I presented this talk (online) at ThinkyCon 2025, on November 6, 2025.
Download the slides for this talk (PDF, 13Mb).
Hi! I’m Andrew.
I’m an interactive fiction person. I’ve written a bunch of IF games and puzzle games. And I’ve done some other random things.
I want to talk about a particular quality of some puzzle games: scaling. Or you could call it “zooming”, or a “fractal” property. I don’t think it has a popularly accepted name. You’ll know what I mean, though.
I’m hoping that I can pin down a description of these “scaling” games. That way, we can talk about it, and maybe wind up with some more games that do it.
My father worked for a tech company, so I got a chance to play this game.
I think I played it on a teletype, so it didn’t really look like this. Regardless, I was hypnotized. I wanted to play this game for the rest of my life.
And I did! A couple of years later we got an Apple 2, along with Zork and the first three Scott Adams adventures, and -- yes -- Colossal Cave. (It was available for the Apple under the title “Microsoft Adventure”. Don’t ask.) I grew up with text adventures. I played all the Infocom games as they came out. And then, as an adult, I wrote a bunch. So that’s my life story.
Another thing I saw in 1978 was this movie. Powers of Ten is a short film about the scale of the universe. It starts with a one-meter photograph, and then zooms out to the scale of galaxies, and then inwards to the nucleus of a carbon atom.
The “real” version of Powers of Ten was made in 1977, but the release I saw was a prototype made in 1968. It was playing on a loop at the National Air and Space Museum in Washington DC. I was hypnotized. I didn’t watch it for the rest of my life, but I did think about it a lot.
What do these two things -- interactive fiction and Powers of Ten -- have to do with each other? Nothing, really. But maybe in my head they do, because several years ago I got the idea to make this game:
Hadean Lands is a time-loop (-ish) puzzle game. It’s a game about alchemy. You spend your time finding spell recipes and potion ingredients.
To summarize brutally: some of the ingredients are rare. You only find one dose, so you have to decide which potion you make with that ingredient. But then some obstacles can be overcome with several different potions or spells. That is, there are puzzles with multiple solutions.
So you might get to a certain point, and realize that you can’t make any of the spells needed to progress. But if you restart and use different recipes earlier in the game, you might be able to make one. Thus the time loop.
(Not a time loop, per se, but the same effect.)
I didn’t want the time loop to be boring, though. You’re going to replay the early part of the game over and over in different ways. So I included an auto-solver. When you restart the loop, you remember all your recipes and puzzle solutions. So you can type CREATE FIRE RESISTANCE POTION, and the game will run around and pick up all the ingredients -- if they’re available -- and perform the ritual. This also applies to puzzle solutions. In the screenshot above, I’ve typed OPEN FIRE DOOR. That requires a key, which requires the fire resistance potion, which requires... et cetera. The game does it all as a single command.
The result is that you start out playing a game about finding and learning recipes. But as your recipe book expands, you shift into a different game: now it’s about resource management. Can you combine different sequences of spells to reach a goal without running out of something critical? The recipes were puzzles in the first game, but now they’re single moves.
Indeed, there’s a bit of a third level. As you repair parts of the starship (there’s a starship I didn’t mention) some parts of the map become easier to traverse. Obstacles disappear. So now it becomes possible to reach more parts of the ship in a single loop. It becomes a game about doing that.
Here’s another game that has a similar scaling idea, but in a completely different way.
Baba Is You is a brilliant game that starts out with Sokoban-like block pushing and then takes it to infinity. This screenshot is from the endgame. Each numbered tile is a level to solve. Nonspecific spoiler: you wind up entering levels, solving them, and then turning the entire level into some other kind of tile. A key, for example. Then you can use the key to unlock a door and continue across the map.
So it’s exactly like Hadean Lands, except that instead of recipes made of recipes, it’s map tiles made of map tiles.
In fact, many games have used the “map tiles made of map tiles” idea. Here’s a few examples. All these games have the notion that you enter a level, solve it, and then the solved level becomes an element in a larger-scale puzzle. I’m sure you can name many more.
(Although, let’s be clear, Baba operates on a whole new level. Pun intended.)
Interestingly, Myst almost has this idea. After all, it’s a game about finding game worlds -- in the form of Linking Books -- and jumping into them. But Myst doesn’t have the concept of picking up a book and moving it around. You can only open and close books; the interface just doesn’t offer the option of picking them up. If it did, interestingly, the entire plot would fall apart! You wouldn’t have to go out and collect pages. You could just pick up a working Linking Book and hand it directly to Atrus.
I’m pretty sure this is why, in the recent remake of Myst, all the books are bolted down.
We’ve seen examples; let’s try to describe what we’re seeing.
The concept I’m getting at isn’t just “levels made of levels”. Hadean Lands was “recipes made of recipes”, but that’s not the right description either. I’m not just talking about recursion. There’s a whole category of casual games where you make elements out of smaller elements. But Hadean Lands is something else again.
The idea (this particular idea!) is that you start playing one game, and then slowly realize that there’s another game inside it. Or wrapped around it. And then it happens again. The key moment is the discovery: “I didn’t realize this part of the game existed, but now I see it was always here to be found.”
The feeling is analogous to (though not identical to) the “metroidbrainia” moment: “I didn’t realize I could take this action, but it was always available -- hidden in plain sight.” This might explain why I am obsessed with both of these tropes.
(Might perhaps the two ideas merge at high energy into a Unified Theory of Surprise and Delight? Someone else will have to work that out.)
The other key element is that when you shift your attention to the new layer of play, the original layer folds up out of sight. You shouldn’t have to fussily replay each lower-level puzzle in order to move it around the map. You shouldn’t have to work through each alchemical recipe every time you need the resulting potion. (If you did, the game would take exponential time!) To play the higher-level game, you must be able to think in high-level moves.
So “things made out of smaller things” is a misleading description. It’s really actions made out of smaller actions. This is why Baba is You is such a good example. Everything in Baba is a word, and words are commands. Or parts of commands, anyhow.
Another subject that works well is electronics. Real-world electronics is entirely a field of putting simple devices together into more complicated devices, and then doing it again, and so on approximately forever. Famously, you can start with NAND gates and get other logic gates, then flip-flips, counters, a four-bit adder, and so on.
Many, many games have taken this as a subject. Here’s the first one I played:
Robot Odyssey was an Apple 2 game about wiring up robots with logic games. You could indeed build a circuit, “burn” it into a chip, and then use the chip in a larger circuit. Up to a point, anyway. The Apple hardware could only handle so much simulation; the signal delays got pretty bad.
These are just a few of the more recent takes the idea. The first two, as you might imagine, literally start with NAND gates and work up.
My only problem with this genre is that we only have one electronics tech tree. It’s always NAND to NOT to AND/OR and so on. Even fantasy consoles with imaginary architectures use regular logic gates. So in some sense that field is pretty picked over.
(I guess INTERCAL is a legitimate exception.)
But I want to pull in one more idea. You shouldn’t completely forget about the lower levels of the game. It should be important to maintain awareness of the entire system. Holding the entire shape of the game in your head is the hard part! Which is to say, the fun part.
This means that you should occasionally have to crawl down and make some low-level change to one of your high-level composite tools. Baba does this really well. HL doesn’t do it as well as I’d like, but you do occasionally have to dig into one of your established recipes and make a substitution.
Here’s another game concept that might fit the pattern. Imagine a game where you’re routing current flow across a level. Then the levels fit together into a larger routing puzzle.
SOLAS 128 is a nice example of this kind of puzzle. Each level is a square that you had to route pulses through using mirrors and so on. Then the overworld is a larger routing puzzle. (It also has a kick-ass soundtrack, by the way.)
So we’ll start with that idea, but place it in the context of a toolbox. Say you can pick up levels, once you create them, and use them anywhere on the map. As you solve levels, you’re accumulating tools that you can use. That’s your basic scaling model.
But now say that you have to modify your tools. Perhaps you find some kind of color filter that modifies the pulses, or a better transmission medium that lets the pulses move faster. Now it becomes possible to dive back into your level tiles and stick in filters or wires or what have you. All your tools have to be re-evaluated in light of the new feature. That’s always good, right?
The point is that the lower-level details remain salient -- they’re still important game concepts -- even as you move up to the higher levels. You can never entirely forget what’s going on in the depths.
I’ll give one more example, which applies our scaling idea to a narrative game rather than a puzzle game. (Or a narrative puzzle game. Imagine. I mean, look who’s writing this.)
Imagine a typical choice-based game. You’re at home, you’re talking with your friends, you’re making a cup of tea. An ordinary day. Your choices affect how the day turns out. The game -- the first level -- has multiple endings. And that outcome affects how events go on your street, among your group of friends, over the coming week.
Now you have a second level: what your group of friends does on the street, that week. In this game, your choices are social -- what your friends do as a group. And the week has multiple endings -- which affect how events go in your neighborhood that month. That affects how events go in your city that year. And so on.
Can we dive back down to the micro-level, the day level? Certainly. Perhaps your neighborhood uncovers some new option, or device... (a teacup?) Something that gives you a new option -- yourself, at home, that first day. A new option for the day produces a new outcome for the week, which produces a new outcome for the month. The game expands, all the way up the tree.
This is just a sketch of an idea, but maybe somebody can do something with it.
That’s all I’ve got. It’s not a prescriptive theory of scaling games. It’s certainly not a recipe. But it’s the description that I see, or the common elements, at least.
Have fun with it.
See my contact page for these links.
Last updated November 11, 2025.