The Stack

[the Sequence] is probably my favorite acquisition of the Summer Sale. It’s a little bit Spacechem and a little bit The Incredible Machine, which kind of makes me realize that Spacechem was really a descendant of The Incredible Machine all along. All these games share a paradigm of setting up a mechanism and then letting it rip, trying it out to see if it does what you want it to do. As such, they’re all really about computer programming. But [the Sequence] doesn’t even pretend otherwise; it’s computer-themed, and its mechanisms, although spatially-arranged, are highly abstract.

Each level has the same objective: on a discrete grid, move a little round thing — officially a “binary data point”, but I can’t think of it as anything other than “the ball” — from a source tile that produces an endless stream of them to a goal. Getting four balls from start to goal without collisions is considered to be adequate proof that you’ve created a loop that can continue indefinitely. You accomplish this by placing objects on the grid that can move the ball around. For example, there’s a type of object that can push the ball one square away, another that can grab it from an adjacent tile and then rotate 90 degrees and drop it, another that’s a shuttle that moves one square in the direction it’s pointing every turn and can drag the ball with it, and so forth. These are represented by really well-designed abstract icons that clearly communicate what they do and where and in what direction. I won’t call them “intuitive”, because I don’t think you could guess their function purely from their appearance, but once you’ve seen them in action, they’re highly memorable.

The real trick, however, is that the objects don’t just affect the ball. They can also affect each other, effectively becoming nouns one moment, verbs the next. Like, maybe you need a pushing device in two different places, but have only one available in your inventory. That just means you need to construct a device to move that pusher around. There are even objects whose only purpose is to affect other objects, like the polarity reverser, which turns a pusher into a puller, or reverses the direction that a rotator rotates.

It’s all very much about lynchpins, flashes of insight about what the rules enable. A typical puzzle makes some aspect of the solution obvious, yet seems impossible: “The only thing I’ve got that’s capable of carrying the ball all that distance is a shuttle, but how do I get the shuttle back to the starting position without a polarity reverser?” And it’s kind of impressive how much it manages to force specific solutions through nothing more than the level geometry and the choice of tools. On the few occasions when I’ve been seriously stuck, it was because I was mistaking a puzzle’s intentions on a fairly high level.

The game’s title derives from the fact that the objects take turns. Getting the behavior you want often depends on adjusting the order in which the objects act. I’ve long felt that coarse placement grids are a good thing in contraption games, making the solutions more certain and less fiddly. Giving the player absolute control over the sequence of action effectively does the same thing for time.

Spacechem is one of those games that intrigued me from the moment I saw screenshots, because it didn’t much look like any other game. I probably would have bought it eventually out of curiosity even if it hadn’t been bundled. I had some problems at first getting it to behave properly: even after exiting the app, whatever dreadful things it was doing to my video card persisted in some way, making Firefox show up split diagonally into normal and all-black triangles. But the Steam support forum recommended a small modification to the config files, and that seems to have taken care of it. I’ve spent a couple of hours on it by now, long enough to get a good idea of how it plays, although in some ways it feels like I’m still in a tutorial. This is a game that keeps on introducing new complexity for a good long time.

It strikes me as a game designed by and for computer programmers. At its core, it’s about creating processes for assembling molecules, using a sort of 2D programming language with two threads of execution, like a concurrent version of Befunge. You have a grid and two cursors, one red and one blue, that move along looping tracks that you can set up however you please by placing arrows that make them change direction. Each grid cell can contain at most one redirection arrow per track, and also at most one command. These commands are what you use to assemble molecules out of atoms. You have commands to release an atom or molecule to a specified input area, to pick up and drop whatever is in the same spot as the cursor, to deliver completed molecules you’ve dropped in the output areas, and to use the “chemical bonder” tool, which has a fixed location on the grid, to connect or sever atoms sitting on it.

The chemical bonder is of course not how chemical bonds are made in real life, but that’s okay, because you’re not making real molecules. You’re making chemical diagrams, 2D pictures made of letters and lines, with everything sitting in a single plane and joined at right angles. But unlike the pictures in your chemistry textbook, it’s not just a simplifying abstraction: the geometry of these pseudo-molecules is important to gameplay. Each atom takes up one cell of the grid, making you shift and rotate molecules on the bonder. The developers call this “fake chemistry”.

A working chemical process, once you get it going, moves like a robotic assembly line, and can be made to go at various speeds, the lowest speed being mostly useful for debugging. It seems like every puzzle requires you to make 40 of the target molecule. At first it seemed like this was just a way to give you an opportunity to admire your machine in operation, but it’s also a test to see if it can iterate effectively. Your first iteration isn’t necessarily like subsequent ones, due to timing issues and the possibility of atoms crashing into each other.

Campaign mode is separated into nine chapters, or planets. I’ve completed the first two, but I don’t think I’ll be rushing through the rest, like I do for so many puzzle games. It’s already becoming intimidatingly complex. I suspect that I’m making things more complicated than they need to be, though, due to inexperience with the optimization tricks peculiar to this system.