FRACTRAN is a beautiful language but takes a bit of work to understand. Mark Chu-Carroll of Good Math Bad Math, described it as “absolutely insanely difficult to program in, but based on one of the most bizarrely elegant concepts of computation”

FRACTRAN was devised by John Conway, creator of the Game of Life, a cellular automaton sometimes referenced in esolangs. A set of rules for the progression of black and white pixels, it takes on complex behaviors under the right conditions, serving as a metaphor for the rise of biological complexity from simple chemical traits, and also of the rise of computational systems from simple rules (e.g. see this insanely complex implementation of Tetris).

FRACTRAN, similarly, is simple and powerful. It was developed before the concept of esolangs emerged, first published at Proceedings of the 1972 Number Theory Conference held at the University of Colorado. It is a and is a theoretical model for computation and does not require or particularly encourage use of a computer to use.

A FRACTRAN program is a sequence of comma-separated fractions. The input to a program is a single number. It is best understood with a simple example, for instance an adder.


In FRACTRAN, we treat each number as the product of its primes. If we start with the number 144 (meaning this is the input to our program), we can factor this to 2 x 2 x 2 x 2 x 3 x 3, or  2^4 * 3^2. In FRACTRAN, the two primes (2 and 3) are variables, and their exponents are their values. We can think of our input this way:

var_2 = 4
var_3 = 2

Serving this to the program 3 / 2, in the first iteration we get:

144 / 2 * 3 = 216 or 2^3 * 3^3, or:

var_2 = 3
var_3 = 3

We have multiplied by 3 (the numerator) and divided by 2 (the denominator), incrementing the exponent of the numerator and decrementing the denominator.

Running our value (now 216) through this fraction again gives us 2^2 * 3^4, or 324. We keep doing this until the value can no longer be divided by 2, at which point we stop, giving us the final value of 3^6, or 729.

So long as we give this program a number that’s a multiple of both 2 and 3, it will add the exponents of 2 to the exponents of 3. 

It may seem like this is a hard way to add two numbers, but we’ll next see how more complex algorithms are enacted in FRACTRAN. It’s important to keep in mind that even though the input is a single number, it’s a set of variables (the primes), with their exponents the data. Also, there’s nothing special about 2 and 3, we use them because they’re the two lowest prime numbers, giving us the smallest possible numbers that will work. If we were to create the program 5 / 3, we could give it input in the form of numbers with exponents of 5 and 3. There’s no magic here.

At this point we’re ready to look at the complete set of rules for FRACTRAN:

  1. For the input value, we consider each fraction in order 

  2. If multiplying it with that fraction will give us an integer (if our number is a multiple of the denominator), we multiply it and go back to the beginning of the list

  3. We stop when there are no more fractions that will give us an integer result

Let’s look at an example with multiple fractions to see how to implement algorithms.


The multiplication program looks like this:

385/13, 13/21, 1/7, 3/11, 7/2, 1/3

(This example is taken from Mark Chu-Carroll’s post here)

This, factored is:

(7 * 11 * 5)/13, 13/(3 * 7), 1/7, 3/11, 7/2, 1/3

For the first fraction, if our input can be divided evenly by 13, it is so divided, and also multiplied by 7, 11, and 5. So in other words, if it has a var_13 (an exponent for the prime number 13), that expinent is decremented, and var_7, var_11 and var_5 are all incremented:

if (var_13) { var_13–; var_7++; var_11++; var_5++;}
elsif (var_3 && var_7) { var_3–; var_7–; var_13++;}
elsif (var_7) var_7–;}

You can see how this is a lot more elegant written as fractions!

If this is hard to parse, start by considering the fraction 7/2; we’re transferring 1 from var_2 to var_7, but only if we can’t divide by 7 (the third fraction), or any of the other steps that come before it.

Then consider the first two fractions. Once we have a multiple of 7 in the numerator (which we think of as var_7 > 0), the activity trades back and forth between the second fraction and the first. We add a 13, then immediately get rid of a 13 – this trades control back and forth between the two steps, initiating a decrement of the 3 (which servers as our counter), increments the 5 (our output of the program) and the 11 (our temporary storage for what used to be in the 3). Once the counter (3) is exhausted, we hit the 3/11 fraction repeatedly, which transfers all of the value of 11 back to 3. Since we never re-incremented 2, we end up with one less value for 2, and 3 is “refilled” back to its original value. This puts us back to the start state, with 2 holding all of what was in 3, and 3 only decremented by one.

If brainfuck makes sense to you (and if it doesn’t, you might start here), you can see intuitively how this might be Turing Complete. We can increment and decrement variables, we have conditionals, program flow, and as many variables as we need, even if these are tied to together in a way that’s awkward for any practical programming.


Conway used this system to build a prime number generator, explained here on Wikipedia. However, primes are so much a part of FRACTRAN. What if we want to break away into other types of programs?

For an astonishing example of the language in action, view this FRACTRAN interpreter written in FRACTRAN itself! Because so many variables are used, we get into some crazy looking numbers:

19228559906211175423748638669483  /  5651764272356337957182213393
4280910472672668030345419  /  2731640537629936180368397
4280910472672668030345419  /  2813898220067976918547621
8966417131313137949541572921  /  5923379139766748474650011041
5278557919448419272929  /  2779624185718793277639611
5069368041531044665019  /  3427403434918364090801
4229240572827076502191649  /  3317872566159321829577
5306449903170735887317  /  3509551586487645786719
4273700394880449185526121  /  2809690409193150378345599
19681282009657046999693471426063  /  5765772982091987384129572961

While FRACTRAN takes some getting used to, it’s a great illustration of the abstractness of computation, the way it can arise in systems removed from  binary, digital systems. We return the algorithm to its roots in math, in a way that can be carried out by hand.