My second-favorite thing about Janice and Stu Turner’s Assembly is that the killer AI is probably right. After a micrometeorite storm introduces a deadly virus to the game’s ship assembly platform, the AI does exactly what every responsible citizen should do when they suspect they’ve been contaminated — it washes its hands. Sure, that involves flushing the station’s oxygen and quarantining the two survivors so they can’t reach Earth, but… when we lose the game, isn’t humanity winning?
Food for thought. At least it made me feel better when I lost over and over again. As for my favorite thing about Assembly, let’s take a closer look.
At a glance, Assembly isn’t quite as space horror-y as it might sound. Killer AI? A luxury liner slapped together in desperation? A ring station on lockdown? Never mind that “lockdown” has taken on a very different meaning over the past couple months, one hallmarked more by boredom and Tiger King marathons than bunkers and cannibalism. Mike Jessup’s crisp artwork doesn’t fall into either category. You get the sense that this station was running quite efficiently before the whole “alien plague” detail. The colors are bright. The icons are legible. Everything seems clean. It’s like how an automobile assembly line conjures an image akin to the service station down the block, all greasy rags and oil sheens, but the real thing is pristine and high-tech. Luxury spaceships were built here. Of course it’s going to be tidy.
Unfortunately — and this is the core of what makes Assembly such an interesting game — assembly lines for luxury spaceships are also meant to be crewed by more than one or two people. In this case, you have a whole bunch of “blueprints,” arrayed like the spokes of a bicycle wheel, into which you must fit the modules that will assemble your ship. The trouble is that your fellow survivor isn’t in the same workstation, so your communications are starkly limited as you attempt to deploy, rotate, and lock the necessary modules into their proper slots.
How limited? Well. Limited enough that the last two pages of the rulebook give suggestions for an ad hoc sign language. Since Somerset and I are the type of people who ruin parties by gushing about how great sign language is for communicating with babies, we stuck to the game’s other proposed rule by only asking a single yes/no question per turn.
It works like this. At the outset the station has twelve blueprints but only one module, and the likelihood of that module being in the right place is quite low. One in twelve, to be exact. In order to successfully escape the station, you need to deploy the remaining eleven modules, rotate them all into their proper blueprint positions, and lock them into place.
Simple enough, right? Not quite, but we’ll discuss the wrinkles later. That’s because even without wrinkles Assembly’s basic quandary is tough to tackle. Both players have a hand of cards, which you can play to take an action. Sometimes you’ll rotate all the modules one or two spaces. Other times you’ll deploy something new, or lock one or two modules into their matching slots, or swap two modules. In any case, this requires coordination between players. Whenever I take an action — say, playing a card to deploy a module — two things can happen. If my partner is also holding a card that lets her take that action, she “verifies” it, but the card remains in her hand. On the table, I take my proposed action. In my head, I now know she’s holding a lock card.
But if she isn’t holding that card? Well, then she’s required to “manually” permit the action. That means she’s forced to discard a card. The action still happens. I still learn a little bit about her hand. But we’ve burned twice as many cards for the same outcome, and she’ll only have two cards at the beginning of her next turn instead of three.
Of course, this is performed against a timer. Three times through the deck and you’re dead. And every turn, you’re only allowed to ask a single question. “Can you deploy a module?” “Can you rotate clockwise?” The pressure is often unbearable, especially as the deck thins and you approach the “scramble,” when the AI will shuffle your remaining blueprints, robbing you of any progress you haven’t locked down.
The thing is, though, Assembly works so perfectly because its puzzle inhabits two expertly interleaved problem spaces. There’s the problem on the table, with all those mismatching blueprints and modules, with the deck edging you closer to a scramble, with additional malfunctions striking whenever you lock particular blueprints. And then there’s the problem playing out between partners, trapped in isolation and only able to communicate in snippets, trying to keep your hand, your partner’s hand, the deck, and the discard pile all straight in your head, along with any multi-step plans you have for the next few rounds; plans which will, by the way, very likely be disrupted by your partner’s own unvoiced intentions. I’d say it’s a parable for dating during a virus lockdown, except Assembly released in 2016 — and anyway, everything seems like a lockdown parable right now.
If that’s too breezy, there are two expansions, both with wrinkles of their own. Glitches adds events that occur whenever you reshuffle the deck, forcing you to choose between terrible outcomes, while Re-Sequence & Override adds two sets of robots that need to be paired, effectively behaving like an additional wheel within the wheel you’re already rotating. But despite multiple layers of complication, Assembly never succumbs to opacity. The problem is always clear, always based on the state arrayed before you, and never hidden beneath any grime, graphical or otherwise. This deliberateness is perhaps the game’s greatest strength. There’s plenty of randomness to consider, but this is randomness that exists to generate troublesome game states that beg for resolution, not randomness that’s constantly nitpicking the decisions you’ve already made.
The result isn’t minimalist, not with character cards and malfunction sets and difficulty variants and everything else it has to offer. But Assembly is something close, a puzzle game that’s both elegant and easily tailored to varying expectations. This is the sort of puzzle I tend to appreciate best. Its parameters are clear. Its core challenge benefits from ideal solutions, but in the absence of ideals novel approaches will have to serve. It knows how to utilize the players above the table as well as the cards upon it, in particular when cooperating partners suffer from clashes of information or intent. And it’s so light, so brisk, that failure is a setback of fifteen minutes rather than an hour.
By which I mean that when you win, humanity loses. That’s okay. The killer AI has met its objective more often than I have anyway.
A complimentary copy was provided.