[Robbie] Most people are familiar with the Rubik's Cube,

but very few of them have actually solved it.

And that's because, if you don't know the moves,

the Rubik's Cube is one of the most difficult

mainstream puzzles ever created.

Frankly, the three-by-three-by-three is too hard, okay?

It really is.

Now, a lot of that has to do

with the fact that there are literally

billions of billions of ways to scramble this thing.

But what's equally incredible is that with a few weeks

of dedicated practice and a little bit of memorization,

most people can learn to solve any of those scrambles

in a few minutes.

But there are some people who can solve it even faster.

A lot faster.

They're called speedcubers, and the best of them

can whip a Rubik's Cube into shape in well under 10 seconds.

As if that weren't impressive enough,

the single fastest solve ever recorded in competition

is a ludicrous 3.47 seconds. [crowd groans]

But could that time go down?

Today, we're gonna look at why solving

a Rubik's Cube in under 3 seconds is almost impossible.

To find out what it takes,

I spoke to a world record holder.

[crowd cheers]

Got a crash course in technique

from some local speedcubers.

Now, the problem here is, you'll notice that

things are on the edge don't match.

Problem here is that I don't even see

that that's a problem here.

And talked about the mind-boggling math

behind the cube with a computer scientist.

There are more positions on the Rubik's Cube

than there are grain of sand on all the Earth's beaches.

[Robbie] The Rubik's Cube was invented

by Hungarian architect Ernő Rubik in 1974.

It actually started out as an experiment.

Rubik wanted to know if it was possible

to design a cube made of smaller blocks

that could move independently without falling apart.

It was only later that he realized

he'd also created a puzzle.

And that puzzle got very popular.

It wasn't long before people were competing

to see who could solve it faster.

At the first World Rubik's Cube Championships in 1982,

competitors took up to a minute to solve the cube.

But speed solvers like world record holder Feliks Zemdegs

have been been chipping away at those times ever since.

I've been to almost 100 competitions all around the world,

broken a bunch of world records,

and won a bunch of world championships.

[Robbie] One round in competition consists

of five solve attempts.

A computer randomly generates the scrambles

to make sure they're difficult enough.

And each competitor solves the same five scrambles.

[Feliks] You get asked, are you ready?

They'll take off the cover, and then you have

up to 15 seconds to look at the cube and start the timer.

So you get that time to look at it,

then you have to place your hands down on the timer

and then solve the cube

and then return your hands to, like, the Stackmat timer.

Now is probably a good place to mention

that diehard cubers actually care more

about average times than single solves.

The fact is, sometimes you just get a lucky scramble.

A better test of a speedcuber's skill

is to have them solve five cubes

and average the three middle times.

Zemdegs currently holds the average solve record

of 5.69 seconds.

But single solve times still matter.

And Zemdegs used to hold that record too,

with a time of 4.22 seconds.

That was until Chinese speedcuber Yusheng Du

demolished it last fall with a time of 3.47 seconds.

So how unexpected was it for somebody to come along

and break your single solve record by,

like, three quarters of a second?

That was pretty unexpected, to be honest.

Like, in speedcubing, you don't often see

those sorts of insane jumps in time.

You can still get like, a really incredibly quick solve,

just on one attempt if you have an easy scramble

and a bit of luck.

And it just so happened that, you know,

there's so many people going to so many competitions

all the time that something like this

was probably bound to happen,

but we didn't really expect it that soon.

Okay, so let's put speedcubing aside for a second.

Because a lot of people, myself included,

have never solved a Rubik's Cube at all,

in any amount of time.

And that's because truly solving one,

like, from scratch, without any help from anybody,

without watching any online tutorial videos,

it's really, really hard.

The truth is, if you see somebody solving a Rubik's Cube,

they are almost definitely using a memorized

sequence of moves to do it.

This is Tyson Mao, he's a co-founder

of the World Cube Association.

And I asked him to teach me the basics.

I would say there are probably three things

to think about in terms of what makes someone

fast at solving a cube.

The first one is the method.

[Robbie] There are many methods for solving a cube,

but they all rely on something the cubers call algorithms.

Memorized sequences of moves that players use

to solve the cube section by section.

Now, as a rule of thumb, the more algorithms you know,

the fewer moves you'll need to solve the cube.

A beginner who has committed fewer

than 10 algorithms to memory,

might solve a cube in, say, 120 moves.

While an expert, who has memorized hundreds of algorithms,

can solve the cube in closer to 50 or 60 moves.

And, as you might expect, fewer moves can translate

to faster solves.

The second thing that contributes to, you know,

how long it takes you to solve the cube

is how fast you can turn the cube.

The fastest speedcubers in the world

average around 10 turns per second

over the course of an entire solve.

And a lot more than that in short bursts.

Take this 16-move finishing sequence for example.

It looks like this, one, two, three,

four with this ring finger, five, six, seven, eight,

and now you gotta push back with this ring finger.

Nine, 10, 11, 12, pull,

13, 14, 15, 16, push.

The top people can execute those 16 moves

in under one second.

Are you serious?

That's amazing!

The third thing that contributes to how long

it takes to solve the cube

is how long it takes you to process the information.

It doesn't help you if it takes you five seconds

to figure out what the next step is.

Your goal is to try to look ahead

and see what the moves for the next step are

while you're doing the current one,

and reducing the pause between those steps.

[Robbie] And then there's the hardware itself.

As you can see, there are lots of different types

of three-by-three cubes.

From the old-school version, which was clunky,

stiff, and hard to turn,

to fancy new models that spin with ease

and include tiny magnets to help

the bases snap into position.

The actual hardware itself has evolved

to the point now where it's like,

it's really, really good, like,

if you give me a cube from five years ago,

it's probably taking off like a second

off my solve times.

If you're going to be good cuber,

back in the day, you had to learn how to make,

you had to learn how to prepare a cube and make it good.

And so, what that would involve

is you'd take the cube apart, you know,

if there are imperfections,

you might sand some of those things down,

you lubricate it with some silicone,

let it dry, put it back together, re-sticker,

you know, adjust the tension on the screws,

it was a big effort.

Over time, as more manufacturers entered the space,

cubes just got a lot better in quality.

So the kids these days, they don't know

how good they have it.

It's like, the improvement in the cube technology

has caused a change in some of these ergonomics.

That move that I showed you earlier,

there was absolutely no way you could

make a turn with this fourth ring finger.

The way that people turn the cube

and, you know, the hand movements that people use

evolved as a results of cubes getting better.

So Mao taught me a method for solving the cube

that he shows to beginners.

Following his instructions, it took me 45 minutes

to solve the cube for my very first time.

What?!

[laughs]

I think if you spent the next two weeks on this,

you'd probably get your time down to about 90 seconds.

90 seconds, okay.

I'm gonna try to get to a point

where I can solve the cube consistently

using the method Tyson has given me today

in under 90 seconds.

And we'll see how that actually happens--

[laughs]

So I took my new cube and got to practicing.

In the first few days, I went from needing

around 20 minutes to solve the cube to just under 3.

Not bad for a newbie.

But that's still an eternity for someone

like Tiffany Chien, a local speedcuber

who averages just under 10 seconds per solve.

She's so good, she can solve a cube

blindfolded and one-handed.

You just solved a Rubik's Cube with one hand

more than three times faster

than I can solve it with two.

When you pick up a cube, you don't look

the way I look when I pick up a cube.

Which is--

[laughs]

Like, it's manic.

You are actually not going full tilt.

Oh yes, definitely

My hands can definitely move faster

than, like, my brain can during the solve.

So, it's limited not by how fast my hands can move,

than, like, what my brain can see, what my eyes can see.

So, are you exercising that kind of restraint

throughout the entirety of the solve?

So, my strategy in competition

is definitely to like, solve it as a smoothly as possible,

like, with no pauses.

Because I find that if I try

to turn quickly when I'm nervous,

I'll turn very poorly.

I had her critique my solving method.

This is so intimidating.

[laughing]

I'm gonna mess up real bad.

Okay.

Ready?

Okay.

Oh no.

That's a mess up!

[taps]

Damn!

That was a slow one!

One minute, 41 seconds, I'm ashamed.

Okay, so, do you have any tips based on that?

For starters, she said I should try opening

with a more efficient move.

The second tip is more general for the entire solve.

I noticed that you spend a lot of time, like,

turning the whole cube around or like,

using your whole hand to turn the cube around.

So, like, during, while you're doing the second layer,

you see the piece up here, so you go like this or something.

But you can see pieces that are like on the other side

of the cube, so you shouldn't need to do

so many rotations of the whole cube.

[Robbie] So, given what we've learned about speedcubing,

what is the lower limit?

How fast can we go?

To figure that out, it helps to understand

some of the math behind the cube.

And for that, we turn to computer scientist Tom Rokicki.

There's 43 billion, billion positions,

which is 43 quintillion, right?

So it's really a big number.

Rokicki has been fascinated with the cube

ever since he was a kid.

And around the turn of the millennium,

he started puzzling over one of the great

unanswered questions of the cube.

Now, this is a little confusing at first,

but stick with me.

Of those 43 quintillion configurations,

there are some of them, like this,

that are very easy to solve.

If I handed you the cube in this state,

you would know that it takes

just one move to resolve it.

But most of the scrambles on the cube

are a lot more complicated than that.

So the question Rokicki wanted to answer was this:

What is the maximum number of moves

that would ever be required to solve the cube?

No matter matter how scrambled it is.

Mathematicians call that figure god's number,

and it went unknown for more than 30 years,

until Rokicki and his colleagues figured it out.

We used a really fast program with all sorts

of clever tricks, it let us solve

about a billion positions a second.

And then we used a billion seconds of computer time.

Now, a billion seconds of computer time

sounds like a long time, and really, it is.

Except for if you've got thousands of computers,

it's a lot less.

So, what is god's number when it comes

to the Rubik's Cube?

20.

No matter how complicated the scramble gets,

you are never more than 20 moves away

from being completely solved.

And it's usually less than that.

Almost all positions require fewer than 20 moves.

18's the most common.

So your typical scramble you're going to get

from your timer is going

to take 18 moves to solve optimally.

Now, you'll remember that the fastest speedcubers on earth

average about 10 turns of the cube per second.

If you divide god's number, which, remember,

is never more than 20, by 10 turns per second,

and you get solve times of under two seconds.

And look, here's proof that it is physically possible

to solve one in that time.

In fact, this robot can solve it in under 1 second.

So, from a purely mathematical standpoint,

a sub-two-second solve by a human should be possible.

There's just two problems with that line of thinking.

First of all, just because a computer

can quickly identify the fewest number of moves

to solve a cube, doesn't mean a human can match it.

There's nobody out there that can look at this cube

and say, ah, I'm 18 moves from solved

and this one takes me to 17.

That's just not something which humans can do.

And second, even if a human could look at a cube

and quickly identify the fewest number of moves

required to solve it, there's no telling

whether performing that sequence of moves

would be any faster than their usual technique.

That's because there can be a tradeoff

between the number of moves it takes to solve a cube

and the speed at which you can execute those moves.

Ultimately, solving the cube requires

executing physical moves, and the faster that

you can execute those moves,

the less time it takes to solve the cube.

But it's not as simple as minimizing the number of moves.

It is not.

You know, ergonomics and other things come into play.

So, one example is, you know,

this state, which we, even though this,

that there are three stickers here that are not solved,

we call this case the U permutation.

And originally, when people solving this,

there was a 9-move sequence that looked like this,

and this is how people solved the cube.

But then, over time, I wanna say around 2004, 2005,

an 11-move sequence became more popular

because it was just faster to execute.

You know, you want to minimize the number of moves,

but you also want to be able to make those moves quickly.

[Robbie] So what do speedcubers think

the limits actually are?

People have asked me to speculate

what the world record will eventually reach,

and every time I've speculated, I've been wrong.

So, I'll take another shot at it again, you know?

You know, I think low fives is probably

at least from an average, you know, maybe five flat.

[Robbie] As for the limits of a single solve?

In the next five years, I would expect probably

something under three seconds.

Given just enough chances.

My best ever single solve at home in practice is like 3.01.

But then there's a couple of people

who've done, like, under three at home.

Again, just pretty much depends on luck,

like, probably I could do a 2.5,

but it's just when will it happen.

[Robbie] Faster times will come down

to a combination of luck, improvements in hardware,

the development of more efficient methods,

and smoother execution.

As for me, for two weeks, I brought a Rubik's Cube

with me everywhere I went,

and practiced at least 20 minutes day.

I got my single solve time down to 59 seconds,

and my best five solution average

down to one minute, eight seconds.

So I actually made a ton of progress

and you probably could too.

And while I'm never gonna compete

with any of the world's fastest speed solvers,

that's totally fine.

Because what they're doing is already almost impossible.