GROW Cube: Why the Order Is the Entire Puzzle

GROW Cube is the game people mean when they say they played a GROW game. It is also the one they quit, because ten items produce 3,628,800 orders and the game gives no hints at all. The interesting part is that it does give you feedback, in the form of a number beside each item, and most players never use it.

That readout is the whole method, and it transfers between games. We leaned on it while building Island Evolution — eight items, 40,320 orders, exactly three of them perfect — and it is why our game ships a requirement chart instead of a walkthrough. What follows is how to read a run as data, with GROW Cube as the worked example. No solution for either game.

What is GROW Cube, and when did it come out?

On Nakayama published GROW Cube at eyezmaze.com in September 2005. Jay Is Games dated it to September 17. You get a floating cube and ten items arranged along the sides of the field. Click an item and it appears on the cube at its lowest level. Everything already on the cube then tries to advance one level, and the cycle repeats until you run out of items and the ending plays.

Each item shows its current level while the run is going, counting up until it reads MAX. Get every item to MAX and you see the full ending, which is the reason people still talk about a two-decade-old Flash game. Get some of them stuck at level one and you see a lumpy, unfinished cube that tells you, quite precisely, where your plan failed.

How many orders does ten items make?

Ten factorial: 3,628,800. At one run per minute you would need nearly seven years to try them all. Nakayama's smaller MiniGROW games use as few as three buttons, where six orders is all there is and you really can brute force it. The main series runs from five panels up to twelve, and twelve gives 479 million.

Those numbers explain the design. The puzzle expects you to reason your way to a short list of candidates, and the number beside each item is the channel through which it talks back to you while you do.

What is the level readout actually telling you?

An item that finishes below MAX failed for one of two reasons, and they are worth separating. Either it went in too late and ran out of turns, or it sat there with its requirement unmet while turns went past. Both look like a low number on the end screen. They are distinguished by what happens when you move the item earlier: a starved item climbs, a blocked item does not move at all.

That single test is the most useful thing in the genre. It sorts ten items into two piles, and the blocked pile tells you what you have not yet found: something else has to come first.

How do you turn three failed runs into a dependency list?

Keep a sheet of paper. Write the order you clicked, then the ten end levels underneath. Now change exactly one position and run it again. Two runs that differ in one place give you a clean comparison, and after four or five you have a list that reads like "this needs that first" rather than a pile of impressions.

The list is what you are really building. Once it exists, most of the 3.6 million orders are gone, because an order that violates a known requirement cannot be right and you never have to try it. The same procedure works on every game in the family, and we set it out step by step in how to play order puzzles without guessing.

Why do early clicks matter more than late ones?

The first item you click gets a chance to grow on every turn after it, and the tenth gets one chance. So the early slots are the scarce resource, and they belong to items with long chains and to items that other things are waiting on. An item that needs five levels cannot finish from the eighth slot regardless of how well the rest of the run goes, which rules out a large block of orders before you test anything.

Cheap items work the other way around. Anything that needs one or two levels can absorb whichever slot is left over, so identifying those early shortens the puzzle from ten real decisions to six or seven. Our own island publishes its full requirement table in the evolution chart, which turns this from a deduction into simple reading.

What do you lose by reading the order somewhere else?

The ending. Playing a GROW game from a written order takes a few minutes and produces a nice animation. What an afternoon of your own runs produces is the moment the dependency list clicks into place, and a walkthrough spends that for you before you sit down.

This is why we do not publish orders, for GROW Cube or for our own game. The method transfers between games. The answer to one puzzle is worth nothing anywhere else, and it costs the reader the only thing the puzzle had to give.

Can you still play GROW Cube in 2026?

Yes. GROW Cube was Flash, and Flash reached end of life on December 31, 2020, with Adobe blocking remaining Flash content the following month. Before that happened Nakayama moved much of the catalogue to HTML5, and GROW Cube also got an Android release in September 2014 and an iOS release in October 2014, so this particular game came through the transition with more lives than most Flash games had.

What did not survive is the release schedule. The main series has had no updates since 2020, and the last new game of any kind was in 2018. If you want the full catalogue with dates, we listed it in what GROW games are, and the island entry from 2007 has its own history in GROW Island explained.

Where do these facts come from?

• The GROW (series) entry on Wikipedia, for release dates, panel counts and the mobile ports.
• Jay Is Games' 2005 review of GROW Cube, for the release date, the ten items and the level readout.

Island Evolution is an independent game with no affiliation to Eyezmaze or On Nakayama. GROW Cube is their work.