 17, 52, 26, 13, 40, 20, 10, 5, 16, 8, 4, 2, 11652
2727, 82, 41, 124, … , 9232, … , 11119,232

The number 27 is the famous surprise: a small start that climbs to 9,232 before finally falling to 1. That wild, hailstone-like up-and-down is why the sequences are sometimes called hailstone numbers.

Why Is It So Hard?

Two things could go wrong: a number could climb forever, or it could fall into a different loop that never reaches 1. To prove the conjecture you must rule out both, for every number. Halving shrinks numbers and \(3n+1\) grows them, and the two steps mix in a way that looks random but is not. Tools that work for most problems in number theory get no grip.

What Do We Know?

ResultWhat it says
Computer checksPublished milestone (Barina, reached January 2025): every starting number up to \(2^{71} \approx 2.36 \times 10^{21}\) reaches 1. Live project progress, checked November 22, 2026: about \(2.39 \times 10^{21}\). This is evidence, not a proof.
Terence Tao (2019)Almost all starting numbers (in the sense of logarithmic density) eventually fall below any threshold \(f(n)\) that grows without bound, however slowly. This does not show they reach 1 or a fixed bound like 100.
John Conway (1972)Generalizations of Collatz-type rules can be undecidable, which hints at why the original is hard.

Was Collatz Disproved in 2026?

No. On July 25, 2026, formal-verification researcher Ramana Kumar posted an AI-assisted Lean project that appeared to disprove the conjecture without using any loopholes. While examining it, he found that Lean would accept a proof of False no matter what: the “disproof” exploited a bug in how Lean’s kernel handled certain nested inductive types. Kiran Gopinathan reported a minimal example on July 28, and the Lean team released a fix (Lean 4.32.2) the same day. The conjecture was never disproved. The episode is a reminder that proof checkers are software too; see our Lean explainer.

Is There a Prize?

Erdős valued it at 500 dollars on his informal prize scale. In 2021 the Japanese company Bakuage announced a 120 million yen prize for a proof or disproof. It has not been claimed.

Common Misconceptions

“It has been checked so far, so it is true.” Many conjectures held for enormous ranges and then failed. Only a proof settles it.

“Collatz is just a puzzle.” It connects to deep questions in dynamical systems and computation, including the busy beaver problem, where Collatz-like machines block progress on BB(6).

Watch: The Collatz Conjecture in Video

If you learn better by watching, these videos are worth your time.

The Simplest Math Problem No One Can Solve – Collatz Conjecture (Veritasium). The most-watched explanation of the 3n + 1 problem.
UNCRACKABLE? The Collatz Conjecture (Numberphile). David Eisenbud explains the problem and why it resists proof.
The Collatz Conjecture was FALSE (for 2.5 days in July) (Alvaro Lozano-Robledo). The 2026 Lean bug story, told by a mathematician.

Check Your Understanding

Five quick questions to see how much of this has stuck. Pick an answer and press Submit; you will see a short explanation either way.

Answer each question and press Submit to see your result right away.

1. Starting from 6, how many steps does it take to reach 1?
Found a bug? Let us know!
2. What do you do to an odd number in the Collatz rule?
Found a bug? Let us know!
3. What is the highest value reached starting from 27?
Found a bug? Let us know!
4. Up to what limit have all starting numbers been checked by 2025?
Found a bug? Let us know!
5. Why did the July 2026 Lean “disproof” fail?
Found a bug? Let us know!

Frequently Asked Questions

What is the Collatz conjecture?

It states that if you start with any positive integer and repeatedly halve it when even or replace it with 3n + 1 when odd, you will always reach 1. It is also called the 3n + 1 problem.

Has the Collatz conjecture been proved?

No. As of November 2026 it is unproved. Every starting value up to 2^71 has been checked by computer, and Terence Tao proved in 2019 that almost all starting values eventually fall below any slowly growing threshold, which is weaker than reaching 1.

Does a very high peak disprove Collatz?

No. Starting at 27 the sequence climbs to 9,232 but still reaches 1. A counterexample would need a sequence that grows forever or enters a different loop.

Was the Collatz conjecture disproved in 2026?

No. A July 2026 AI-assisted Lean project appeared to disprove it, but it exploited a bug in Lean’s kernel that accepted a proof of False. The bug was fixed in Lean 4.32.2 on July 28, 2026.

Why does 27 take so long?

Starting from 27, the sequence climbs as high as 9,232 and takes 111 steps to reach 1, because many odd steps in a row keep tripling the value before halving brings it down.

Keep Exploring

Sources and Further Reading

Related to This Article

What people say about "Collatz Conjecture Explained: Why 3n + 1 Is So Hard - Effortless Math"?

No one replied yet.

Leave a Reply