Still Unsolved

← Interactive · Open problems in mathematics · October 2026

Still Unsolved

Sixteen questions nobody on Earth can answer yet. Each one comes with a small toy that shows what it is about. They are ordered by how easy they are to understand, not by how easy they are to solve. Several of the easiest to state are the hardest to crack.

Bounty is any cash prize on offer. Our shot is my honest guess at whether a small team with computers could add something real: No chance Long shot Real chance
1

The Collatz conjecture ¥120M PRIZE

Asked in 1937 · Lothar Collatz

Pick any whole number. If it is even, halve it. If it is odd, triple it and add 1. Keep going. Does every number end up at 1?

Numbers bounce around wildly on the way. 27 climbs past 9,000 before it crashes back down. Every number anyone has tried lands on 1 in the end, but nobody can prove they all do.

Bounty
¥120 million (about $800K) offered by Bitmasters, Japan, in 2021. Also £1,000 from Bryan Thwaites (1996) and $500 from Paul Erdős.
Known
Checked by computer for every number up to about 268 (295 quintillion). In 2019 Terence Tao proved that nearly every number falls to a small value.
Our shot
Paul Erdős said mathematics is not yet ready for problems like this.
Read more
2

Odd perfect numbers

Asked around 300 BC · Euclid; still open since Descartes

A number is perfect when its divisors add up to exactly itself: 6 = 1 + 2 + 3. Is there an odd perfect number?

Every perfect number ever found is even. Odd numbers can fall short or overshoot (945 overshoots), but none has ever landed exactly. Try some.

Bounty
None.
Known
52 perfect numbers are known, all even. An odd one would have to be bigger than 101500 and have at least 101 prime factors.
Our shot
Two thousand years of attempts.
Read more
3

Goldbach’s conjecture

Asked in 1742 · Christian Goldbach, in a letter to Euler

Can every even number bigger than 2 be written as the sum of two primes?

100 = 47 + 53. 1,000 = 3 + 997. The bigger the number, the more ways there are to do it, and the chart of those counts spreads out like a comet. Nobody can prove the count never drops to zero.

Bounty
None today. Publisher Faber & Faber offered $1 million in 2000–2002 to promote a novel. It expired unclaimed.
Known
True for every even number up to 4 × 1018 (2013). The odd-number version, using three primes, was proved by Harald Helfgott in 2013.
Our shot
The computer checks are already far past anything we could add.
Read more
4

Twin primes

Stated in 1849 · Alphonse de Polignac (the idea is much older)

Are there infinitely many pairs of primes just 2 apart, like 11 and 13, or 17 and 19?

Primes thin out as numbers grow, and twin pairs thin out even faster. Jump the window up to the millions, billions and trillions. The pairs keep showing up.

Bounty
None.
Known
In 2013 Yitang Zhang proved there are infinitely many prime pairs less than 70 million apart. A worldwide team cut that gap to 246. Getting from 246 down to 2 is the whole problem.
Our shot
The known methods are believed to stop at a gap of 6.
Read more
5

Sums of three cubes

Asked in 1953 · Louis Mordell

Can every whole number be written as a³ + b³ + c³, where a, b, c may be negative? Which ones are impossible?

Numbers that leave 4 or 5 when divided by 9 can never work. That part is proved. For the rest, answers can be gigantic: 42 took a planet-wide computer search in 2019. Tap any number below 1,000.

Bounty
None.
Known
Below 1,000, only 114, 390, 627, 633, 732, 921 and 975 are still unsolved. Searches with numbers up to 1016 have found nothing for them.
Our shot
Cracking one of the seven would be a headline, but it took huge distributed searches for 33 and 42.
Read more
solved open impossible
6

No three in a line

Asked in 1917 · Henry Dudeney, puzzle writer

On an n × n grid, can you always place 2n dots so that no three sit on one straight line? Slanted lines count too.

2n is the most you could hope for, because a row can hold only two. Small grids work. Tap the grid and try it yourself, then let the solver show you an answer.

Bounty
None.
Known
Answers have been found for every grid up to 46 × 46, plus a few even sizes beyond. Many experts suspect very large grids cannot reach 2n, but nobody can prove it either way.
Our shot
A clever computer search could push the record past 46. That would be a new result, though not the full answer.
Read more
7

The Erdős–Straus conjecture

Asked in 1948 · Paul Erdős and Ernst Straus

For every whole number n of 2 or more, can 4/n be split into exactly three fractions with 1 on top: 1/x + 1/y + 1/z?

Ancient Egyptians wrote all their fractions this way. Type any number and the solver finds a split. Some numbers make it work much harder than others.

Bounty
None.
Known
Checked for every n up to 1017. Proved for most kinds of n, but a few stubborn families of numbers are still open.
Our shot
The open cases are well studied. A new idea is needed, not more computing.
Read more
8

The party problem, R(5,5)

Asked in 1930 · Frank Ramsey

How many guests must be at a party to guarantee that 5 of them all know each other, or 5 are all strangers?

For groups of 3 the answer is 6. Tap any line between two guests to mark them as friends or strangers. With 5 guests you can avoid a same-type triangle (the starting picture does it). With 6 guests you never can. For groups of 5, the answer is somewhere from 43 to 46.

Bounty
None for R(5,5) itself. Erdős offered $100 and $250 for related questions about how Ramsey numbers grow.
Known
43 ≤ R(5,5) ≤ 46. The upper limit dropped from 48 to 46 in 2024 (Angeltveit and McKay, with heavy computing).
Our shot
Erdős joked that if aliens demanded this number or war, Earth should put every computer on it. The search is astronomically large.
Read more
9

The happy ending problem

Asked in 1933 · Esther Klein; Erdős and Szekeres, 1935

Scatter dots with no three in a line. How many dots guarantee you can always pick k of them that form a convex shape, with no dents?

5 dots always hide a convex 4-sided shape, 9 a pentagon, 17 a hexagon. The pattern says 33 dots for a heptagon: 2k−2 + 1. It got its name because Esther Klein and George Szekeres married after working on it. Tap to add dots.

Bounty
$500, offered by Paul Erdős for a proof of the 2k−2 + 1 formula.
Known
Proved up to hexagons (17 dots, by computer, 2006). In 2017 Andrew Suk proved the formula is almost right for large k.
Our shot
Settling 33 dots for heptagons by computer is far beyond today’s searches.
Read more
10

Graceful trees

Asked in 1964 · Gerhard Ringel and Anton Kotzig

Take any tree of n dots. Can you number the dots 0 to n−1 so the differences along the lines are 1, 2, 3 … n−1, each used exactly once?

Every tree anyone has tried can be numbered this way, which is called graceful. Grow a random tree and the solver numbers it. The orange numbers on the lines are the differences.

Bounty
None.
Known
Checked by computer for every tree up to 35 dots, and proved for many tree families (paths, caterpillars, stars). A weaker version, the Ringel conjecture, was proved in 2020.
Our shot
Pushing the computer check past 35 dots is a realistic search project.
Read more
11

The lonely runner

Asked in 1967 · Jörg Wills

n runners start together on a circular track, each at a different steady speed. Will every runner, at some moment, be at least 1/n of a lap away from all the others?

The shaded arc is the lonely zone around the blue runner. When nobody else is inside it, blue is lonely. Change the speeds and watch it always happen eventually.

Bounty
None.
Known
Proved for up to 7 runners (2008). Each extra runner has needed new ideas. Recent computer-assisted work claims a few more.
Our shot
Computer-assisted proofs for the next runner count are a plausible research project.
Read more
12

The inscribed square

Asked in 1911 · Otto Toeplitz

Does every closed loop you can draw, one that never crosses itself, have four points on it that form a perfect square?

Draw any loop on the sheet. The solver hunts for the biggest square with all four corners on your line. It seems to find one every time, but nobody can prove it must.

Bounty
None.
Known
Proved for smooth loops and many others. In 2020 Joshua Greene and Andrew Lobb proved smooth loops hold rectangles of every shape. Jagged loops that wiggle forever are the open case.
Our shot
The open case needs deep topology, not searching.
Read more
or draw your own on the sheet
13

The perfect cuboid

Studied since the 1700s · Leonhard Euler

Is there a box whose edges, face diagonals, and long diagonal through the middle are all whole numbers?

Boxes with whole edges and whole face diagonals exist. They are called Euler bricks, and the smallest is 44 × 117 × 240. In every one ever found, the long diagonal through the middle misses. Drag to spin the box.

Bounty
None.
Known
No perfect cuboid has been found. Computer searches into the trillions have come up empty, and nobody has proved one can’t exist.
Our shot
Extending the search is doable. Finding one, or proving there are none, is not.
Read more
14

Coloring the plane

Asked in 1950 · Edward Nelson

Color every point of a flat plane. What is the fewest colors you need so that two points exactly 1 unit apart are never the same color?

Drag the ruler, which is exactly 1 unit long. On the 7-color honeycomb its two ends never match. Switch to stripes and turn the ruler upright to see a coloring fail.

Bounty
None.
Known
The answer is 5, 6 or 7. In 2018 Aubrey de Grey, an amateur, proved 4 colors aren’t enough, using a computer-checked pattern of 1,581 points.
Our shot
Amateurs with computers have made real progress here. Pushing it to 6 would be a major result.
Read more
Drag the middle to move, an end to turn
16

The Riemann hypothesis $1M PRIZE

Stated in 1859 · Bernhard Riemann · Clay Millennium Problem

The primes look random. Do they follow a smooth curve as closely as possible?

The blue staircase counts the primes up to x. The dashed curve is li(x), the smooth prediction built from Riemann’s work. The hypothesis is the same as saying the gap between them never grows faster than about √x · ln x. The shaded band shows that limit.

Bounty
$1,000,000 from the Clay Mathematics Institute.
Known
Trillions of zeros of Riemann’s zeta function have been checked, and all of them are where the hypothesis says. Thousands of other theorems are proved assuming it is true.
Our shot
Widely called the most important open problem in mathematics.
Read more