This is the ninth part of The Ouroboros Sequence, a series on puzzle games.

Archaica: The Path of Light

Last year I developed an interest in the qualities of beam reflection games. I’d never really had a hankering for them until I tried Archaica: The Path of Light (Two Mammoths, 2017) and it got me thinking about whether the ideas contained therein were actually unique. The levels were tight and buzzing with ideas: beam splitters, beam generators, mixing different colours of light, portal-type objects that teleport lasers…

What were the origins of the reflection puzzle? I began to dig.

I asked Archaica developer Two Mammoths about their design inspirations and they pointed me towards the Chromatron by Sean Barrett which was released in 2002. I think we’ll call Chromatron’s style Windows 3.1-esque and given the visual choice today I’d plump for Archaica despite a few UI fumbles such as the environment occasionally obscuring a puzzle.

The overlap between Archaica and Chromatron is significant but what stands out to me is Archaica’s efficient layout and Chromatron’s “agoraphobic” presentation: one large, fixed grid. Archaica scatters the pieces about the level – instead of having to drag them out from a toolbox – and the number of available spots to deploy them is tightly constrained in most levels. I am attracted to small puzzles that seem impossible because the difficulty is divorced from the “cheap” complexity of going large: it feels clever that a solution is effectively hidden in plain sight instead of obfuscated behind a wall of complexity. However Chromatron’s grid is fixed so the impressions of size are often untrue.

Chromatron 4

Chromatron has three sequels, containing 200 different levels in all. Looking back, Barrett feels too many of the levels recycle the same ideas and he would do things differently today. He also told me he used a Flash game called Reflections as his starting point for Chromatron which he did not realise at the time was an out-and-out clone of Aargon (Twilight Games, 1999) down to the actual level designs. This is an early example of a free game grabbing coverage that the commercial design source didn’t and Reflections did not even put in the level design work. This reminds me when someone remade my Atari 8-bit puzzler The Citadel for a commercial release on PC and used all the levels I had designed (although they did create a lot more of their own). Reflections did not build on Aargon and thus had no value beyond spreading the word about beam reflection mechanics.

This is the reason why Chromatron never got a graphical overhaul. Sean Barrett: “The art for Chromatron 1 was originally placeholder art meant to be replaced by something fancier and I was planning to charge money, but then I discovered I’d reinvented Aargon‘s mechanics and just killed it as a pay product and released as is.”

I asked Steve Verreault of Twilight Games where they had conjured the idea for Aargon from. How far did the lineage go back? Verreault designed Aargon with Curtis Monroe and explained they had been unaware of any earlier reflection games – which actually go back to the 1980s – and it was The Incredible Machine (Kevin Ryan, 1993) that had inspired them. “We didn’t look at other games when coming up with the rules for our game or the way it played. It was trial and error mostly. At the time we weren’t even sure if it would be possible to make something puzzling by aiming beams of light around. Ironically Aargon ended up being quite a difficult puzzle game focusing on a niche of people that liked a real challenge. The first test level was created by Curtis who created a puzzle of splitting a beam into 3 colours and back again.”

Aargon Deluxe

Okay, this is important.

The Incredible Machine asks the player to construct bizarre Rube Goldberg machines to solve specific objectives. For example: “Stop the shootout between Pokey the Cat and Bob the Fish. You must let both guns fire, but you can’t let the bullets hit Bob or Pokey.” The player has a prescribed set of tools for each task and must place each item so that once the action gets underway, the problem solves itself. Drop a bowling ball onto a scissors which cuts a wire, which releases a bucket, which bounces on a trampoline, which blocks a fired bullet…

What Twilight Games built on this was the idea of a steady-state laser puzzler. The player is not working through a sequence but effectively composing a jigsaw. If the pieces are all in the right place, it’s job done. However, with The Incredible Machine you do not know if you’ve assembled the jigsaw correctly until you start the action; in Aargon, the solution is immediately evident.

The first games featuring reflection mechanics I can find are Deflektor (Vortex Software, 1987) and Laser Chess (Mike Dupong, 1987). Most of the reflection games prior to 1999 operate in a similar way, the player twiddles with lasers to take out objects and clear the screen. I say most because I found someone else almost got to Aargon’s steady-state design innovation. Five years earlier, DOS game Laser Light (Pixel Painters, 1994) asks the player to set up a steady-state solution but the player still needs to activate the laser to see if they’ve pulled it off a la The Incredible Machine. With Aargon and Chromatron and Archaica, the lasers are always active.

Laser Light

And here’s the ironic bit.

Barrett told me he was attracted to this design because unlike Sokoban which is laborious and involves remembering a sequence of moves, it’s just a task of assembly and less frustrating. But there is a pitfall.

One of the classic gotchas of the beam reflection game is to require a solution in which a single reflector bounces two different beams at the same time. The player, at first, is convinced they do not have enough reflectors to get through the level. The flaw with the steady-state design is that once the pieces are in the correct position, the results are instantaneous: several beams which were firing into dead ends suddenly splinter and sweep the puzzle in a kaleidoscope of colour. If you accidentally solved the problem through experimentation – easily done for a player not familiar with the gotcha – the level is simply done.

This can rob the player of understanding. You have to study why the problem was solved yet some games are pretty eager to get the player moving along. Well done! CLEVER GIRL. Chromatron and Archaica both give you a chance to review and continue playing with the level; Aargon locks down the level and the player can only observe. Lazors (Pyrosphere, 2010) is a decent, tight zen game for the daily commute, but it covers up a completed level with a bravo. If you wait, you do get a chance to play with the level, but I can see players just clicking straight through onto the next challenge.


Last year’s Operator Overload (Benn Powell, 2017) took a step back from steady-state embracing the old-skool two-phase approach of Laser Light and The Incredible Machine. The player must fire the lasers then watch as the game lazily draws the beams from source to destination, so you can see how everything works. Operator Overload’s designer, Benn Powell, told me this was important to impose clarity on levels that support lots of lasers, although I was wary of that particular revelation. It was done to facilitate big complexity… and we’ve already covered that more things on the screen does not necessarily translate into more fun.

Operator Overload

One last thing before I finally stop talking about these beam reflection games. Puzzle games involving balls bouncing around are, mechanics-wise, often similar to beam reflection games. Take a look at Micron (Apparition Games, 2012) which has you lay down reflectors to bounce balls into an exit bumper. I initially assumed this put it in the same camp as Operator Overload, that it was just a clever way of seeing the path of your beams, but the first clue that all was not what it seemed was the lack of an undo.


Many levels in Micron do not support a steady-state solution. For example, players need to drive balls to hit a bumper, which opens a chamber, then have balls hit another bumper inside that chamber. This can only be done if the player adds reflectors to a puzzle in progress. Some levels go further forcing the player to siphon off a set of balls to be used separately; all of which means Micron cannot be sensibly translated into a laser reflection game.

It might look like a laser reflection game, smell like a laser reflection game, sound like a laser reflection game… but it ain’t one.

Alright, that’s enough about reflection puzzles. Next time, I want to discuss some puzzle designs which feel fresh.


  • There’s a little mathematics here that might be of assistance, although I haven’t followed through the details. David Kempe showed that laser reflection games, based on the set of mechanics in the Reflections clone, are NP-complete which indicates a certain threshold of complexity and resistance to algorithmic solution. Joseph Culberson showed in 1999 that Sokoban is PSPACE-complete, which implies it is even more complex. This suggests one reason for how they endure as puzzle templates.
  • The story goes that Costa Panayi, the developer of Deflektor, was apparently inspired by a science programme about lasers he had seen.
  • Lazors, like Archaica, does not use a toolbox from which you drag pieces out from; they are scattered across the board already. Lazors exploits this in the “Darkroom” set of challenges where there are seemingly too many pieces on the board, obstructing your lasers.
  • Alan Hazelden’s PuzzleScript game You’re Pulleying My Leg is a Sokoban/laser reflection hybrid which gets my agoraphobia going. It’s one large connected puzzle.
  • I like Micron as it does not have billions of levels and does not rely heavily on size for complexity. I did get a bit tired with all the portaling puzzles, though.
  • Updated 6 May to include quote from Sean Barrett regarding Chromatron‘s graphics.
  • Update 28 June: Reader Jon_Dog told me that The Incredible Machine featured multicoloured laser puzzles. I followed up on this and the information was almost correcct: the 1994 sequel, The Incredible Machine 2, featured them as you can see on this YouTube video. Which raised the possibility that Aargon’s inspiration was actually TIM2 and its laser puzzles… which upsets the picture presented in this post. I went back to Steve Verreault of Twilight Games to find out. He could not be sure when the team played the sequel, but Verreault was adamant that “the only real inspiration from TIM was the idea of mixing and matching lots of different pieces to solve a puzzle”. So while TIM2 might have got in there with a little laser puzzling before Aargon, the chain of design inheritance is unchanged.

Download my FREE eBook on the collapse of indie game prices an accessible and comprehensive explanation of what has happened to the market.

Sign up for the monthly Electron Dance Newsletter and follow on Twitter!

21 thoughts on “Reflections on a Design

  1. Instant feedback please. I hate waiting around for a level to do its thing. Auto-scrolling Mario levels seem like the same thing, and are the worst.

  2. Interesting! I think I prefer the instant feedback approach too, although Operator Overload’s approach has value because you can control which lasers you want to see fire – it allows to you analyse the information piece by piece.

    I think this is a step up from the two-phase “one button fires all” which means players have the tools to can parse the complexity in the same way as instant feedback. Operator Overload’s approach may even be superior as it allows you to manage laser clutter.

  3. I think it comes down to design intent. Steady-state removes all doubt about an attempt. Two-phase can be useful if the developer is trying to get some value in distancing the setup from the execution. Increased difficulty, second guessing, introducing other mechanics which might interrupt the execution, etc. If it’s not one of those, then don’t separate them.

    I like the Mario analogy the more I think about it. The auto-scroll levels are frustrating (and sparse, thankfully), but they introduce two new mechanics – a barrier ahead of you to slow you down, and a barrier behind you to force you forward. Both are killer if you’re on a platform or near a wall when they’re in your way.

  4. In a reflection game I’m designing using PuzzleScript, I take a hybrid approach; the effects of the moves on the lasers are instantaneous, but the player has to move to the exit once the level appears to be solved (in some cases, a pseudo-solution makes it impossible to reach the exit).

  5. Josh I just want to see how Archaica would have felt like with a “fire laser now” button so you can trace the path like Operator Overload. Would it feel too as I throw my weight around in the responsive Archaica?

    I think that’s an interesting take about auto scroll and waiting around. I guess I’ve never been a fan of vertical platforming either (and I can’t think where I saw it first, must *surely* be before Mario…?)


    Uh, oh, PuzzleScript alert!!!! Good luck with it.

  6. Sorry that comment went out before I’d finished.

    Josh: I’m saying would it feel slow.

    Marcos: Sounds in the vein of Alazn Hazelden’s linked game although aprobably without the gravity!

  7. That Hazelden game was very cool! It didn’t trigger my agoraphobia because of the magic words “Reset to last checkpoint,” which gave me confidence to focus on solving the problem before my nose without worrying that it would doom me later. So it felt like it was broken up into levels. The individual puzzles were still pretty complex but unlike the most agoraphobic of puzzles they eventually resolved into smaller goals–figure out the configuration of mirrors, see that this thing has to go there before that thing does, figure out what needs to get stowed out of the way and how to do it, etc.

    I did have one frightening moment near the end: V unq chfurq gur svefg zveebe gb gur raq bs gur fperra ohg V jnfa’g dhvgr fher jurgure ivpgbel jbhyq erdhver trggvat gur trz nyy gur jnl bss gur fperra, va juvpu pnfr V unq qbbzrq zlfrys naq jnf tbvat gb unir gb haqb n ybg n ybg bs fgrcf. But it worked out.

    …and yes, mega-agoraphobic games like Corrypt and Promesst give you mechanically similar checkpoints, but that’s just to mark the point when from then on you can totally screw yourself over with the slightest misstep.

  8. A moment of honesty, here, Matt. I haven’t played Alan’s game too much. First time I saw it my immediate reaction was WTF no no no. But when writing this up I gave it another go and realised it wasn’t so bad although I did get stuck on pretty soon. I had an article to write so I just abandoned it 🙂

  9. Yay! I think it’s important that you implement the solution one move at a time, so the player can experiment.

    I guess implement-one-move-at-a-time naturally goes with instantaneous-effect. The opposite paradigm is that the player has to lay out the whole solution and then press “Go” to see if it works. And you could implement the mechanic that way–it’s a valid solution only if everything is lit up and there’s a path from start to finish. But why do that in this case?

    Hmmm, the only pure lay-out-the-solution-and-try-it-out game I can think of that I’ve played offhand is Cosmic Express, though there the solution takes time to work itself out. Eets Munchies is like this sometimes but has some elements you interact with while the solution is running, though you still can’t place new elements after you’ve started the solution. I think maybe The Incredible Machine is a pure let-the-machine-work-itself-out game?

  10. Oh hey, possibly because of the new Firefox update, my save game issue is fixed! I can continue from where I left off!

  11. Marcos, I am very bad at this game! I only managed to do a couple of levels. Although to be honest, it’s very difficult to read – the result of PuzzleScript’s imposed 5×5 aesthetic I guess?

    Matt, in terms of puzzles-as-algorithm-design, you’ve also got Trainyard and anything by Zachtronics. I’m thinking about chucking out a small Ouroboros piece about “solve-at-once” vs “solve-in-sequence”. There’s also a special case of “solve-in-sequence” where you’re funnelled down a solution – such as Hexcells or RYB – which is about finding the correct next step. I’m playing with terms and I have all sorts of thoughts, especially as a results of our conversations down here in the comments.

    I don’t know if it’s the Firefox update because I think the PuzzleScript engine went through a big update recently? Don’t quote me, but that was the impression I got.

  12. Don’t worry! Several others tell me the same. It’s true that the 5×5 tile is a limitation for these types of games. Perhaps I will re-make it using something graphically more permissive.

    Speaking of Ouroboros, another of my recent experiments is named after it:

  13. *facepalm* Of course, Spacechem. And there was a sequence of puzzles in Saira that was also like that (really just a literal programming exercise, you have to program a thing to move around to a bunch of collectibles and avoid a bunch of obstacles).

    I was thinking of talking more about Saira in the thing where I was talking about ambient games. It is not an ambient game, definitely not as much as some parts of Knytt Underground; besides all the puzzles it also has a couple of worlds that are just speedruns, though there’s this nice aspect where you can wander around the course and plan your route before you start which makes it feel more atmospheric. I get the impression it is Not Beloved as much as Nifflas’s other games and I can see why, partly the speedrunning, partly the blatant Oil And Water mix of platforming and puzzles, partly the weird controls, but I did like it. Gregory Avery-Weir had an eloquent defense of the unusual jump/climb mechanics (having to repeatedly walljump up a chimney feels more like shimmying up it than holding the up arrow) and even of the genre-mixing.

  14. Sorry I have not written up something this week – I’ve actually been working on two different pieces but it’s been soooo hot here that my brain has just died in the evening. I’ll try to get both out next week to make up for it. I’m sweating just looking at my screens. This is ridiculous.

    I found Ouroboros’ Eggs interesting, Marcos, but after I finished the first level I felt done with the concept! However, I did crack some of the puzzles in the second level before walking away. There’s much more planning required here because you have to do all the eggs in one go rather than go in and out to change orientation, for example.

    I’m not sure why, Matt, I’ve not really played any of Nifflas’ later games much. It’s just bad luck really – there are hundreds of games I want to have a go at and Nifflas has just got the short straw. And I even met him in an interview with Doug Wilson. He was absolutely LOVELY. You’re always very effusive about them. I should just get on board and I feel like I’m missing out.

  15. Marcos, I hope you realize that it is entirely with affection that I ask: Why are you like this?

    (Finished year 1. It took me a while to work out the mechanics, I mean that you had to do everything within a burrow and I think that it was leaving the burrow that triggered a reset? At first I thought I had blown a time limit.)

  16. Following up and agreeing with Matt’s comments on Ouroboros’ Eggs: yes, it took a fair bit of time to work out “how” you make eggs. It was not obvious that you had to create an enclosure with the snake’s body when the instructions say just bite the tail. The reset on leaving the burrow caused some confusion too.

  17. I feel like the Overly Obsessed Girlfriend with Nifflas’s games. The later stuff just hits me in the right spot–I love platformers (especially ones that are not all about walljumping, SUPER MEAT BOY, though ironically Knytt/Saira are much more walljumpy), and these two are very atmospheric and immersive; especially Nightsky (on regular difficulty) where there’s so much effort in the sound and design put into making you feel like you’re in a space rather than in making things difficult for you. And if you like Knytt Underground it’s very easy to get lost in it–I didn’t mind the immense size because I wanted to spend more time there. (Also you can easily finish it without 100%ing it.) And as you know it is just beauuuutiful, my non-gaming mother-in-law remarked on that when she saw me playing. (That’s another thing Saira had going for it.)

    But! These are itches I have, they are not necessarily itches you have. My son asked me what my favorite game was and I instantly said “Knytt Underground” and he said “I thought you were going to say the sausage game,” which is perhaps a sign that you and I have different itches.

    Also I just have a weird relationship with games these days where I just want to fall into something familiar over and over. Part of that is that I don’t have much space on my hard drive to download new things, part of it is just that giving over an evening (and then another and another) to a narrative game seems like giving up when I should be working. So I downloaded Oxenfree and played through the first scene and haven’t gone back to it. So I go back to my old atmospheric favorites. I would like there to be more like them, though, or basically them again except not the ones I’ve played umpty times.

  18. Sooo I’ve been replaying English Country Tune and it made me think of a similar steady-state vs. dynamic puzzle thing. There’s a family of words which are about tiling the world–you flip over the grid and leave a “bush” behind every square you flip away from, and the goal is to leave bushes on every square (or every square with a certain marking). This gets intense when it moves to three dimensions, where some bushes can cover more than one square (on an interior angle).

    Anyway, I was thinking about how to classify these levels, and it struck me that they’re basically line-drawing puzzles like Cosmic Express. The goal is to find a path through the level that covers everything you need to cover. (Carl Muckenhoupt points out that the red-gate puzzles in DROD are like this too–you have a gate that opens when you’ve dropped every trap door in the room, which means walking across it exactly once.) Elsewhere in these comments (to Discussion: Ethan Carter’s Stories Untold) I mentioned a game called LYNE which is about bridges-of-Königsberg problems, and that’s the same family, though you have to draw more than one line most of the time.

    Anyway–the relevant difference here is that English Country Tune and DROD are dynamic while Cosmic Express and LYNE are steady-state, which may just be to say that the first group is keyboard-controlled and the second is mouse-controlled. In Cosmic Express and LYNE you just draw the lines, and you see the lines drawn as paths. In ECT and DROD you have to step your avatar around the level bit by bit until the path is formed, and the path remains behind you in a physical form (planted bushes, dropped trapdoors) like the traces in Donnie Darko. In the first group, the paths are drawn in space and you can retract them by sweeping the mouse backward, and in the second, the paths are drawn in time and you can retract them by holding the undo key.

    Notable also, in ECT and DROD the engine incorporates other kinds of puzzle, which requires the interface to be dynamic. Also in ECT the playing field keeps rotating in three dimensions to show where you are, which would be hard to do in a strictly line-drawing interface. And the temporal aspect of the paths makes it harder to edit your solution, since you can’t actually see the path all at once–if two adjacent squares are planted you don’t know if you got from one to the other.

    Also this helps explain why I don’t like these puzzles as much as the other levels. I just want to use my square to move things around, man.

  19. Hey Matt.

    So I had another read of Repetitive Strain and reminded myself I categorised this kind of puzzle as being degenerative path-dependent. The solution depends on the sequence of action but your possibility space degrades with every step.

    I think I can just recall this was one of those points during Repetitive Strain where I had to hesitate. If you step back far enough, everything looks the same. Technically, Cosmic Express has some equivalence to the “degenerative path-dependent” levels of ECT but the process of solution is quite different as you point out.

    Cosmic Express allows you to edit the total solution from an omniscient viewpoint and it’s just not as fun if you are forced to step it out. Not least because finding the solution will entail extensive backtracking. I take onboard that these puzzles cannot be easily translated to a path-independent version. Thanks for the data point. Always useful!

Comments are closed.