GROW Island Explained, and the Island Puzzle That Came After

Search for GROW Island and you get two kinds of pages: a walkthrough that hands you the answer in the first line, or a dead Flash embed. Neither tells you what the game is or why it was built, and the walkthrough quietly removes the only reason to play it. Here is the game itself, and a way of thinking about it that survives being read.

It is also the game we were answering. GROW Island has eight items and 40,320 possible orders; so does Island Evolution, which we built in 2026 with the same skeleton and three deliberate changes — a published requirement chart, a score you can read while the run is going, and exactly three perfect orders instead of one. Where those choices came from is easier to explain once you know what the 2007 game did.

What is GROW Island, and who made it?

GROW Island is a browser puzzle that On Nakayama published at eyezmaze.com in September 2007. Jay Is Games dated the release to September 13. It works the way every GROW game works: you are shown a small world and a panel of items, you pick them one at a time, and after each pick everything already in the world tries to advance one level. You never aim at anything, you cannot run out of time, and the only decision you make is which item goes next.

The island itself starts as a lump of rock in the sea. Place things in a workable order and it fills in with roads, water, power and buildings until the ending animation plays. Place them badly and you get a half-built island that stopped growing several turns before the game ended.

Why was GROW Island made for a university?

This one was commissioned. The Shibaura Institute of Technology in Tokyo paid for it as promotion, and Nakayama built it as part of the JOBs section of his site rather than as a numbered entry in the main series. That is why the eight panels are named after engineering disciplines: mechanical, civil, architecture, aeronautical, environmental, electrical, computer and chemical.

The commission changed how the puzzle reads. In GROW Cube you are clicking a pot, a bone and a ball, and nothing about those names tells you what they do. In GROW Island the names describe real fields, so a player can guess that civil engineering lays groundwork and that computer engineering needs something to run on. Jay Is Games made the same point in its review: the descriptive names make the logic feel less arbitrary than earlier entries. The puzzle is still a puzzle, but your first guess is an educated one.

How many orders can eight panels make?

Eight panels means eight factorial, which is 40,320 orders. If you played one order a minute without ever repeating yourself, you would need about four weeks of continuous clicking to see them all. Across the series the count ranges from five panels to twelve, and twelve is 479 million, so eight sits at the friendly end.

Friendly is relative. 40,320 is small enough that people finish the game in an evening and large enough that nobody finishes it by guessing. The gap between those two facts is where the genre lives: the search space has to be reduced by reasoning before it can be searched by hand.

What makes the island setting easier to read than a cube?

An island has geography, and geography implies order. Water goes somewhere, roads connect things that already exist, and a tall structure needs ground under it. None of that is stated in the rules of GROW Island, but all of it is true inside the game, because the artist drew a coherent place rather than a floating abstraction.

That gives you something a cube does not: a sanity check. When you suspect that panel A has to come before panel B, you can ask whether the island would look wrong the other way around. The check is not proof, and it will mislead you a couple of times, though it usually points the right way. Our own island puzzle leans on the same instinct, which is why the evolution chart exists: once you can see which item is waiting on which, the geography stops being a hunch.

How do you work out an order without a walkthrough?

Write down what you saw. Every GROW game tells you the level each item reached at the end of a run, and that readout is the actual output of the puzzle. One run gives you eight numbers. A panel that finished at level one either went in too late or was waiting on something that never happened, and those two cases look different across two runs.

Then treat each run as one experiment with one variable. Move a single panel, keep everything else where it was, and compare the numbers. If you reshuffle four panels at once and the ending changes, you have learned that something you did mattered, which is close to learning nothing. Three careful runs beat twenty frantic ones, and the method we set out in how to play order puzzles without guessing works on any game in the family.

Which panel deserves the first slot?

Not the one that looks most important. The first slot is worth the most growth opportunities, because an item placed first gets a chance to level up on every turn that follows, while the last item gets exactly one chance. So the early slots belong to whatever has the longest chain to climb and whatever the rest of the board is waiting for.

In practice you are looking for one panel that needs nothing and unblocks several others. In an engineering-themed island, ask which discipline the others cannot start without. In a game with a published chart, read the chart as a family tree and find the name that appears in other items' requirements but has none of its own. Whatever that is, it has earned the first slot.

We are not publishing an order here, for GROW Island or for our own game. The solution is the entire product, and a page that gives it away has spent something it cannot get back for the reader.

What should you change after a run that nearly worked?

A run where nothing died and the score stopped moving is a good run with a scheduling problem. Something ran out of turns, which means two adjacent panels want to swap, not that the plan was wrong. Work at the end of the sequence first, since late items are the ones starved of turns.

A run where something visibly broke is a different animal. Look at what was already on the board before the thing that broke arrived, because the hostile check usually runs every turn rather than only on the turn you press the button. Our island has five distinct ways for a placement to kill something, and four of them punish an item that was already sitting there.

Where can you play an island order puzzle now?

The GROW games were Flash, and Flash reached end of life on December 31, 2020. Nakayama moved much of the catalogue to HTML5 before that deadline, so the series did not vanish with the plugin, and several titles got Android and iOS versions in 2014 and 2015. The main series stopped receiving updates after 2020 and no new entries have appeared.

That scarcity is why people keep searching. Island Evolution is our answer to it: eight items on a deserted island, conditional growth, five ways to lose something, and a chart that tells you the rules instead of making you guess them. If you want the background to the whole family first, start with what GROW games are.

Where do these facts come from?

• The GROW (series) entry on Wikipedia, for release dates and panel counts.
• Jay Is Games' 2007 review of GROW Island, for the release date, the Shibaura Institute commission and the eight engineering panels.

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