The Stack

It’s taken me a while to admit this, but I think that I’m going to have to do something I’ve never done before: shelve a main-line DROD title while it’s still incomplete. Oh, I’ll get back to it. But for now, I’m playing other things, including one thing that I had previously intended to start after beating The Second Sky: Wonderquest, the only known DROD imitation outside of that one room in Frog Fractions 2, the one that looks misleadingly like Nethack. FF2 does that a lot, hiding a game pastiche under the skin of a different game pastiche.

Wonderquest isn’t like that. It doesn’t hide what it’s imitating at all. It’s about as blatant an imitation as you can get. You hit orbs to open doors, your initial enemies are roaches which move exactly like DROD roaches, there are roach “spawners” that only differ from DROD roach queens in that their spawn cycle is 24 turns instead of 30. The initial player character even wears the same color of shirt as Beethro. So let me introduce Wonderquest further by describing how it’s different from DROD:

Instead of a dungeon, Wonderquest is set in a jungle. It’s functionally equivalent to a dungeon, just replacing stone walls with impassable forest. The player characters are from various places on Earth, and have no idea how they wound up there. Instead of a Smitemaster, your initial character is identified as a “Beggar”.

It’s not nearly as slick as DROD. The tiles are noisy and repetitive. The music and sound effects are low-fidelity. The controls are a little awkward, even after you rebind them to match the DROD keys — holding down a movement key to go faster here results in zooming out of control, and is almost always a bad idea. Movement is animated continuously rather than moving in discrete steps from tile to tile, which probably seemed like an improvement to the designer, but I find it just muddies my understanding of what’s going on. The dialogue, revealed entirely by stepping on scrolls, is pretty terrible. The whole thing is a little amateurish, but amateurish in a way that I find a little comforting, hearkening back to a less commercial age of indie game development. Retro, but in a way that doesn’t feel like it’s doing it on purpose.

This extends to the puzzles as well, at least in the early stages. Lots of repetitive action, with enemies in quantities that don’t make the puzzles harder or more interesting. I suppose it’s really no worse than some of the stuff in King Dugan’s Dungeon, but even that comes as a shock after playing so much of The Second Sky. For example, here’s a puzzle with two chambers, where you have to do exactly the same thing to both chambers. TSS had a number of puzzles that looked like this at first glance, but some slight asymmetry — maybe even just the placement of the entrance — meant that the seemingly-identical parts had to be approached completely differently. Here, there’s no such subtlety.

That said, there are ways that it’s going beyond DROD, even in the first few levels. As I noted before, Wonderquest had tiles that change your weapon before DROD did, something that’s also true of force arrows that can be disabled like doors. Also, there’s an element that Wonderquest introduces early on, a rolling ball that moves every turn in the same direction, unless stopped by a wall or redirected by a force arrow, potentially pushing crates or crushing monsters as it rolls, or activating orbs and pressure plates. This gets used a lot, sometimes in large quantities. There are rooms dominated by the balls bouncing around, a flurry of mechanical activity that ignores the player. Balls in isolated tracks are the main thing powering things like time limits and cyclical door-opening. One level in TSS had something similar to balls, rodent monsters called “lemmings” (presumably inspired by the game Lemmings) that move forward, ignoring the player, destroying everything destroyable in their path, but they were never as ubiquitous as the balls are here, or used with such versatility.

As I write this, I’ve reached a level with giant butterflies that move in knight’s-moves, an idea found nowhere in DROD — probably because there isn’t a lot you can do with it, puzzle-wise. But Wonderquest has lower standards for its puzzles, and is content with providing multiple rooms where you’re simply mobbed by a swarm of the things.

I know there’s a great deal more to come. I’ve played a few levels further than this, years ago, and remember some additional characters, with unique abilities. Also, the main UI has spots for a resource-gathering mechanic that I never played far enough to see. I hope to see it this time around.

More than a week since my last post! I’m still playing The Second Sky, but I’ve slowed down considerably. I don’t play every day, and when I do, I consider it a good session if I finish even one room. A far cry from my reaction to The City Beneath.

It’s all because the puzzles — or at least, the endgame puzzles, which are still somewhat optional unless you’re stubborn enough to insist on the best ending — are so much harder this time around. Before this game, Journey to Rooted Hold was definitely the most difficult, even if you ignore the Challenges that consumed so many days on this blog. And given how much easier Gunthro was than The City Beneath, I really wasn’t anticipating such an extreme reversal. But it makes sense. We’re at the final chapter of the final chapter. If you’ve come this far, you’re an expert on everything the puzzle designers can throw at you. That means they can freely exploit minutiae like movement order.

See, every monster has a number, which you can see when you right-click on it. When you move, the monsters move in order from lowest to highest, and this can have effects on what happens. Consider a row of Roach Queens lined up against a wall, as in the screenshot. Roach Queens run away from the player, and prefer north/south to east/west movement when they’re blocked from moving diagonally. What happens when you move one square to the north? You can’t answer that from the screenshot; there isn’t enough information. If they’re arranged from east to west, such that the easternmost one moves first, then each roach will simply move one square southwest, except for the last, which will have no empty space to move into. The next turn, the same will happen again, except the last two will be blocked, and so forth until they’re all hugging the western wall. But if they’re arranged from west to east, the leftmost one will move south first, then the next one will be unable to move southwest and instead move south, and so forth. They’ll remain in a horizontal row and march south in lock-step.

The latter is in fact what happens in this puzzle, and it’s pretty easy to keep the roaches neatly ordered as you chase them south, then back north, then southwest, timing your actions to their spawn cycle. Which is good, because you need them in a row at the very end, for reasons I won’t go into. But before you get there, the geometry of the room acts against you, splitting the first and last roaches out of the line. By the end, you need to get the roaches in a different order, probably 3124, presumably by using the right-hand walls in just the right way. The uncertainty of my words comes from the fact that I haven’t yet solved this puzzle.

I’ve known since the very first DROD that movement order can make a difference, but I never needed to pay attention to it precisely. “Move these guys against walls some to get them to shuffle around” was always adequate. Here, I don’t just need to know movement order, I need to deliberately manipulate it. And that’s what this chapter is like.

One of the larger (but not the largest) of the very large levels in The Second Sky‘s eleventh chapter is all about Fluff so let’s talk about Fluff. Fluff is depicted as cottony white clouds, but these clouds are as impassible as any other obstacle. It’s like tarstuff, in that it needs to be two tiles thick at all points, and when it isn’t, the thin parts break off into monsters, called Puffs. Or maybe it’s wrong to classify Puffs as “monsters” — they don’t count as such for level completion. Sometimes completing a Fluff room means fleeing just before the massed Puffs overwhelm you. Which they can do, because, like Wubbas and Serpents, you can’t hurt them with your sword.

Oh, they’re not unkillable. Explosions will take Puffs out, and if you have access to a pushing weapon, you can crush them against walls or obstacles, including other Puffs. Similarly, Fluff can’t be cut with sharp weapons, but only with blunt weapons or explosives. But not all Fluff rooms provide these things. Instead, the chief weakness of Puffs is that they’re slow. They only move once every five turns. They’re the only creature with this property, and it completely changes how you deal with them in puzzles. Among other things, it means they can’t cross Hot Tiles unless they’re pushed. Leading a Puff to an opportune spot involves a lot of waiting, and often you have to deal with other things while you’re waiting, so it’s more like just making sure you wind up in the right place relative to the Puff after every fifth turn.

Fluff, I say, is essentially a tarstuff, but it’s a peculiar form of tarstuff. It’s the only form of tarstuff that flies; in both Fluff and Puff forms, it can go over pits and water, and doesn’t trigger pressure plates. Also, there is no such thing as a Fluff Mother. Instead, there are vents, an architectural feature that cannot be killed. If a vent is covered by Fluff, then that Fluff expands every 30 turns by the normal Tar Mother expansion rules. If it’s not, then the vent simply emits a Puff instead. But vents tend to become covered by Fluff over time, because of Fluff’s most notable unique feature as a tarstuff: it can reform. Puffs adjacent to Fluff can merge with it. If four or more Puffs gather in a tar-stable shape, they meld into new Fluff.

I think the most important thing about Fluff, though, is that it hates all life. Puffs are the only thing that will actively pursue and kill both the player and the monsters — which is another good reason for excluding puffs from the “monster” category, because they don’t kill each other. (Slayers will sometimes kill monsters in pursuit of the player, but only because they’re in the way. They don’t hunt down monsters the way Puffs do.) There are multiple rooms with no monsters other than Puffs and the passive and immobile Brains, where your goal is to bring them together.

The great thing about all these unique features is that so many of them can be used in puzzles as either obstacles or solutions, depending on context. Puffs being slow is a good thing, until you need to get them somewhere within a time limit. Puffs attacking monsters is a good thing, unless you need the monster alive to weigh down a pressure plate or something. Puffs forming into new Fluff is good if you need to create an obstacle to keep that monster standing on the pressure plate, but bad if it blocks your only way out of the room. Vents are bad if they fill the room with inescapable Fluff, but good if they’re the only way to kill monsters. This is rich puzzle fodder, exactly the sort of thing DROD thrives on.

If a Puff dies over water, it lets out a burst of cold that freezes its tile into thin ice, which is equivalent to a trap door: you can walk on it once, but it collapses as you step off. I learned this in the first Fluff level, but there’s so much about Fluff that’s peculiar that I had completely forgotten this one detail the next time I needed it. I think this is the first time I’ve found the monster descriptions in the help menu really useful.

The Second Sky is the largest DROD by a very comfortable margin. That much is clear to me now. It’s like the designers, knowing that this was going to be the final episode, decided that they had to use all their ideas now. The thing is overflowing.

Previous episodes grouped rooms into levels, but TSS additionally groups levels into chapters. I suppose that’s essentially what Gunthro and the Epic Blunder did, too, but this time it’s more formalized, and there are more chapters. A chapter can be linear, going through a sequence of levels one after the next, or it can be open, just giving you access to a bunch of new levels and letting you flit between them as you like. I’ve been doing a lot of such flitting about, because the puzzles in this game aren’t just numerous, they’re also very hard.

I said earlier that I was hitting the end of the game, and just taking some time to finish up side-quests before finishing it off. The game encouraged this notion by providing an isolated campsite where I could go to wait out the Turning whenever I was so inclined. Obviously I wasn’t going to do that — not without finishing everything I had started! So over the past few days, I have taken care of unfinished business, even going back and clearing out the Tar Recycling Annex, which has nothing to do with Beethro’s goals at this point in the game but which does have a RCS stamp for my collection. It was here that I needed a bit of a nudge from the Caravel Forums hint boards, but only for one of the four rooms I had previously given up on.

And, having done all that, and having set the plot in motion again, I find that there’s an entire additional chapter, with some really big levels. I’m not as near the end as I had thought. The campsite is still there, in case I want to give up, but nah.

Without going into too much detail, Beethro’s plan for saving the surface-dwellers from the Turning in the chapter I just finished involves posing a really difficult question to the Truth Vessels, one that they’ll need to keep spawning more Truth Vessels sorcerer’s-apprentice style to answer. It’s another of those metaphor-for-game-design moments, Beethro trying to come up with an adequately difficult puzzle for a whole bunch of other people to solve. And there’s a neat bit of resonance there, because if the player decides to end the game early, it means Beethro gives up on looking for really tough puzzles for his audience, and the consequence for the player is that you miss out on a bunch of really tough puzzles. We are the Truth Vessels.

I’m in a position in The Second Sky that’s more familiar from JRPGs than from puzzle games: very near the end, but holding off on completing the game because I want to finish more side-quests first. In a RPG, there are usually practical justifications for this: completing those last few quests could give you items or other boosts that help you against the final boss. At the very least, you can expect to get a little extra XP in their pursuit. That doesn’t apply to DROD puzzles. Neither will the bonus puzzles become unavailable after I win. Nonetheless, it is lodged in my brain that this is the proper order to do things in. Side-quests, then victory, and then, because this is DROD, going back to hunt for the secrets I missed and unlock the Mastery area.

One thing I discovered in my last session: a secondary office for the new First Archivist, who Beethro calls “Arky” to distinguish him from the First Archivist who attacked the surface, containing a note explaining his puzzle design MO. I should note that Beethro and Arky are reconciled now; towards the end of the story, Beethro finally gives him the apology that was all he wanted all along. Beethro’s really grown as a person over the course of this episode. But even after he’s your friend, there are still “Arky rooms” to contend with. The “Inventory” room back at the train hub even tracks them as a special category.

Arky rooms are always secret rooms, and thus optional. Their hallmark is a note from Arky describing what makes the room impossible to solve. An actually impossible room is trivial to make, but the point is that these rooms look like puzzles, and this tricks delvers into trying to solve them. In fact, the rooms are perfectly solvable, and Arky’s explanations of why they’re not contain hidden false assumptions. The effect is to make Arky seem humorously incompetent, and this is heightened by the way that the notes describing tricking delvers have been left around for delvers to find.

There’s one other purpose for the notes: misdirection. If it weren’t for the notes, the player might not notice that the room is “impossible”. There’s one note that describes how opening the room’s tar gate requires clearing three invulnerable 2×2 bocks of tarstuff with only one powder keg. I read that note on entering the room, and sure enough: the room has blocks of gel, tar, and mud. The sole powder keg could be placed between two of them, but had to miss the third. I assert that it is the effect of the note that it took me as long as I did to realize: Hey, wait a minute, a 2×2 block of mud isn’t invulnerable! I can clear that with my sword! It was like the room’s punch line. Arky seems all the more incompetent for making such a stupid mistake, but then, through the note, he managed to pass his stupidity on to me. Perhaps he’s cleverer than he seems?

The note in the secondary office says that the main trick behind his puzzles is to make ones that he personally can’t solve. If someone else finds a solution, he pretends that it’s what he had in mind all along. If no one finds a solution, it just makes him seem cleverer than everyone else. And it’s got me thinking: This actually might not be a bad approach for puzzle design in DROD. Obviously you want to not actually release puzzles that no one has been able to solve, but the “Let someone try to solve something I think is impossible” part? DROD players have proven their ability to find solutions that the designers didn’t think of. That’s how we got the kill-the-Slayer Achievement in Journey to Rooted Hold. Or consider taking the same approach with yourself as the player: design a puzzle, then remove something that the solution to the puzzle relies on and try to solve it anyway. I suspect that some similar process is behind all the theme-and-variations puzzle designs in this game. I mentioned a pair of levels called Easy Way and Hard Way. I have since discovered that it continues into Harder Way and Hardest Way, four sets of the same puzzle designs with the solution to each made impossible in the next.

One of the big surprises in The Second Sky is that the Empire has an underground rail system. It was built mainly to facilitate the collection of surface-dwellers, but Beethro can ride it. He doesn’t even have to trick anyone or sneak on board or anything like that. Because he still holds the position of First Slayer from back in The City Beneath, the Empire’s resources are at his disposal, to some extent, when they aren’t trying to kill him. Once again, it must be remembered that the Empire has no coherent system of policy, and Beethro can easily be considered an enemy of the state one moment and a VIP the next.

After a certain point in the story, after helping Tendry get into some trouble, Beethro flees the Patronage’s robots —

There are robots in this game. I haven’t mentioned them before, but I’ve been seeing a lot of them lately. The game calls them “constructs”, to fit them a little better into the basically-still-a-fantasy-despite-all-the-sci-fi-stuff setting, but they’re clearly robots. They’re a little like smarter versions of golems, in that they leave a pile of debris when killed, but with one major difference: until you clear the room, robot debris comes back to life every 30 turns. They remind me of the trolls in Nethack in that regard, and, like Nethack trolls, one way to keep them from reviving is to push them into a body of water with a pushing weapon. Another is to push them onto tiles infested with oremites, which they otherwise avoid. That’s a good example of how the game creates new exploitable complexity through special cases in combinations of elements, something that also reminds me of Nethack. There’s so much of this going on in the game that I don’t have time to describe it all in the kind of detail I gave to the earlier episodes.

Beethro flees the Patronage’s robots by getting on a train without an intended destination, just “Get me out of here”. He winds up at a forgotten station, which becomes a hub for side-quests. Once you have access to that, you can go anywhere in the world — okay, I talked about a world map before, but it turned out to be divided into sections, with only one section available at a time, and that limitation is gone. (Except that it won’t take me to the distant past, which is a bit of a shame, because I still have at least one unsolved puzzle back there. But we have the Restore menu for that.) In fact, the sections I passed through before seem to have sprouted some new levels, where I can go to solve bonus puzzles for collectibles, which are tracked at a special room in the train station. One of these collectibles: RCS tokens. I had found one of these earlier, before I had its context. “RCS” is the name of the train system, and what the tokens do is give you access to rooms containing nonogram puzzles laid out on the floor. Solve these, and your reward is more bonus levels.

I commented before about how each title in the DROD series does more with the bonus content than the last, and I was wondering how The Second Sky would manage to one-up Gunthro and the Epic Blunder, with its trail of clues leading to an entire extra level. It turns out that something like half the content of the game is extra levels this time.

The thing I really like about this is that it gives the game a sense of breadth. We’re not in a linear sequence of levels any more. We’re out exploring, pursuing whatever challenges we come across. It’s a design pattern I associate with JRPGs — most of the Final Fantasy games start off linear, but then you gain access to an airship and you’re suddenly free. I find it significant that the opening up of the DROD world is also accomplished by obtaining a vehicle, albeit one more suited to the largely subterranean setting.

Welp, I’ve seen the secret at the heart of the world.

It’s one of those peculiar moments that calls the mindset of the whole game into question, like the end of Prince of Persia (2008), or confronting Gehn in Riven. As I anticipated, Beethro gets back to his own time just in time to stop Former First Architect from starting the Turning early. Beethro doesn’t even really have to do anything once he’s there; just showing up is enough to derail FFA’s plans. The thing is, though, this just stops the Turning from happening ahead of its proper time. Its proper time is 98 days from now.

I’ve mentioned before how Beethro is, at core, a seeker of knowledge, or at least of explanations. His chief motivation, from Journey to Rooted Hold onward, is a desire to find out what’s really going on. Heck, that’s why he went back underground in the first place: he just had to know what was beyond that unopenable door in King Dugan’s dungeon. The Second Sky has been surfacing this side of his personality even more than usual. And so it is that, on meeting the mythical being responsible for the Turning, a being made visually impressive simply by existing at a much larger scale than the rest of the world, his reaction is to ask a whole lot of questions. “Tell me about the Turning”, he insists, “I need to know!” And what he learns is that it’s inevitable. It’s not a product of villainy, not something he can fight or puzzle his way through. It’s just a part of how his world works, a pattern far older than his civilization, no more stoppable than the tides.

This makes Beethro contemplative. What do you do when you know your world is about to end? “I always assumed there’d be something useful to do”, he says. “Maybe all that’s left is to be ready for the worst.” For once, he doesn’t bother asking the Truth Vessel that stops by any more questions. It’s a sense of resignation you don’t often see in videogame heroes.

But he does at least decide to go consult the Patrons, on the basis that they seemed to be on his side before, and to have some kind of plan, and any plan is better than no plan. The player who’s been through all the DROD games and has seen a number of events from points of view other than Beethro’s has a better idea of what’s going on: the Patrons are already in the process of evacuating the surface. When this was revealed back in DROD RPG, I thought they were doing it to protect everyone from First Archivist’s army, but no, it’s always been about the Turning.

However, nothing can go smoothly in this world, due to either Beethro’s interference, or the Empire’s incompetence, or both at once. General confusion down below, helped along with suspicion about surface-dwellers, has resulted in the erection of defenses intended specifically to keep Beethro from returning to Rooted Hold. Meanwhile, Tendry from DROD RPG has a cameo in which he declares his intention to free his countrymen, who he still thinks of as prisoners rather than evacuees, and Beethro, not knowing any better than him, wishes him luck. Things are primed to go wrong in the grandest way. And once they do, Beethro will have something useful to do again: picking up the pieces, resolving the mess he helped to create.

After sending you through the heart of the temporal storm, The Second Sky gives you a break. Beethro wakes up on a beach — something that the game makes a point of identifying as cliché — and we get the first puzzle-free wandering-around-talking-to-people segment in quite some time. It leads back into more puzzles, of course. A couple of the NPCs have fetch quests for you, sending you to puzzle dungeons on the periphery of the beach area. But it’s all a great deal less tense than the previous chapter.

I don’t think it’s less difficult than the previous chapter, though.

Partly it just feels less intense because of context and presentation. You’re not in a broken world on the verge of collapse any more. The art is a great deal gentler, almost pastel in its softness. I haven’t said much about the art in DROD before, but it’s gotten really good over the course of the series. The original DROD tiles were brutal and garish, and that art style is still available when it’s what the level designer wants, but now we’ve got excellent small-scale pixel art, delicately detailed, coupled with mood-setting lighting effects. Important objects, such as monsters, gates, bombs, etc. have a simpler and more stylized look than the backgrounds, with higher saturation colors, the better to stand out. Anyway, just the fact that the beach dungeons are well-lit makes a huge difference to the experience, letting you see all the detail clearly.

But there are also ways of varying the intensity of the puzzles that are orthogonal to the difficulty. Tense puzzles put you in danger. They send monsters after you, and arrange the terrain so you can’t just bottle them up with obstacles. They impose time limits. They put you in the position of continually reacting to things, and try to overwhelm you with urgency. This is why the Temporal Aumtlich was such a good foe for high-tension puzzles: by splitting, he easily adds new complications when you thought you had him under control. A relaxed puzzle is characterized by stability. Monsters tend to be immobilized, or at least locked away until you decide to let them out under your own terms. Rather than overwhelm you, they give you the time you need to prepare.

The difficulty in a tense puzzle comes from trying to avoid getting killed while accomplishing your goals. The difficulty in a relaxed puzzle comes from your goals appearing to be impossible. Sometimes a puzzle will be so stable that it’s hard to see how you can change things at all. By my earlier terminology, tense puzzles are tactical and relaxed puzzles are strategic.

In the classic Sega ninja arcade game Shinobi, the player can use a special “ninja magic” attack once per level, clearing the screen of enemies. Depending on the level, you get either a lightning storm, a whirlwind, or a “doppelganger” attack, in which duplicates of the player burst out from him and bounce around with a thundering noise. Despite being surrounded by other ninjas, the player character has exclusive access to these magics — until the final boss. He has all your spells, and uses them against you. Except, because the game is designed to be winnable, his versions are weaker than yours. His whirlwind attack is dodgeable. His lightning is localized. His doppelgangers emerge one at a time, and can be fought and overcome like any other ninja.

This was brought to mind by events in the swirling temporal maelstrom that Beethro finds himself in after jumping through a few time portals. It’s a place of chaos: precipices over a foggy void, lighting that’s dim and erratic enough that I can’t tell what color a gate is until there’s a flash of lightning, segments of floor that appear and disappear. The ever-useful clock in the UI doesn’t work properly here, sometimes moving backwards. This is the home of the Temporal Aumtlich.

In a sense, there’s only one Temporal Aumtlich, but you have to kill him over and over. That’s because he’s capable of spawning time-clones of himself — just like Beethro does when he steps on a temporal split token. This comes on the heels of several levels all about time-cloning puzzles, so this is one of those Shinobi moments, where you encounter an enemy with powers similar to yours, and it somehow doesn’t seem fair. Time-cloning is a very powerful tool, even if the puzzles are set up to make it difficult to apply correctly, and now that tool is being used against you. And unlike the Master Ninja’s knock-offs, the Temporal Aumtlich’s version is better than yours. Unlike Beethro, killing one branch doesn’t kill the others. Also, he doesn’t need a time split token; his power is intrinsic, and he can apply it anywhere, as many times as he likes, turning himself into an army.

In practice, though, his splits are predictable, linked to his movement, which is as controllable as that of any other monster. What he does is: Whenever he has a choice of two equally-direct paths towards you, he takes both. It reminds me a little of a non-deterministic Turing machine, an abstraction used in theoretical computer science that’s sometimes described as a computer that’s capable of splitting into two identical computers so it can explore two execution paths simultaneously. But the branches of a non-deterministic Turing machines aren’t capable of getting in each others’ way like the Temporal Aumtlich does.

So this is a tense set of puzzles, based around a problem that can keep getting worse, but it’s nonetheless something that you can learn to manage. Some of the more difficult puzzles actually require you to deliberately spawn new Temporal Aumtlich instances under your control.

Another level of those Braid-style cooperate-with-your-past-self puzzles. Let me say a little more about how this mechanism affects and is affected by DROD.

This type of puzzle is fundamentally about executing a plan without being able to see its effects. In the simplest cases, you go forth from the temporal split point and do things that will benefit you after you rewind, like standing on pressure plates to open doors for your future self to go through, taking care to anticipate how long you’ll need to stand on each plate. But your future self can do things that affect your past self, too. Suppose there’s a pressure plate that opens a door, and on the other side of that door there’s an orb that releases a dungeon roach near the pressure plate. If it kills your past self standing on the plate, you die. So your past self has to be prepared to deal with that roach, even though it can only play back a repeat of recorded actions. When you set out from the temporal split point the first time, you wave your sword in the air in the place where you know the roach will be. The actions will remain the same, but circumstances will change their meaning. This is a rich ground for puzzles.

Now, roaches are simple and predictable, and easy to plan for. But there are monsters that run on subtler and more complicated algorithms, like serpents and goblins and soldiers. My usual approach when fighting these things is to take it a turn at a time and undo a lot. I don’t have a perfect model of their behavior in my head, so I have to keep correcting my course. Temporal split tokens can make this habit unworkable. You can’t take things a turn at a time when you have to do everything in advance. Thus, puzzles using temporal split tokens have the potential to require complete mastery of the game’s increasingly complex systems in a way that other puzzles don’t.

Mercifully, the game so far hasn’t much explored this potential. I’ve seen one puzzle that pits you against future goblins, but that was in a secret room, which means it’s allowed to be unusually hard.