My thought process while solving the cube

A speedcuber averaging 18 seconds walks through his thought process on several solves, explaining how he evaluates the cross, tracks F2L pairs, and uses look-ahead. He emphasizes that F2L is where most improvement happens and recommends cubers below 40 seconds deliberately slow down to practice tracking pieces while turning.
Chapters
Back when I was starting out, I remember trying to search for a video that would be about the thought process that a speedcuber who is reasonably good goes through as he's solving the Rubik's Cube. I wanted to see what kind of things he's noticing and just generally what he's thinking about as he solves it. Unfortunately, as much as I tried, I was unable to find anything like it, and my suggestions to some of the better speedcubers to make a video like this went unanswered. About 9 months later, I'm now averaging around 18 seconds, so I would say I'm reasonably good to make a video like that on my own. So in this video, I'll go through a couple of random solves of the Rubik's Cube, and I'll try to comment on everything that is going on in my head. Hopefully some of you will get something out of it. As a side note, this video is meant for those of you who are ranging around 20 to 40 or 45 seconds. If you are worse than that, then you might find this a little too fast, so I'm not sure exactly how helpful it will be for you. Without further ado, let's get to it. So I'm scrambling the cube, let's see what happens. Okay, I see I have 2 solved and they're correct with respect to each other. This is quite lucky, actually. This never happens almost. And we've got a nice 2gen thing going on here, which means that I'm only gonna need 2 moves to place it, I mean 2 sets of moves, in this case R and U. So here, I'm gonna place the blue first, so blue up, this up, and fix. So that was a nice 4-move cross, I was lucky. I see that this is a pair right away, so I got lucky, and I'm thinking that this might be a pair here that is misoriented. But I can't act on it because I'm not too sure. So I'm gonna place this first. I'm looking for other stuff. I see this. I saw the blue— I saw the green-red in there, so I know that that wasn't the right corner for it. This is. So I know that this goes with that edge there, so I'm gonna use that. I see that I just paired up these two, so when I place this, immediately after that I'm placing this one, and I see that these two are a pair. So I'm tracking then what happens, and this is a simple case place. Well, from this position, this is probably a 3-cycle. Yeah. Solved. Okay, so I'm scrambling the cube again. Let's see what it gives us this time. Okay, green is placed, so I'm gonna fix it. That will make this orange, red, blue. I see I can place the blue in one move, and if I place the blue, it's gonna orient the orange, and I'm gonna be able to place the orange in one move. However, the problem with this is that if I put the orange first in, then I'm gonna disorient the red, so I don't want to do that. I don't want to do this and then place this because it's gonna disorient this red. So I want to place the red first, so what I would do in an actual solve here is to make a very quick insertion— well, that didn't work out, well, that didn't work out either, anyway, like this, and then I would do my business because now I can just do it without problems, and that's a cross. Right away I see one edge and its pair, corner, so I can use those. I see corner here. I see that those two are most likely a pair. I see these two are a pair for sure. I see that these two are a pair as well, but this is an uglier pair than this one, so I'm gonna do this one first, and I'm hoping that I'm somehow gonna break this pair so that I can use it. I see that there's another corner that entered the U layer there. Now this didn't break up, so I'm not gonna do anything with it still. The edge for this guy is not on the top layer, so it must be here. So I see that this is red, so this must be green here. So I'm gonna pop it out, set it up, and see, I broke those two. That's a good thing, I'm happy about it. I insert the green, and I know which case this is. Place. This is an OLL and PLL G permutation. Solved. Let's see another one. I'm scrambling the cube. This is so exciting. I wonder what's gonna happen. Is it gonna be good or is it gonna be bad? And how bad? Okay, I see red here. Orange, green, and blue. I see that I can place the blue in one move, and it's going to be with respect to the red in its correct place, so that's— I'm very happy about that. This one can be placed in one move, and this one needs two moves to be placed up. However, I can use the orange to, to orient it first, so I can see the cross already this case. First, I'm gonna place the blue. Next, I'm gonna put the green under that orange there. I'm gonna fix the up layer, and now as I place the orange, see, it oriented my green. So I use the fact that I'm placing the orange to orient the green. Anyway, so that was a nice cross. I see that these two are most likely a pair. This might be a pair, so I'm hoping that's a pair. It's not. That sometimes happens, that's okay. I don't see anything else, so I'm gonna work with this pair right now because it's not a good pair, but it's all I have. I'm looking for other stuff. I see that a pair has formed here, which I'm happy about because it's already paired up for me nicely. I see a corner here, but I'm not sure which corner that is. Okay, right now I arrived at a point where I have nothing. This sometimes happens, and Ideally it shouldn't. Anyway, this— I see that these two are a pair. I see a corner here and I see its edge and I'm placing it. I see which orientation this is, I form it, and from here this must be a 3-cycle. This is not a Z, this is not an H, and I know that from experience somehow. So I need to check which— where's the bar, here or here? It's here, so it's this one. Done. Okay, let's do one more. Scrambling the cube. Okay, let's see this. Green, blue, orange, red is solved. That's nice. But this thing here is not too nice. So in this case, what I would do is I'm gonna place this one first. Because by doing so, I take out the blue and I'm gonna be able to place it here, and then I'm gonna go the other way. This was not too little— this was a lot of moves, but I think it would be fast if it was executed, so that's how I would do the cross. Corner, one of these is the one that we want. Hopefully this one, because otherwise we have to set this one up. And it's not it, and it's not even this one, so sometimes that happens, but I'm lucky and I see that this is actually the k— the one that we want. We place them both here. I see edge corresponding to this corner, so I'm gonna solve that right away. I see that these two are a pair as well. I'm tracking them around the cube, and I see I need to do a setup move first, and now I'm placing them, and these two. Of course, case as well. So we have an orientation, and this looks like a G. It cannot be just a 3-cycle of corners because otherwise this would be solved already. So, a G. So I hope this video was at least of of some used to. As you see, most of the thinking happens in F2L, and therefore F2L is the most important part of the entire solve, and it will be the main source of your success as you get better in speedcubing. As a side note, even though it seems like a contradiction, if you are averaging below 40 seconds, you should try to slow down while solving the F2L. Try to look ahead exactly like I tried to show you in this video. Track the corners and the edges around the cube as you do the other moves. If you get stuck, take your time and decide slowly on what you think is the best way to proceed and just go through it slowly. You should find this to be a very useful tip and that will help you in learning F2L. And that's all for now. Bye. So, I filled it. This is an easy— well, it would have been an easy one if I didn't fuck it up. What the hell is this? This is nothing. I screwed up my entire cube. What's going on? Oh, never mind.
More from Andrej

Skill Issue: Andrej Karpathy on Code Agents, AutoResearch, and the Loopy Era of AI

Andrej Karpathy: Tesla AI, Self-Driving, Optimus, Aliens, and AGI | Lex Fridman Podcast #333








