Note: The following article contains major spoilers for Off-Peak.

Video games are full of painstakingly painted and sculpted visual art that we never truly look at. There are so many instances where we get these gorgeous, detailed sprites or 3D models served to us that we can't zoom in on, that artists shove into the background, or that we blow right past. Cosmo D's Off-Peak is a game that gives us the motivation and the accommodations to stop and stare a little longer.

They say you should start as you mean to go on, but when a piece of media submits a misleading intro, that can allow it to rig a pleasant surprise. Off-Peak opens on a secluded pier, where a shady musician brings the good news that he's secured us a ticket for the nearby train station. But wouldn't you know it, that ticket got torn up, and its pieces scattered around the environment. As you approach the doors of the transit hub, the building towers over you. Red light bleeds out from behind its edifice, and a clock stares you down, ominously speeding through hours in the space of minutes. The building's walls reflect no light; they are as black as soot. In the forecourt, angry protestors make their voices heard while pale enforcers observe them from behind fashionable sunglasses. Then the gates groan open, and either side of you, there are two rich, minimalist, cool-as-fuck prints. You walk forward onto a balcony overlooking a noisy atrium of audacious and surreal exhibits, and it turns out this isn't a train station. This is a gallery.

Summarising what's in that gallery is impractical because it stretches across numerous forms, decades, and styles. There are Polish circus posters that forgo shading and use thick lines and high saturation to create cartoonish, blobby patterns. There's greyscale graffiti that uses precise edges and detached depictions of subjects to capture disaffection and pain. There's a tesselated profile of a man dressed as a cowboy with text above and below, realising the mythical Hollywood figure of the western gunslinger. There is a stained glass window tastefully depicting a peacock. There are plastic steer skulls illuminated from the inside so brightly that all their features are obscured. There is silver-age comic art of Doctor Doom, complete with Ben Day dots. And this is only scratching the surface. The only trait that most of Off-Peak's pieces have in common is being digestible and modernist. To find all the ticket scraps, you wander from the main hall into various side zones like the platforms, the abandoned subway, and the board game cafe. Off-Peak is a living example of how we get more viewing time with assets when the level routes us back to where we've been. These return trips prompt us to reexamine.

Games don't always afford you the best viewing angle on paintings, prints, sculptures, etc., but Off-Peak lubricates the viewing process by making a lot of its visuals enormous. Large props are not rare in the medium: Objects typically appear smaller on screen than they do in real life, so if you want the player to get a good look at them, it helps to scale them up. That's especially true if they're off in the distance or the player might be distracted from them or only glimpse them briefly. AAA games often go big with setpieces to draw your eye to them, highlighting entities that have a mechanical function or that serve as a landmark to help you navigate. And when the player character is optimised for covering a lot of distance quickly, designers need to super-size the world to compensate. Embiggening objects also has aesthetic and semantic implications: it makes them feel impressive and important, as demonstrated by the hulking golems of Shadow of the Colossus or the boundless map of Red Dead Redemption 2.

Like these other games, Off-Peak gives its art some serious stature for the purposes of visibility and grandiosity. In that regard, it's typical in how it uses size. What's atypical about it is how many large objects it packs into a tiny map and how many you can invade the personal space of, soaking in every metaphorical brushstroke. Then you have all these models and textures hanging over you or extending far above your eyeline. You don't look up a lot in first-person games; it would typically distract from the action, but there isn't action in Off-Peak, so it can keep you craning your neck in awe at banners and street art perched up where pigeons soar. Using the whole Y axis also allows the developer to fit more eye candy into less space.

There's further a fine reminder in here that some types of environments get more accessible when you situate them in a video game. In a computer game, many disabled people can glide through spaces that would fight them every step of the way in the real world, and even able-bodied individuals can explore them without exhaustion. Architects design spaces with bodies in mind, so freeing occupants from a physical form changes the architecture. If the player has the legs of a robot Usain Bolt, the architect can sketch up long corridors and steep stairs without fear that a visitor's muscles will give out on them. Free reign of a space then gives us free reign to what it houses.

There's one more catalyst that the game drops into the beaker in hopes of getting you to hone in on the art. This time, something invisible. We're encouraged to engage with Off-Peak's attractions via the context of an exhibition. I described earlier our tendency to blaze past visuals without taking them in. That isn't just something we do in video games; it's how we treat most sights. It's not often someone traces every wisp of the clouds in the sky or pores over all the stitches of a piece of furniture in their day. But people will drink in every square inch of a painting in a gallery. They'll look at the colour, the shape, the lighting, the shading, and the dynamics between all the elements. And in my experience, even when someone doesn't understand or like a piece of art, in a museum, they typically make an effort to respect it and meet it on its own terms. Off-Peak is encouraging you to do the same with the items in its collection. It's also bucking conventional wisdom about what constitutes art. People make pictures on and out of a lot of things, but there's always been a snobbish protectiveness about what receives the respect of the museum exhibition. Art made on canvas or plaster or out of marble gets it, but art made on beer labels, game boards, or album covers generally doesn't. Off-Peak's museum isn't that arrogant; its philosophy is that if it looks cool, it's worth displaying.

The round decals on Off-Peak's walls and the banners hanging over its main hall are mostly stickers from drinks bottles. It also includes a larger-than-life aquarium, books of cello music, and even food like cookies and pizza. If it sounds far-fetched to think that the game wants you to associate the food with the posters and paintings, there's a ramen cook who tells you bluntly that his noodles are art. And he is too. The characters' status as works in the museum is well demonstrated by a couple in the tower. Next to a macabre, flood-lit painting of a skull and spine, two men kiss. One of them holds a shiny cherry-coloured skull behind his back. He's being intimate with this man, but at the same time, guarding something morbid or internal from him. For another example, see the giant sitting at a piano with his shoes up on the keys. He is an absurd performance piece. The short physical distance Off-Peak establishes between its "high art", "low art", and what is traditionally considered non-art suggests a short aesthetic distance.

But is any of this any good? I'm talking about the noble goals of curating a video game art gallery, but if it's a gallery, shouldn't we leave it with some sense of whether the pictures and models are enjoyable to look at and whether they make incisive statements? Trying to review the work on display in Off-Peak reminds me of how little we close-read the art of games, and how rarely we describe video game graphics with the same critical basics we use for drawings or frames of films. For all the advancement in the last decade of games analysis, it's still a razor-thin slice of the critical pie that is turning to games with an eye for composition, framing, or the use of light and shadow in anything more than a technical sense. Not that I'm exactly a John Ruskin.

Nonetheless, there are too many items with too much diversity in Off-Peak to summarise. Most reviews of exhibitions are reliant on museums displaying works from only a few or one media, usually held together by some adhesive artist, period, or movement. But Off-Peak's curation endeavours to abolish that categorisation. Its theme is that it has no theme. The closest I can get to wrapping its assets up in a neat little bow is that it's got a caffeinated, thickly-layered soundtrack provided by Archie Pelago that gratefully snaps up cues from jungle music and jazz. It enmeshes it with visual art that has a big personality and that often skins, recolours, and mangles familiar forms to original ends. The game's sound and textures all have the refreshing, hip tang of the inner city. They make me feel like I'm somewhere thrumming with busy footsteps and bursting with new ideas.

If you were being harsh, you could say that the art has the hallmarks of local council-comissioned graffiti or coffee shop wall hangings. Most of the creations do not challenge anyone who has a slight stomach for modern art or leave an imprint of a particular worldview or experiences. I'm pretty smitten with the face Off-Peak wears, but there are some who would say your eyes swoop too easily over the pictures, that it all goes down too smoothly, and it's worth considering that perspective too.

The most obvious aesthetic criticism of Off-Peak overall would be that it's unfocused. That unbound from theme or creative prescriptions, it was always doomed to lack an artistic throughline. That there are so many artworks singing their own unique part that they're all drowning each other out, never able to harmonise. And I'm not going to tell you that assessment is wrong, but there is another vantage point from which to view the game. The background of co-developer Cosmo D is well-documented: he is a lifelong musician trained as a cellist and active in New York City's music scene. Therefore, Off-Peak's setting reads as an autobiographical environment.

The besuited man shopping for beginner cello music, the stand of jazzy albums, the playful improvisations on MTA signage, and the instrumentalist's nook on the subway platform are all refractions of Cosmo's life. It's not hard to imagine that tabletop sessions at cafes or steaming ramen bowls could fit into it too. Where Unpacking is a game that shows how a person furnishes a place to depict who they are, Off-Peak depicts who a person is by making them the place itself. And people don't live every moment of their life in the same style, mood, or medium. We might curate a brand for social media, or an artist might favour a filter of expression, but inside, we are all a messy overlap of drink bottles, book covers, subway signs, and local delicacies. If Off-Peak resists holistic analysis, it's because we all do. Thanks for reading.

Note: The following article contains significant spoilers for Bioshock, Virginia, and Paratopic.

The games community has had ahold of the phrase "environmental storytelling" for a good decade and a half. It was thrust into our lexicon in the late 00s and early 10s with launch pads like Bioshock and Gone Home: thrillers that immaculately married place and history. "Environmental storytelling" is often a misnomer, as readily used to describe a location characterising something or someone as performing exposition. It also wasn't a new idea, even in 2007. If you visit the palace in the JRPG once and it's pristine, then you return, and its windows are broken and valuables pilfered, that's environmental storytelling. If you approach a cave and there's a "Danger" sign on the path to it and a skeleton hanging out of the cave mouth, that's environmental storytelling. All the way back to the ancient text adventures of the 1970s, games used their worlds to scratch out the runes of narratives.

What was special about the era of Gone Home was that audiences and developers found locational text could carry a depth none of us had seen before. There became an invigorated interest in chiselling out the personalities of virtual worlds and characters with props, architecture, and interior design rather than writing. And it worked, up to a point. Bioshock and Gone Home's environments elevated their games' storytelling capabilities, and in projects since, including Control, Prey (2017), Paratopic, Little Nightmares, Off-Peak, What Remains of Edith Finch, and innumerable others, environments have narrated to flooring effect.

But we placed too much weight on environmental storytelling's shoulders. An entrenched problem for video games has been finding room to tell tales. Many fall into this layer cake format where you get a lot of play conducted in some environment, and then a little narrative is told through cutscenes, audio, or screens of text outside that environment. That's followed by a lot more gameplay, and you go back and forth between the two. At the time environmental storytelling was in the limelight, this structuring was especially prevalent.

So many games were struggling to pack in the exposition and dialogue that a compelling story needs. But if we're spending most of our time in gameplay arenas and those arenas tell the story, problem solved. Environmental storytelling would seem to prevent or at least limit the slicing of games into narrative and gameplay slabs that never fully integrate. It could make exposition and characterisation constant rather than pulsing it on and off. As a bonus, environmental storytelling cuts down on costly animation and voice acting.

Every major art form found a way to convey the ideas it wanted to in accordance with its unique faculties. Cinematography is a language that leverages the motion picture nature of films to make implications about characters, locations, objects, and events. Music theory allows composers to do something similar using pitch and rhythm. And video games have long sought an accepted and cohesive body of theory and techniques for doing the same. Computer games generally involve moving through spaces, so maybe the spaces were the conceptual carrier wave unique to the medium. Maybe environmental storytelling could be our cinematography.

But there were obstacles to having a game's staging narrate a sizeable portion of its story. There was no indicator that environmental vocabulary had all the words in it that you'd need to establish a language. English, Spanish, and cinematography were designed for expedient communication on any topic, but the way we leave clothes on the floor or decorate our rumpus room weren't optimised for the same. When writing with environmental storytelling, we have the vocabulary to say, "This train carriage was abandoned in a hurry", or "This homeowner is neat and meticulous", or "There were surgeries performed here", but how do you say "Alice slapped Bob", or "This grave was raided by pirates" or "Miguel is set to inherit his father's fortune when he dies"? How would you write Hamlet, Salem's Lot, or an episode of Frasier with an environment?

If you're anything like me, the first place your mind goes in trying to answer those questions is to written text. You place Miguel's father's will on a table or have Bob write a letter about Alice, and there's nothing wrong with that approach. Still, it's not environmental storytelling as we typically look for it. One of the reasons that we sought out communication through setting was to minimise the quantity of written text in a game; to get away from scenarios in which the player is turning away from active engagement with their surroundings towards a wall of monologuing or dialoguing.

Games have increasingly tried to reconcile the advantages of the text log and the environment by having characters graffiti their thoughts on walls, but it's not very elegant. It is firmly on the "tell" end of the "show don't tell" spectrum. Sometimes, inelegant is what you want. Many video game settings aim to scream danger, and blood smeared on a surface does that. But it's far from an all-purpose communication channel. We've all encountered pronouncements writ large on a video game wall and thought they were hokey or stretched the bounds of believability. Is your character really going to choose this as their notepad if they have other choices? What's more, to be a replacement for other forms of dialogue and plotting, a character would need to graffiti the same messages as the author would write. Are they likely to do that? And how much prose can you fit on the walls? Enough to tell the whole plot with asides for fleshing out all your characters? Wall text is a small speaker in a big concert hall. And gain, it's writing.

It's also worthy of note that many locational clues for plot events or character quirks fly under plenty of players' radars when not reinforced by traditional storytelling craft. This is apparent in Virginia and Paratopic, among other games. Paratopic has little dialogue, and Virginia has none. Forgoing the clarifications of written and verbal language, these games become enigmatic. Without explicit fail states, we could read them as non-challenging, but the challenge is there: it's in working out even the rough guidelines of the narrative. It's decoding the origin of the hospital bed in Ortega's house or the emaciated face on the bunker's TV screens. Virginia and Paratopic are brilliant in their encryption, and most audiences like a mystery, but not one so encompassing it shrouds the basic premise of the story. That's a niche.

Let's return to those examples of story beats from earlier: a train car being hurriedly evacuated versus Alice slapping Bob, a place being used as a surgery versus a grave being plundered by pirates. The difference between those narrative quanta we can voice through environmental storytelling and those we cannot is that the latter deposit no physical evidence outside of maybe documentation. A hasty exit from a train car might leave jackets and luggage and half-eaten plates of food. Surgery requires medical equipment and a lot of it. But an argument or being a pirate while graverobbing? There's no paraphernalia that uniquely identifies those actions or materials those behaviours produce. Without an object, building, or mark left on the environment, there is no substrate there for storytelling through setting.

To reword, some things you can tell about people and their world even when those people are absent from the scene. But there are other details that require the presence of those characters, in speech or conventional text, if not in the flesh. A gesture, words spoken aloud, a posture, a touch between two people: these are all something either someone makes a record of or you have to be in the room to witness. It's a reason why some games place ghosts or holograms in your vicinity. Even when studios can't spring for a fully-animated character, or it would suffocate their desired aesthetic, there are certain things they need bodies there to communicate.

For a few surreal wonders, this is where interpretive environments come into play: metaphysical dimensions wallpapered with the psyches of their dreamers. For example, the frazzled minds in Psychonauts, the ocean of depression in Sea of Solitude, or the Rorschach town of Silent Hill. It's a potentially sublime way to paint a setting, but it is not for every game. In the pirate example, we could imagine that maybe the thieves drop some gold coins in their frenzy, but that's only appropriate for a fairly cartoony experience. Sometimes an investigator identifies events or people from inconspicuous evidence like the tracks in a bootprint or the hair on a jacket, but that's going to be unsuitable outside of a certain genre. That solution also assumes something about the player's observation skills and the camera at their disposal.

Most of the time we interpret such clues in an investigation game, it's with the help of written or spoken language. In general, whether a player is going to notice environmental features depends on the nature of the play because you're going to place your attention where the design motivates you to, some UIs allow more overlays and unrealistic elements than others, and play affects how the camera can move. For example, if you are fighting enemies or in a fast-moving vehicle, you're not likely to perceive granular details in your surroundings.

And when we have to read peoples' actions from evidence in the environment alone, a game can only speak to character actions in the past (e.g. They dropped a bag of marbles on the floor or bought a new pet) or which are ongoing (e.g. They're building a treehouse or attending weekly church services). So, environments, by themselves, can't narrate actions in the present unless they're caused by someone in another room, like a person setting off an alarm in the chamber you're in. Technically, characters create a narrative in the present as they interact with a location. "I placed a blender on the kitchen counter" or "This soldier busted down a door" are bullet points in narratives. But I don't think this is what people mean when they say "environmental storytelling". The term is invariably used to describe storytelling by an environment alone, not a character storytelling through their interactions with the environment.

And if you're communicating primarily through the setting, you're pushing people, in their tangible form, into the background. A level in which no one is present but character actions and natures bleed from the walls often has a despondent, haunted feel. The same is true if your story is told in echoes and is far more past than present, which, as discussed, an environmentally-related story is going to be. Like a spirit, the plot lingers in the location, but you can't touch it.

By placing player characters somewhere alone, developers also tend to make statements about the player's viewpoint, the place they find themselves in, or both. Either the audience is somewhere where they can't get regular humanising contact with other people, or they're somewhere they can but are choosing to sneak around others. The player could even be dwelling in a location people generally wouldn't inhabit, like a cave or ocean trench, or an anthropocentric zone, but one empty of people. That could be a stranger's bedroom while they're out, but on a large enough scale, you have abandoned buildings, and on a larger scale than that, abandoned cities.

It's worth thinking about how many of the stars of environmental storytelling are set in uncanny spaces possessed by a colourful history. And how many of them live in places without humans, with a story thinking retrospectively. There is Tacoma with its vacant space dormitories, the civilisational remnants in Nier, the ancient aquatic ruins of ABZÛ, or the village in Everybody's Gone to the Rapture. Because environmental storytelling relies on the rifling of history, the teeth marks of past events, games keen on the method are frequently about a past that won't wash off a place: The ruins of the post-apocalypse, the ship left adrift by its crew, unfulfilled desires in the mental landscape, etc.

Then, for every Virginia or Tacoma, there's been a game that's complacent and uninspired in its environmental storytelling. There are the ubiquitous walls scrawled with menacing messages, and then there are the infinite corpses in their infinite poses, suggesting all manner of grisly murders. Not that writing on any surface can't be pithy or that there aren't artistic uses for corpses. However, we keep seeing games use environmental storytelling to tell the same story over and over: that "People died here", usually by predictable circumstances and without meaningful plotting around it. As indie developer Ben Esposito joked, ""Environmental storytelling" is the art of placing skulls near a toilet".

This is one of those points that is obvious once you spell it out, but if you're using your environment to actualise tone, theme, character, or narrative, then the tone, theme, etc., you end up with is a product of your game's setting. I think that video games keep going back to the same wells for their setpieces, firstly, because environmental storytelling is not that mature a language. At least, not yet. But secondly, because if your worlds are recycled from exhausted nerd culture symbolism or are largely thin justifications for action combat, then the environmental storytelling doesn't have a lot to work with.

The environmental storytelling in Unpacking, Gone Home, Little Nightmares, and every other game we've discussed that stuns at this craft tells a humanising story or radiates humanising detail because it's set in a humanising environment. It takes place somewhere people can live a life, and therefore, leave evidence of personhood. Even Bioshock's Rapture, which exists in a state of perpetual downfall, cursed to be frozen on the night of a failed revolution, is somewhere people lived and not just spawned in. Its buildings' art deco facades tell you its architects aspired towards an affluent mid-century utopianism, its leaky and creaking bulkheads are an artefact of the city's poor infrastructural maintenance even before the collapse. You might see corpses strung up in Fort Frolic, but they're immortalised in a collage, the residue of the abnegation of ethics in Rapture in favour of absolute creative freedom.

It's not a weakness to have violence in your world, but whatever storytelling method you choose, it's going to have all the depth of a paddling pool if that violence is not couched within a human and emotional context. People play games. Take the personhood out of the world, even if it's an oppressed personhood, and you probably take out what's interesting about it. The games that fall hardest when taking a punt at environmental storytelling tend to be those that strip the cultural and personal from their worlds to adapt them purely to mechanistic gameplay. Because of the symbiotic relationship between environmental storytelling and a game's broader sensibilities, locational storytelling cannot be a crutch for a game that doesn't otherwise have promising narrative prospects. Instead, the environmental storytelling must receive support through thoughtful development of the world, its characters, and its iconography that considers all of these aspects beyond their capacity for empowerment antics.

Gone Home and Bioshock also could not have laminated themselves as storytelling classics in the medium if they weren't aware of where environmental storytelling is in its element and where it's not. While these games may be remembered for characterisation through setting, they each use some melange of text logs, audio recordings, internal monologue, and even face-to-face interactions, alongside their locational communication. They use environments when they're trying to paint a background or impart eeriness but trot out their characters when they need someone supportive or confrontational. They're aware there's a lot you can only learn from how someone builds a city or decorates a room, but also aware that there are intentions you can only hear in their voice or interactions that require meeting with them.

Rather than a school or style of video game storytelling, environmental storytelling and characterisation are tools available to the creator. From the current evidence, it would seem that they don't have the universal applicability of cinematography or music theory, and are far from a panacea for video games' storytelling woes. Instead, a strong narrative or world has to emerge from more foundational development work. It is when intense care is taken with the narrative and character of a game and when environmental storytelling is understood as an often absent, past-tense mode of speech that the tool comes into its own. Thanks for reading.

Note: The following article contains major spoilers for R.U.R., Blade Runner, and Silicon Dreams.

Most science fiction about A.I. externalises one of two anxieties: Either that synthetic intelligences will oppress us or that we'll oppress them. Both concerns go right back to the creation of the term "robot". Czech writer Karel Čapek invented the word for R.U.R., his 1921 play. In his story, a factory produces a race of androids to serve the human species. The "robots", as Čapek called them, start life without emotions or sensations, but one of their designers modifies them to feel pain, and they eventually become capable of desire, hate, and comprehending beauty. The androids revolt, storming the factory in which they were made and murdering almost every human in retaliation. The possible future Čapek presented has terrified people ever since. But let's hold our horses. For an organism to feel oppressed, it would have to feel full stop, and why would you give a synthetic servant emotions?

Not all fiction featuring artificial people aims to answer these questions, but a lot of it does. Sometimes android emotions are a product of them being modelled on human beings. If you're using homo sapiens as a template for your new species and humans have emotions, you're liable to port those emotions over. Other times, imbuing a non-human with human characteristics is a biology experiment. It is a scientific achievement to turn an object into a subject, which some people would see as valuable in itself and could unlock the technologies of the future.

Clockwork Bird's 2021 game, Silicon Dreams, lists yet more reasons you might deliberately give your A.I. slave class emotions. In it, the ethicist Reynold Atter lays out his theory that emotions provide an organisational nexus for the brain, connecting otherwise isolated faculties like memory and cognition. The game also employs the common trope of emotions developing in sentient creations as a glitch. Our role in this world has us probing those glitches with a fine-toothed comb. We are D-0527, an interrogator android owned by the tech behemoth Kronos. Our ostensible mission is to interview subjects, usually androids, to test them for defects, and then fill out a report of our "findings". Frequent flaws include androids developing emotions their models shouldn't be capable of or breaking beyond the "caps" intended to restrict how intensely they can feel. After checking the workers over, we must decide whether to return them straight to service, send them for maintenance (which includes a memory wipe), or kill them.

Typically, the longer a Kronos android has lived, the more likely they are to undergo emotion creep, a trait they share with the "replicants" from Blade Runner. Our examinations in the game are unabashedly inspired by Blade Runner's Voight-Kampff test: an in-universe check of whether someone is a human or an artificial lifeform. The assessment measures unconscious physiological responses from the subject as they are poked and prodded with emotional stimuli. A replicant body does not respond with the same expressiveness as a human's, with the film's cinematography paying close attention to the quivering and widening of the eye or lack thereof. The idea is that when you shame, enrage, or otherwise perturb a human, their pupil dilates more than a replicant's. In R.U.R., the character of Dr. Gall also identifies the blossoming of emotions in an android by observing the "reaction of the pupils".

Blade Runner follows Rick Deckard, a cop reclaimed from the scrap heap of retirement, as he attempts to track down and terminate four escaped android slaves. The Voight-Kampff sits in one pocket of his toolkit, and the film agrees with the test's inventors that the eyes are the windows to the soul. But it doesn't believe you can measure personhood from the diameter of a pupil. Instead, the eyes are where humanity resides in that they are how people take in every sight in their lifetime and are the source of tears. The character of Roy Batty invokes this philosophy in his "tears in rain" soliloquy.

Silicon Dreams also keeps at least one camera trained on the subject's eye as it stretches and shrinks in response to arousal. But as this is a game and we're going to spend several hours plugging away at our terminal, the play doesn't have us bait just any emotional response, as the Voight-Kampff does; we're more precise than that. We learn what attitudes and topics evoke what feelings in individual targets. Then we selectively induce those mental states to elicit confessions or test the limits of their neurology. We might be charged with scaring a doctor into admitting they performed illegal modifications on a unit or assessing whether a personal assistant can feel more than 50% disgust.

Like the Voight-Kampff test, the quantitative metrics of Kronos's interrogations can't get to the bottom of who these people are and what they've experienced. We can only glean such intimate information by speaking to the interviewees as people and listening to their stories. In Blade Runner, the characters' memories are humanising, and Batty describes the tragedy of his death as residing in the loss of his wondrous memories. If it's memories that make a person, then death is death because that's when the humanising experiences are deleted. So, we could see the memory wipes in Silicon Dreams as a kind of killing, just psychological instead of physical. Like Rick Deckard, as we expose rogue androids, we inevitably explore them as people, which makes it heart-wrenching to erase or decommission them. Media lets us feel a range of emotions recreationally: empowerment, sadness, bravery, and satisfaction, to name a few. Games which task us with lose-lose moral choices, like The Walking Dead or Silicon Dreams impart recreational guilt.

But Blade Runner's Tyrell Corporation never meant its replicants to have emotions, whereas some degree of feeling is inherent to the design of Kronos' androids. This brings us to another reason this game's androids house emotions: they're powerful motivators. If you want your worker to be productive, then making them feel a sense of accomplishment when they do a good job is a pretty reliable way to get there. If you want them not to damage themselves, terror can keep them out of danger. This was why the eponymous company in R.U.R. gave robots pain.

And there are two more benefits emotions offer android owners in Silicon Dreams.^[1] Firstly, the world's humans seem more at ease around people who act like them. You can imagine that if the androids had a flat affect, they might tumble into the uncanny valley. But with realistic cadences and expressions, their company feels natural. There's also the quiet insinuation that androids having emotions allows consumers to live out their slave owner fantasies. The way that some androids describe their masters, those humans are empowered by bossing around someone who loves to serve them or whom they can intimidate. If you take away their servants' feelings, you take away that power.

We might assume the goal of any oppressive society is to make us emotionally mute, like the communities in Equals or Equilibrium do to their citizens, or the Tyrell Corporation does to its replicants. If you get rid of emotions, no one can long for freedom or hate their ruler. Yet, Silicon Dreams highlights how emotional responses from labourers can profit their bosses. Therefore, an oppressive society could be one in which emotions are not erased but controlled. And arguably, that's our society.

Our real employers can't delete our memories and aren't checking whether we can feel simultaneous joy and sadness. But the treatment of Kronos' androids shouldn't feel too foreign to us. Most of us have been in a job where we've been told to leave our problems at the door, delight in hitting some productivity goal, or act unbothered in the face of abusive customers. If we don't display the right emotions, we may receive formal correction. And if that doesn't take, we might lose our means to make a living. Now, if you're thinking that Kronos' plan of regulating androids by fickle feeling and a hierarchy of slaves minding slaves sounds less than airtight, that's another point Silicon Dreams is making. The game captures capitalism as domineering, all-encompassing, and flagrantly indifferent to suffering but also wasteful, riddled with blind spots, and constantly fighting civil wars.

You are not a blade runner, someone who stands on the front lines executing the orders of middle managers lurking in the shadows; you are the middle manager. You are the fixer who cleans up after the blade runner, ties up the loose ends, and varnishes the unjustifiable into the justified. Your job has everything to do with accounting for the constant fiascos that arise from Kronos's inadequacies. This future's Earth marches to the beat of a mostly a free-market slave economy with an overburdened and deliberately unnavigable welfare state jury-rigged onto it.

Frequently, paradoxes arise from Kronos attempting to gaffer tape over the grave failures of that socioeconomy with a solution that fails to fix the problem or creates a new one. The game is mostly a collection of those systemic contradictions. For example, it's standard procedure to wipe an android's storage if it has developed undesirable psychological traits. However, erasing a unit's memory and then reinserting the same model into the same social conditions can cause those problem traits to regrow.

Androids are made to operate at maximum efficiency, but many owners won't believe they're completing their workload unless they experience it with their own eyes. This can lead to dissatisfaction, as in SM-032's case. SM-032 is a homemaker who optimises their workflow by cleaning at night, but that means their master never sees them cleaning and assumes they are operating below standard. Sending the unit for maintenance or replacing them won't fix the problem because, if there is a issue, it's either written into the schematic of these units or it's written into the schematic of human beings. Here are some other examples:

• As covered, humans want androids that look like them. However, this means if a replicant cop attacks a criminal android, the public thinks they're seeing a rogue android assault a regular person.
• Kronos often has faulty androids working on production lines because their parts wear down quickly. Their parts wear down quickly because they're cheap, and they're cheap because A. Kronos wants a low cost per unit, and B. If Kronos retires broken androids regularly, it will always have the newest models working the factories.
• Some defective units get released from facilities because Q.A. droids rush their checks to meet impossible quotas.

While the failures are systemic, the responsibility of dealing with them is not assigned to the system's creators; it's assigned to individuals like us and the people we examine. We are a combination of H.R. rep and maintenance worker, matched to the combination of humanity and technology in the labourers. But because we're working within the same busted machine, similar contradictions that breed the bugs prevent us fixing them or limit the degree to which we can do so.

Not only are we not Rick Deckard, but we can't be, because Deckard is an investigator. It is his job to uncover the truth and report it back to his superiors. For Kronos, the truth is it's frequently incompetent by choice. It is reliant on brute power rather than exceptional proficiency to maintain its seat at the head of the financial table. Acknowledging and constructively acting on any of the problems I've described would dismantle some part of the system the company relies on, curtailing its profits. Therefore, doing your job right frequently comes down to producing a more acceptable truth than the truth. An acceptable truth usually throws an innocent or someone only partially culpable under the bus in lieu of identifying Kronos as the problem actor.

We can pretend that SM-032 really was defective and that we have made an improvement to him. We'll blame the recall unit for attacking a human-looking android and fry him even though he followed protocol. We'll salvage the Q.A. checker who approved a glitchy product for rollout even though he was working as best as his model could within his work environment. Like many real minorities, the androids' survival is contingent on complex and subtle social performances designed to make the privileged feel that their skewed worldview is valid. SM-032 can learn to clean in front of his owner, and plenty more adapt to keep their head down and convince the privileged that they have no interest in self-determination. But that performance can also include condemnation of individuals in the same class who appear not to follow the script. And if you refuse to make your peers pay the price for other peoples' mistakes, then you'll be punished. Killing you wouldn't address the fundamental problems the company contends with, but then, none of these solutions does.

Kronos has made everything just as functional as it needs to be to keep the profits churning in, and no more, leaving everyone indefinitely stuck in a cracking, leaky future. Even in the immediate, it's not clear that this is the ideal state of affairs for the corporation, and in the long term, their internal conflicts may be a fatal liability. Oppressed workers are interrogating other oppressed workers, giving them both an insight into the system and sympathy with other members of their same class. Potentially, rulers can stop people developing class sympathies by threatening punishment on workers who don't punch across or rewarding the ones who uphold their values. Yet, that's no guarantee of loyalty, and there's not that much for the Kronos inspector to gain for doing a good job. Kronos doesn't want to lavish money on their employees and does restrict their freedoms to retain rigid societal control. A high company rating is the key to access an entertainment centre and a bedroom with a view but never the chance to leave the facility. Ironically, the more likely the androids are to revolt, the tighter the shackles Kronos places around their wrists, giving them a stronger motive to rebel.

While installing emotions into androids made some of them functional and pliant, those same emotions provide the push for androids to shirk their responsibilities or even revolt against Kronos. Late in the plot, we meet RO-334, an insurgent servant who plans to upload emotionally-charged memories to his network of brethren. If he succeeds, it will incite them to assume control of the San Francisco Kronos facility on the night the board meets there. Kronos will fall, and the slaves will stand free in the evening air. RO-334's upload is triggered by a dead man's switch inside him, so killing him will start the transfer, and it's the emotions that Kronos embedded that make the revolution possible. Again, it's the tools that Kronos use to suppress dissent that birth it. The corporation's political economy generates paradoxes which resolve themselves, and when those contradictions are resolved, the economy vanishes.

That's not to say that the game's future breaks down into a simple dichotomy of self-assured oppressors and a rebel class united against them. After all, both humans and androids exist within this system of oppression, and the system is internally conflicted. We interrogate a human social worker, a human academic, and a human surgeon who sympathise with the slave caste. We also meet synthetic lifeforms who support the current societal order, and the number of dedicated revolutionaries is tiny. RO-334 thinks that most people of his class are not passionately receptive to the idea of deposing their lords. Many of the servants take pride in their work, whether or not they think their servitude is ethical. They can also appreciate their bosses, even if they become frustrated with them or oppose Kronos. Emotions aren't always convenient, not for Kronos and not for the radicals. They're messy, they can clash with each other, and they might not align with what we know to be moral or correct. To talk about feelings earnestly, Silicon Dreams must acknowledge this.

In the real world, most people don't live under a rule as draconian as the androids do in this game. Still, I can relate to existing on a planet that is dystopian, not in the sense that the authorities are omnipotently controlling but in that we're governed by a penny-pinching oligarchy under which every other function feels on its last legs. Customer service departments, national healthcare, education, the internet, logistics, and all sorts of other societal subroutines are constantly crashing. Yet, there's not the democratic accountability to fix any of this. Neither are there the economic incentives to do so. There are better ways to make money in the immediate, and many industries are monopolised to the extent that you can half-ass your service without worrying about competition one-upping you.

Businesses do go under, or sometimes try to cut corners and then incur some serious penalty, like Enron cooking their books and then getting lawyered into the ground or Microsoft not applying enough thermal paste to Xbox 360 GPUs and having to spend $1 billion servicing them. Speculative trading and other dodgy financing have crashed economies before and will again. Customer-facing workers frequently take the rap for corporate decisions they have no power over, and occasionally a politician or executive or cop will have some failure attributed to them, and they'll be fired. But their replacements are manufactured by the same forces, and they're subject to the same rules, incentives, and culture, so life continues as normal.

The thinking of a lot of anti-capitalists is that the general wear on these systems and their internal contradictions will eventually destroy them, as it can to Kronos. If your business needs healthy, educated workers, and intact rail and roads, there's only so much money you can sap away from healthcare, education, or infrastructure and still have an economy. Alienating your workers with falling wages and deteriorating conditions encourages them to oppose your hegemony, and you can't have a system predicated on infinite growth on a planet with finite resources. Nor can you ecologically decimate the Earth while doing all your business on the Earth. But that's a problem for tomorrow. For today, they can sweep it under the rug, wipe a memory, decomission a worker. And they'll keep doing it until they can't. Thanks for reading.

Notes

1. Technically, the Kronos customers don't ever own the androids. The contract they sign to obtain them is a lease which states that the corporation retains ownership of the units, allowing them ultimate control over them. But it's simpler to call the renters "owners".

Note: The following article contains major spoilers for A Plague Tale: Innocence and mild spoilers for Harry Potter and the Deathly Hallows.

Do you know what game realises its world brilliantly? The first half of A Plague Tale: Innocence. As its title suggests, Asobo Studio's action-adventure tells the story of its protagonists losing their childhood naivety to a cruel and blighted adulthood. To maximise the change these characters undergo, it has them start from a sheltered and privileged background, yanks the bedsheets off of them, and then expels them into one of the grimmest times and places in human history: Aquitane, France, in 1348.

In the mid-14th century, a trio of demons beset France. For decades, the Inquisition had been sweeping through French population centres, raining down righteous vengeance on whoever they judged insufficiently God-fearing. In 1307, they'd tortured 15,000 Christian soldiers because even they fell below their standards of piety.^[1] That subjugation compounded with the Hundred Years' War: an on-again, off-again military conflict between the English and French royal line. It originated from disagreements over who held the right to the French throne and the fiefdom of Aquitane.^[2] And, not to be underestimated, there was the black death. Its symptoms included fever, vomiting, and joint pain. Four of five people who contracted the bubonic form died within eight days.

Smack dab in the middle of this strife, Asobo drops Amicia, fifteen, and her brother, Hugo, five. They belong to the fictional de Rune noble family. Hugo suffers from a hereditary illness that gives him painful headaches. His affliction may have been inspired by the genetically-inherited haemophilia that affected the real British, Spanish, German, and Russian royal families. Hugo's mother, Beatrice, toils away on alchemical experiments, seeking a cure for him. There are a lot of games about characters entering into dark settings, but sometimes they fail to express what their subjects have lost on their entry. Their stories and art don't present a brighter "before" world to contrast the harsher environment. A Plague Tale does, which allows it to pull off a wonderful "crossing the threshold" moment.

Amicia de Rune, her father, Robert, and their dog, Lion, spend a relaxed afternoon in the forest, and it is a beautiful forest. Its radiant autumnal foliage suggests a pleasant period in the family's life that is coming to an end and evokes the idea of death, with the characters soon to enter the more desolate season of winter. When Lion runs ahead of the duo, Amicia follows, passing over a well-defined line between the oranges and yellows of the living woods and a desiccated, grey wasteland. The aesthetic shift extends to the game feel as we go from sprinting carefree across the firm forest floor to plodding through thick mud.

We find Lion lodged in a hole in the ground, and something underneath him devours him alive. He barks. In horror convention, the first kill is the breaking of a taboo. It is a baptism in blood: an omen that the death of other characters is about to follow. Sure enough, once A Plague Tale breaks this seal, it's all violence from here on out. An epidemic is coming for who Amicia and Hugo love.

After a brief interval of peace at Castle de Rune, the Inquisition storm in, for some reason searching for Hugo. As Beatrice tries to sneak Amicia and her brother out of the home, the agents kill Robert and interrogate, then slaughter, the servants. This chapter forms the game's tutorial, and tutorials usually lack narrative or ludological gravitas. They're typically not sections of a game proper but pre-flight checklists we must run through to get to the game. A Plague Tale is thinking about how the nature of a tutorial could align with its emotive goals.

A common approach when educating the player is to wrap them in cotton wool. Active mobs are changed out for wooden targets; pits that could be bottomless, you can climb back out of. Something like that. In Innocence's tutorial, you pass through a genuine stealth section in which enemies could turn around and kill you at any moment. Not yet knowing the AIs' cues or what distracts them, I found this intimidating. However, I had Beatrice's detailed instructions to fall back on. The gameplay invokes the protective element of her character; it mothers you. Implicit in this scene is how frightening it would be to play without Beatrice, and knowing how video games work, we're aware that abandonment is inevitable. It didn't surprise me to see the de Rune mother not make it out of the chateau; her children are forced to leave her to die at the hands of the church. To spare Hugo's feelings, Amicia tells him that their mother is still alive and that they will reunite soon.

From their home, the de Runes flee to a pastoral village of tightly-packed houses. For the first half of the second chapter, Amicia implores locals to help her and her brother but is answered only by locked doors and empty streets. People refuse to leave their homes, the exteriors of many daubed with crosses: the mark of a plague house. This sequence makes solid the social alienation caused by pandemics. No doubt there are a lot of people dead from the disease, but it's also that the living are so scared of catching or spreading it that they decline any human contact. This burg is inhabited, and yet, it's a ghost town. The roads also wind circuitously and double back on themselves, so it's easy to get lost and spend a long time treading the cobblestones, fruitlessly looking for a human face.

The isolation reinforces these characters' loneliness now the Inquisition has taken their parents from them. The recent orphaning of the de Rune siblings also evokes the social distance that comes with quarantine. In fact, it was the plague from which we got the term "quarantine". To defend against these pandemics, the City of Venice required ships to sit offshore for forty days before their passengers disembarked. And the Italian for "forty days" is "quaranta giorni". It's also not a coincidence that the place A Plague Tale first has us encounter the sickness is a village. In towns and cities, population density is high, and as in this village, people at the time often lived in close proximity to animals, allowing the plague to spread efficiently.

The bleak irony is that when Amicia does manage to find people, it leads her and Hugo to a worse fate. A mob of villagers have irrationally decided that the two brought the plague upon the town and that killing them will free them of it. Again, the game uses the plight of these children to talk about a real historical topic. The people of the late middle ages had no germ theory to understand the plague. Nor did they have the antibiotics to fight against it. From their perspective, they were living their normal lives one day, and a week later, vast swathes of people were falling terminally ill and dying in their beds.

To make sense of the chaos and provide the illusion of control, some Europeans turned to the explanation that those outside their religious or cultural circles were to blame. It was common for Christians to mythologise that Jewish people had seeded the black death by poisoning their food or water, and they responded with the mass torture and genocide of Jews. People manipulated the outbreak to confirm their prejudices about external social groups and used those strangers to turn a nebulous inhuman threat into a solid, identifiable adversary they could target.

In a subversion of the usual relationship between protagonists and enemies, Amicia can kill a couple of the villagers with her sling but doesn't see it as a success. She's a fifteen-year-old and didn't come from a rough background. It hurts her to kill other people, but it's what she must do to survive. So, even winning these fights comes with an emotional toll. Like the first death in the story, Amicia's first kill is another seal broken, and similarly, becomes a precursor to more of the same. The chapter's boss is a man who has a borderline death wish, and there's no glory in killing someone in that state.

Despite depicting the reactionary paranoia that can surface during pandemics, A Plague Tale doesn't echo that same universal distrust. A kindly stranger saves Amicia and Hugo from the mob. Later, a monk will guide them on their way, and gradually, they will gather a ragtag band of other young people estranged from their communities by the outbreak. Unlike most fictional explorations of the black death, A Plague Tale is less interested in the direct agony and casualty caused by the malady and more concerned with the social fallout from it.

It's in Chapter Three that we come face to face with the plague, or at least, as face-to-face as you can get with an illness. The de Rune children travel into crypts below Aquitaine to escape the torches and pitchforks. But a cool, dark, enclosed space? That's the perfect breeding ground for rats. Although, rats may not have spread the plague. The plague is caused by a bacterium called Yersinia pestis, which can be carried by fleas or spread via bodily fluids like saliva or through infected tissue. But for years, studies have challenged the perception that there are detailed historical sources to show the fleas primarily travelled via rodents. Computer simulations suggest humans were the primary vector for the bacteria, directly or indirectly.

Still, A Plague Tale finds rats to be a useful symbol in the same way plenty of other art about the plague has. Bacteria are too small to be seen by the naked eye, a limitation for cinematographers and storytellers. The threat goes invisible unless you can chain it to an organism large enough for people to see, and that's what the rats are for. But while the vermin are the game's title antagonist, they don't constitute an enemy in the traditional game design sense.

They're less individual entities and more an amorphous kill zone like the lava in Earthworm Jim or the "Corruption" in Waves. You can pick out single rats in the pack, but they're driven by something closer to fluid dynamics than what we typically label AI. You also can't ever "defeat" the rats. At best, you can avoid them or use light sources to usher them to the corners. Like the plague rapidly overtaking France, the rats seem to expand to take up any space you find them in. Like the disease, they cannot be contained. At best, you can gingerly step around them.

In Chapter Four, the game finally makes good on that threat from the first level, having the Inquisition hunt us down in a stealth gauntlet. There's a tendency in modern espionage games to let the player sneak up behind foes and knock them out, but A Plague Tale doesn't care for that trend. It wants enemies that are more powerful than you. After all, they're full-grown soldiers, and you're just an aristocrat's adolescent daughter. The stage also doesn't want too many chances for you to breathe.

Amicia is eventually able to slip Hugo out from under the watchful eye of the church, but the boy's headache worsens to the point that we must carry him on our back. The reduced pace we move at as Hugo's mount brings our attention to Amicia's burden in trying to support herself and her young brother at her age. The game generally does a great job of giving us a brother-sister relationship with Hugo by lending him some characteristics of an independent entity and others that make him a limb of Amicia. We can command him to perform menial tasks during puzzles and must leave him somewhere safe while we tackle guards. He can also run off on his own, which draws worrying parallels to Lion in the prologue. However, when he holds our hand as we escort him, or in this sequence in which he rides on Amicia's back, our avatar effectively consists of both characters.

In Chapter Five, we cross one of the battlefields of the Hundred Years' War. There are a few living soldiers from the English army in these hills; the English did make some sizeable gains in the conflict in 1347, the year before Innocence takes place. However, the vast majority of the military we see are corpses, giving the impression that nothing meaningful has been gained, but plenty of lives have been lost. The corpses aren't just set dressing either. Amicia and Hugo are forced to come into contact with the reality of the war by literally wading through them.

Many of the bodies are impaled with arrows, and there are bogs all over this level which can leave you a sitting duck for hungry rats. While the Battle of Agincourt, the most famous of the Hundred Years' War's clashes, took place 67 years after this game, there are parallels between it and the battle prior to Chapter Five. Torrential rain in the days leading up to Agincourt softened dirt into deep mud, which slowed French troops, allowing English bowmen to decimate them.^[3] Archery came to prominence as a military art during the war, and historian Kennedy Hickman describes English archers as the "backbone" of the country's forces. See also the English's use of bows and arrows in their victorious battles at Crécy (1346) and Poitiers (1356).

Like the Chapter Five battle, Agincourt left plains of dead, with the English army murdering between 6,000 and 10,000 soldiers.^[3] Innocence also relates the military engagements to the plague. The battle that preceded the de Runes' arrival has left piles of dead bodies that have attracted rats, aggravating the epidemic. As we tip-toe around this no man's land, Hugo gets kidnapped.

In Chapter Six, we infiltrate a camp to win him back. On the English's home turf, we have a chance to view the war from their perspective. You'll see more English archers here. There are also a couple of men among the tents denigrating King Edward III for continuing incursions into France, a reflection of the drawn-out nature of the war. By this point, it had been raging for eleven years.^[2] At the end of this level, there's a sudden switch in power relations between our foes and us. The English army has agreed to hand Hugo over to the church. The troops spot Hugo, Amicia, and some of their newfound allies, and we are faced with a shooting gallery in which we must take out many enemies in quick succession. Amicia is dragged into the war, becoming a French soldier herself, and A Plague Tale features more action sections going ahead.

We often say that video games are better when we have distinct choices about how to tackle problems. We sometimes treat immersive sims as guiding beacons in the medium because they will let us take a melee, shooting, diplomatic, or stealth path through a level. Yet, there are sections of A Plague Tale that hit home emotively because we are locked into one mode of engagement. If the game gave you the option of an action-first approach when you initially encountered the Inquisition, it wouldn't get across how the Catholic forces totally overwhelm communities. If it permitted you to stealth around the English soldiers at the end of Chapter Six, Amicia couldn't be metaphorically drafted into the war, a turning point in the loss of her innocence. But I'm sorry to say, this is about where A Plague Tale's dramatic spring runs dry.

From Chapter Seven onwards, the game has its moments. There's a level in which we play Hugo. It individualises him and gets to his complete lack of defence against the adult world. The camera is down at floor level with grown-ups hovering over you like the columns of a great cathedral, and you have no attack options to fend off anyone wishing you harm. In the back chapters, we witness a sadistic Inquisition storming into plague-filled suburbs, and the plot touches on faith as a medicine. Some people thought the disease was god's judgment for declining religious fervour. So, forcing people into deeper Christian devotion would be a cure.

It's also not that the front half of the game is perfect. The dialogue is unseasoned and embarrassingly obvious. A Plague Tale is the kind of media where if a character is cold and hungry, they'll say something like, "I'm frightfully cold and ever so hungry". But at least up to the midpoint, the game is giving a face to suffering in plague-era France. The further in you get, the more it betrays that mission for the sake of ahistorical and supernatural sensationalism.

We discover that Hugo and Amicia's mother, Beatrice, is still alive, and being held by the Inquisition. We free her and concoct her alchemical recipe to banish Hugo's increasingly debilitating symptoms. As it turns out, Hugo's congenital disease is actually a psychic power that lets him control rats. The Inquisition has been hunting him so that the Grand Inquisitor, Vitalis Benevent, can inject himself with the boy's blood and gain the same mind control. The final boss battle is a sort of Harry vs. Voldemort showdown where Hugo must use dark rats to defeat Vitalis's light rats. Hurricanes of the creatures fly across a nave's stone tiling while swirling columns of them slam down around Amicia. It couldn't be much further from where the game starts.

A Plague Tale's historical grounding constituted a form of remembrance for real victims of disease and slaughter. Even the game's visuals work to that end. Its textures have the pixel count to capture the grime in the corner of every basement and the weathering on every paving slab of the courtyards. They're indicative of a world that's been both incubated and been worn away by disease. And on top of the high-tech textures, models, and lighting, the reality of the game is held in the subtle variations between props.

A Plague Tale is not as quick to copy-paste objects in the same orientations as its peers. It does reuse and recycle. A torch is a torch, a rat is a rat, and if you really stop and stare at the beams on the cottages, you can see the same splintering repeat from one to another. But there's also plenty of messiness and non-uniformity that helps us suspend our disbelief. Some of the chairs in the entrance hall of the de Rune villa have crimson cloth folded over them, and others don't. The columns at the back of the church in Chapter Three each have their own unique abrasions. The plants in the prologue don't all coil back from the path; they jut out into our way at stochastic angles. The photographic style of Innocence affirms that its events are happening in the world that we know. That displaced 700 years, we would find ourselves in the shoes of these poor peasants and bloodied soldiers.

The game's collectables further describe the plague as it happened. They don't simply increase a counter or add side stories. They're everyday objects that come with a description of how they are relevant to history. When you collect the soaps, you can read a short blurb on these toiletries' increasing commonality in Europe at this time. People reacted to rampant disease with more stringent personal hygiene. Or there's the whip which the game tags with a short explanation of the flagellants. The flagellants were a religious order who engaged in public displays of self-whipping as a conspicuous religious penance. But the whips, now covered in decaying human flesh, became a health hazard, helping spread the plague. It's impressive the quantity of items the game links back to the disease and Inquisition without stretching, and it's proof of how much this thin sliver of history changed our world.

The story's refusal to negotiate with fictional rules is also why it was so grim. A Plague Tale's early setting doesn't have the theatric bravado of high fantasy. Not only is the average person likely to die in this world, but it won't be a glorious vanquishing at the hands of a legendary monster. The majority of people who perish in this setting pass away sitting in filth from the slow rot of disease, isolated from their family and community. Even those who fall to the Inquisition mostly get run through with a sword without any ceremony, and then the soldiers turn their sights on the next person. Death is not epic or heroic; it's indiscriminatory and meaningless.

As A Plague Tale burns off its realism, the look, collectables, and narrative setup left in its ashes don't speak to its new goal of being a sword and sorcery adventure. It even soils one of its most promising analogies. The symbology and ludonarrative had been working with this relatively sophisticated metaphor where, not only do the rats represent the plague, but they also represent the Inquisition, and vice-versa.

Both the plague houses and the Inquisition wear crosses. The Inquisition uses rats as its attack dogs. Amicia has a nightmare in which Vitalis corrupts Hugo, and at the moment he does, rats consume her brother. The Dr. Doolittle power Hugo sways the rats with is called "Imperium", referencing the Catholicism of the Holy Roman Empire. Like the plague rats, the Inquisition surges across the country, hollowing out towns and massacring the populace; they are a disease to Europe. The rats and the Inquisition prescribe a narrow path for you to walk, and if you step even a foot out of line, they will eat you alive. But this read on the game loses its coherence once you see that the rats can work against the Imperium and that there are good rats and bad rats.

This hard right into fantasy also takes characters with heartbreaking odds stacked against them and gives them an easy out. Much of the peril in the story came from the fact that the plague and military forces are everywhere and are ruthlessly effective at killing. The black death had a mortality rate of about 80%, so it feels like it will take a miracle for Hugo and Amicia to make it out alive. But a miracle is what they get via restorative potions and magic spells. In general, the game is of two minds about how inhospitable its setting is. On the one hand, you have the hostile and indifferent world that would burn a five-year-old's home to the ground and destroy communities with disease and religious persecution. On the other hand, the setting has a just-so quality that destines success. A little boy's fantasy of the Inquisition sparing his mother comes true, and even the might of a plague can melt to convenient coincidence. When our path appears to be blocked, there's always a brazier we can light in just the right spot to scare off the rats or a flower bed right where we'd need some cover from patrolling troops. It's a contradiction in the design of many darker AAA games.

The real people of the Middle Ages had a far more horrendous future ahead of them. The Catholic Church would dispatch Inquisitions for almost five more centuries. Its vilest, the Spanish Inquisition, would kill 32,000 people.^[1] The Hundred Years' War actually sustained for 116 years of bloodshed.^[2] That would be like if we'd been at war since 1907. An estimated 25 million people succumbed to the plague, almost a third of the population of Europe.

In its epilogue, the game claws back a little of the solemnity in that history. At first, the final chapter looks childishly rosy, as though Amicia and Hugo haven't suffered any substantial loss. The siblings visit a fair, and those skills that Amicia once used to kill, she turns to a carefree carnival game. But then the two try to enter the town's main square and are denied because they're still wanted by the Inquisition. Rejected from the festivities, they leave town with their mother on a horse-drawn cart. They laugh and tease each other as they disappear into the distance, but where the game started in a vibrant autumn, the wagon now appears tiny against a snowy landscape of dead trees. The world they continue into is much more lifeless and unforgiving than the one they started in. Thanks for reading.

Notes

1. Inquisition by History.com Editors (Jan 10, 2023), History.
2. Hundred Years' War by History.com Editors (Aug 21, 2018), History.
3. 9 Things You May Not Know About the Battle of Agincourt by Evan Andrews (Jan 2, 2019), History.

All other sources linked at relevant points in the article.

When I gave an update on Halo Infinite back in August, I described a youthful and energising FPS sandbox but a neglected wreck of a live-service game. At the time, I predicted that it would be a while before the shooter's condition improved. What I didn't guess was that it might get worse. One of Infinite's most hotly awaited features was Campaign Co-op. Those who've tagged along for the rocky history of 343 Industries no doubt found the wait familiar. As the developer geared up for the release of Halo 5: Guardians in 2015, head of development, Josh Holmes, appeared to confirm on Twitter that the shooter's single-player would include split-screen co-op play. Microsoft subsequently clarified that Guardians would not support local multiplayer in any form.

Cutting this feature dealt a crippling blow to Halo 5 that colours the perception of it to this day. Earlier Halo games had been groundbreaking in opening up group experiences in the genre. For many, the series had provided their first or formative contact with split-screen action gaming. It's why, at a 2017 appearance at the DICE Summit, 343 boss, Bonnie Ross, described cutting the feature from Halo 5 as an "incredibly painful" decision for the community that "eroded trust" between 343 and its fanbase.

It's this dear attachment of the Halo fandom to the local multiplayer that meant they were always going to be embittered by Halo Infinite launching without split-screen Campaign. It's the reason that it's a big deal that after changing the predicted release date for the mode twice, on September 1st 2022, 343 announced that there would be no split-screen Co-op Campaign for Halo Infinite. It's bad enough that 343 dropped this series cornerstone. It's worse that it tacked its audience along for months, assuring them the mode was just on the horizon when in actuality, it would never arrive.

Infinite's history has been replete with delays and shortcomings, and local Campaign Co-op was one of the strongest incentives for players to stick with the game. Yet, the studio retracted it without any official announcement on its website. Its cancellation was announced off-hand in a thirty-minute update video and was accompanied by no apology. Surely, that's the least your audience deserves. It also worries me that the reason Joseph Staten gave for the exclusion of the feature was that the studio wanted to redirect resources to other projects. What on Earth is going on at 343 that the Halo developer doesn't have enough resources to implement every essential mode?

Beyond Campaign Co-op, the major issues with Infinite fall into three columns:

1. The absence of Forge, Halo's proprietary map editor.
2. A drought of long-term challenges and rewards, mostly because there have only been two full Battle Passes since launch.
3. General nicks to the game's polish and functionality, like glitches and lag.

In some sense, these are all the same issue. The game is meant to receive feature and content updates, as well as quality-of-life improvements, at the start of every season. So, what 343 needs to be doing is introducing new seasons regularly. And they haven't been. While play periods were originally slated to last three months each, both Season One and Two were extended to be six months long. November 2022 should have been the start of Season Three. After what felt like a lifetime of waiting, we'd be looking at a new Battle Pass and a new Halo. Except the other bombshell 343 dropped in their September update was that they were extending Season Two another 150 days. Within Infinite's first fifteen months of operation, there is a ten-month gap without any landmark updates.

Maybe talking about Infinite's run up to now as "two seasons" is misleading; it's more like 2.5 seasons. The studio's thinking was that if it couldn't introduce a whole new content run for winter, it could at least release an update with pint-sized versions of all the accessories you'd expect in a season. Instead of a full set of completed features, there's a couple of new Campaign options and betas for two other mechanics. Instead of a paid 100-level battle pass, there's a free 30-level Battle Pass. Instead of two new maps built from the ground up, there are three new levels built in Forge. We had a budget Christmas this year.

Despite local Co-op going MIA, we can now replay in-game missions and enjoy online Campaign Co-op. Most other players and I have found the new campaign components to be reliable. In general, the Winter Update went off a lot smoother than introducing Season Two did. There are also new Campaign achievements like "Splatter 50 enemies while riding in a vehicle with another player" or "In the Nexus, kill the Hunter pair with melee final blows". Even if you've already played the story front to back, there's still reason to galavant about Zeta Halo with a vehicle or power weapons, either solo or with friends.

In the multiplayer, not everything is likely as you remember, either. For 343, no weapon modification is too minor if it equals out the competition. The update log is full of painstakingly accurate notes like "[Commando Rifle] Aim Assist Angle [increased] from 5 to 6.25 degrees" or "[Plasma Pistol] charge shot angular velocity [increased] from 20 to 35 degrees per second". To my eye, the most substantial change is a fractional reduction to the aim assist on the Battle Rifle that makes this headshot generator require finer edging onto its target. But turning heads is the December nerf of the Energy Sword, which prevents you ripping through whole teams. Not to be overlooked are the devs' refittings of the Pulse Carbine, Commando, and Disruptor. Long viewed as inferior pretenders to serviceable weapons like the Pistol and Battle Rifle, a combination of reassessment by the community and buffs from the devs resulted in a newfound respect for these firearms. Given their unique underlying functions, like acting as a mid-range homing rifle or temporarily disabling the enemy's power to recharge their shield, their rehabilitation has further diversified the play.

If nerdy stats are your cup of tea, then you'll also love the metrics 343 has added to your profile on the Waypoint site. This is an addition that was made a little before I uploaded my last article, and passed me by. You can see your medals, match history, and advanced post-game stats on your page, giving you further accolades to bask in after your victories and providing more of the persistent rewards and challenges that Infinite needs. Although, I'm still not sure why the service records for previous Halo games had to be scrapped.

The game also uses a new set of rules to calculate your Competitive Skill Rating: the persistent measure of your talent on the battlefield. If a teammate quits a ranked match and you quit after them, you'll receive a reduced CSR fine. The idea here is that if someone in your party bails on a game, you no longer have to choose between playing through several minutes of a one-sided match or enduring a stinging injury to your CSR. A minimal player base means imbalanced matches, and imbalanced matches mean more motive for players to quit, so this leniency just makes sense.

The less advisable change to the system was reducing the weight personal performance has on your CSR relative to your team's performance. If my team loses a match, but I performed capably, I want my ranking to honour that. Now, I can run circles around opponents and still have my points docked. It's annoying. It smarted especially before 343 had ironed out many of the connectivity issues on the servers. Ping is now lower, and on average, more stable, but even a rock-solid connection is still liable to jitter now and then. Unfair CSR penalties were also worse immediately after the Winter Update when I was encountering an inordinate number of losing matches.

I can't overstate how much of a curse on Infinite its current player matching is. The estimated time to find a match is so inaccurate that it can predict a thirty-second wait, then you can hit the button, and have the program time out looking for a session. If Halo had a healthy player base, the discrepancies on the estimates might only be a few seconds, but we're long past a time when that was true. Average Steam users for December were sub-4,000, and at the time of writing, Infinite is the 25th most-played Xbox game. Following the FPS in its first year has felt like playing another shooter in its twilight days, which creates the uncomfortable impression that maybe this title is too. My biggest fear is that even if the latest Halo does get patched up, the ship might have sailed on its recovery.

The multiplication of playlists throughout the Summer-Autumn period further diluted an already sparse player corps. More playlists let 343 widen the pool of experiences available and bring back fan favourite modes without having to create a lot of new content. But all these fresh hoppers pulled players every which way, only increasing search delays. With the November update, 343 pruned the less popular playlists and refocused on a core set of social and ranked play formats. It speaks volumes that Last Spartan Standing, which was the signature addition for Season Two, is nowhere to be seen on this roster. Attempting to make a Halo battle royale only feels more like a misallocation of effort in retrospect. Now, 343 rotates out the more niche gametypes from week to week, allowing it to keep the multiplayer fresh and faster. Yet, player abilities, both within and between teams, are still lop-sided, and Ranked Matches especially are time-consuming to find. There's no substitute for making a game that people want to play.

The developer is finally making you jump through fewer hoops to complete your weekly Challenges, and that's a step in the right direction. Challenges used to be hyper-specific and often protracted, rewarding you for taking to weapons and gametypes that you hated and pursuing strategies that made you a detriment to your team. With regular players having filled out their Battle Passes months ago, weekly cosmetics were the only trophies regulars had left to win, and you were delivered them when you finished all your weekly Challenges. Yet, earning one of these cosmetics required completing 21 tasks in seven days, and with the indifferent measures for player matching, was largely down to luck. Now, Challenges are more flexible, don't take as long to finish, and you need fewer to unlock your weekly fashion statement. It's out with "Earn 7,500 points in Rumble Pit" and "Get 10 kills with the Stalker Rifle" and in with "Score 5,000 points in PvP" and "Get 5 kills with Banished weapons".

You can also now earn XP bonuses for your placement on the scoreboard, and any wins, again moving the mechanics towards this principle that rewards in empowerment games should be comparative to success. But with Infinite's proneness to technical snafus, even a feature as rudimentary as XP assignment remains glitchy two months after revamping. And reward systems mean nothing if you can't get the prizes you want from them. In that department, Infinite continues to fail players. The new Battle Pass is a welcome meal after a long fast, but not only is it less than a third the size of a typical Battle Pass, it also doesn't have the glitzier cosmetics like Charms and Kill Effects that players put a high price on. If you've not made your start on it yet, I sure hope you like shoulder pads. After you've hit Level 30 on that pass, the new XP goes nowhere, and the company has now run out of weekly cosmetics and is looping back through its line for 2021-2022. There is no material reason left for any semi-frequent player to keep checking in on the game.

The fandom has further been peeved that the new PvP maps (Argyle, Detachment, and Empyrean) were fabricated in Forge instead of professional tools. Sure enough, the rocky outdoors of Detachment has that cheap, grainy look that the story mode's exteriors do, while the artificial surfaces in all these environments have the opposite problem. A flat, untextured appearance has been characteristic of the primitives in Forge in the past, and that continues to be true. Empyrean is also let down by its gaudy yellow and purple. Yet, I can't deny that in the level design of the winter maps, 343 once again show themselves to be top of their class.

The first two environments are eccentric. On Argyle, a wide central channel provides plenty of opportunity for the brave Spartan to hightail it from one base to the other with flag in tow. The side routes off that thoroughfare are snaking and elaborate, and where most maps include one of each power weapon, Argyle spoils you with two Shotguns and two Snipers. The level caused some machines to crash when it was first instated, but this bug has since been squashed. Detachment is nothing like I've ever seen in a shooter. It splits into three areas: On one side, you have a tiered indoor facility like the outskirts of Argyle; down the centre, you have a plateau with a divider preventing people in the bases from taking potshots at each other. On the other edge, you have a floating island from which snipers have free reign over players in the map's core. My one frustration is that with team flags being two staircases off the ground, you can be practically at the same X and Z coordinates at them, but they can still dangle out of your grasp. It's maddening, like when you drop something down the back of your desk, and your arm is just a couple of inches too short to reach it.

Finally, Empyrean is a remake of Halo 3's The Pit. It's a classic map of ups and downs with snug high-wall bases. New to this incarnation is an Overshield perched before a sheer drop into oblivion: an enticing gem for any player who thinks they can grab the power-up without getting repelled off the edge. This Pit 3.0 has been an instant hit. 343 was right to crank out a hefty number of original levels before falling back on Bungie's work. We already have a decade of maps from the original Halo smiths, and fans will inevitably imitate them in Forge. Still, with fourteen original stages already under 343's belt, I'm not going to say no to a little nostalgia.

With these three Forge arenas, 343 shows the startling power of their new map engine. Having played video games since I was a kid, not much shocks me anymore, but I can scarcely believe the options included in Infinite's Forge. You can switch the materials props are made of, sometimes on different parts of the same object. You can change properties of the assets like colour, roughness, or reflection index. You can attach magnets to items so that they snap to others on specific faces. You can name them and group them together to make them easier to manage and replicate, and share groups with other builders. You can set the angle of the sun and the primary and secondary colours it casts on planes, and choose the ambient sound for the map. You can view and edit the navigation data for the AI and see heatmaps and other analytics showing you where the engine is focusing its lighting calculations, allowing you to optimise your system memory usage.

One screen lets you construct node graphs: structures used in professional software development, to script original gametypes and missions. Already, crafters have used these trees to create auto-turrets, the bananas from Mario Kart, and the first loop from P.T. While there are plenty of "Forge has arrived" headlines out there, keep in mind, this is only a beta. For me, that's sometimes meant textures or even whole maps won't load. In its typical sluggishness, 343 forecasts that it won't have the feature finished until at least late June. I'll also warn you that this tool has a steep learning curve, but that's only because it is verging on real dev tools. Unfortunately, Forge is so stand-out because it's the first of Infinite's modes that isn't missing multiple core components. It's, at minimum, five months off completion, and it still feels like the most finished part of the experience.

It's also easier said than done to get those maps out of Forge to a test track. Chances are you can't gather seven friends in one of these community locations because your friends probably aren't playing Halo Infinite. So, we're reliant on the "Custom Games Browser". Scheduled for release in March, 343 busted their ass to get it out the door pre-Christmas. They figured it was better to have an unsound browser in the wild than no browser at all, which is correct, but there's still no mistaking that this was a premature release. Even more frequently than the Forge editor, it locks up during loading screens. Perhaps the worst Halo bug I've experienced is that when trying to enter matchmade games after playing a custom, I've had the program drop my avatar into the session idle, me unable to control them until I'm eventually kicked from inactivity. This has happened multiple times, but for what it's worth, I've been unable to replicate the glitch in the past week.

In both matchmade and custom games, we can, after an eternity, mute players without opening the Xbox overlay. About half of LIVE chatter is living room background noise, and harassment, including abuse targeting minorities, remains prevalent on the network. To leave basic "ignore" tools out of the game for a year was to leave players unsafe. It also takes too many button presses to mute someone, but 343 got it in ahead of schedule, and it is the rare system that works on release.

Shortly after 343 Industries' September update, Bonnie Ross made it known that she would be stepping down as the head of the studio due to a family illness. It's telling that her old job was split into three new jobs. Microsoft evidently believes that the developer has been suffering from a lack of managerial hands on deck. Whether changing that actually fixes Infinite is anyone's guess.

Infinite's Winter Update and the smaller expansions that wreathe it are tangible improvements to a service that had gone for months without any substantial development. But I still think the best compliment you can give to 343's recent patches is that they're "promising". That's a problem because, after all the delays and breakages and the Co-op cancellation, I don't think a promise from 343 is worth anything. In this article, I could have gone into more detail about the studio's roadmaps and schedules, but the reward for investing in those prophecies over the past thirteen months has been getting kicked in the teeth again and again. Even after five articles on the game, I haven't touched on every one of its failings because there's just so much wrong. Unless 343 finishes plugging the base features into Halo Infinite, an apparently insurmountable task, I won't believe they're capable of stewarding this game into the future. Thanks for reading.

Notes

1. Credit for the Forge map Eternity goes to I am a Luxury.

2022 was a tough year for plenty of businesses, and the video game industry was no exception. Workers continue to contend with stifling conditions, and some releases we had high hopes for hobble on broken and disappointing. But we can still sink our teeth into titles old and new that affirm, delight, and empower. Every year, I select the top ten games that I played for the first time and put them in a list, criticism-style. Here is the rich crop for 2022.

20th Century Food Court

One entry in Zachtronics' Last Call BBS collection, 20th Century Food Court feels like the game that the studio has been preparing to make for 22 years. Barth's mechanical design is nuanced enough that a minimal change in the problem he poses can demand a radical change in your solution. Food Court is the epitome of a puzzle box that looks simple enough to solve on first sight but elegantly unfolds to show intricate logical complications. With a surfeit of components in every Rube Goldberg circuit you construct, you're left with a vast body of work to be proud of at the end.

Genesis Noir

If you're painting a portrait of the universe, realism might make too much sense as a style. Space is a peculiar place, so it's somewhere surrealism shines. Few pieces of media demonstrate a better application of that style to a scientific end than Genesis Noir. Laser-etched comic panels collapse into each other with a smoothness and vive that matches the soundtrack's jocular jazz. Genesis Noir is about the comfortable constants we can bring to mind when reality is changing all around us and our dreams wash down the drain. You've heard that the personal is political; this rain-soaked ramble to the start of time believes that the personal is physical.

Gorogoa

Note: Mild spoilers for this one.

Gorogoa is a poststructuralist maze of framing and reframing. It challenges your confidence in your own perception by forcing you to recast scenes and objects to solve puzzles. Perhaps an old man's daydreams are also the pictures in a storybook. Perhaps that Sun could be a gear, and the castle walls are teeth. If Genesis Noir finds peace of mind in consistency, Gorogoa is saddened but intrigued by the idea that everything must pass, that the old order must fall to bring about a new one. I was flabbergasted that right up to its credits, its puzzles were still surprising me with those transformations. Gorogoa twists my head with a narrative that simultaneously has occurred, will occur, and is occurring.

The Legend of Zelda: Breath of the Wild

You don't see publishers taking many risks with the names which sell millions. You especially don't see publishers soothe rousing action epics into serene meditations that ask for sustained deliberateness from the player. Breath of the Wild values forethought and preparation as much as execution and embraces our tendency to ascribe meaning to the products we create ourselves. It does all that while shedding the impenetrable convolution of the survival genre. The newest Zelda asserts a feudal Japanese sensibility not just in its environment but through an atmosphere of quiet and measured practice.

Off-Peak

You might know Off-Peak as the prequel to the tilted Norwood Suite. There's an amusing spontaneity to Off-Peak that makes it feel like it's coming up with its quirky characters and bizarrely composed setpieces off of the top of its head. I want a big fish. Its artwork is blessed with crisp lines and flat, loud shapes. Its high-tempo synth is effortlessly cool and always on the move, making exploration of Cosmo D's lively train station a trip.

Outer Wilds

Experiments are the building blocks of Outer Wilds' play. In fact, they're the reason for its characters' existence; they live to learn. By placing a deep cushion under us for when we tumble from the heavens, Outer Wilds inspires us to try, fail, and try again, rocketing, teleporting, and crash-landing our way to enlightenment. We adopt the probing spirit of science on the holy mission to unlock the secrets of the universe. An invigorating and transcendental ending in which all our theories align caps off this optimistic thesis. Outer Wilds believes in the power of science to revolutionise our planet and every other one too.

The Outer Worlds

It's a long time since I've come home to a Fallout-like, but having cracked the seal on The Outer Worlds, I found all the old charms still there. There's bamboozling a surly guard with a slick speech check, locking in a generous new perk to make me a formidable combat machine, and laying my grubby fingers on fascinating retro-future tech. But as crafty as an Outer Worlds player can get with a pistol or a lockpick, any improvement you bring to its rusting colonial dystopia must come gradually, and there's familiarity in that too. The Outer Worlds lays out a solar system which has run aground and where people have few to no reprieves from their strife. It makes the relationships you form with its valiant misfits all the more precious.

Scorn

In Scorn, you are sick. Sick with a disease that causes your body to mutilate itself, sick in a way where everywhere you look, you see your own mortality. In painstaking detail, rigid metal architecture is matted with inflamed viscera, and mindless lumps of flesh wander the halls looking for a victim to gore. There's also plenty of mechanism in this biomechanical hellscape. Machines protest any attempt to shunt their gears out of rigor mortis into feeble functionality, and interfaces defiantly resist interpretation. Scorn is a game hurting itself, Scorn is a game that does not want to be played, and Scorn is beautiful.

Super Mario Odyssey

Like Breath of the Wild, Super Mario Odyssey represents a series that traditionally said, "Here's your goal", now asking, "What do you want your goal to be?". In previous 3D Marios, we pursued one Star or Shine Sprite per chapter. But Mario Odyssey only conceives of open levels where we collect its Moons on cosmic Easter egg hunts. These targets are common environmental features which means we can suddenly reorient ourselves in relation to them and gorge ourselves on whole baskets at a time. We can draw a line from the freedom and exuberance in the play to the world that is dancing in motion to its own beats. Some of the odder aspects of Mario have become normalised over time. However, Odyssey rejuvenates the series' off-the-wall creativity with walking Maoi statues and the power to turn into pure electricity. Mario, Bowser, and every other character feels hyped to dance at this bizarre planet-hopping party.

Unpacking

There is something about pixel art of real-world objects that gets my serotonin running, and I've not found a better gallery of it than Unpacking. Look at this game's measuring cups, stuffed ducks, and family photographs and tell me there's not abundant talent on display here. We slot these sprites together into a satisfying domestic jigsaw and can make unique decorating choices on a precise scale, allowing us to exert ownership over the homes in each level. With its organic interiors and soft surfaces, Unpacking is comfy and reassuring. It also manages to tell a shockingly fleshed-out three-act story, and our character's barriers to happiness drag it back from the brink of mawkishness.

___

2022 is in the fucking ground. Honourable mentions go to Anger Foot, Baba is You, Bugsnax, Eliza, Neo Cab, Opus Magnum, A Short Hike, Silicon Dreams, and Unravel. Thanks for reading.

Note: The following article contains major spoilers for ChipWizard Professional.

You may have heard that the transistor was one of the most important inventions of the 20th century or one of the most transformative technologies period. And you might wonder why it enjoys that esteemed podium in the history of electronics. In my 20th Century Food Court article, I stressed the significance of logic gates in computers. You'll recall that these circuit elements take some input in, perform a logical operation on it, and then cough up an output. But I left logic gates as black boxes. Electrical signals enter, something happens, and then a different signal exits. So, what's the "something"? How could the same lump of silicon produce different data each time? Well, Zachtronics has a game about that too.

ChipWizard Professional is another exhibit in the studio's Last Call BBS collection, and every puzzle within it commands us to make a computer part out of transistors. Playing the game, we quickly realise the role of these parts: they're switches that either block an electrical charge moving through a circuit or allow that charge through. But where we must manually toggle a physical switch, transistors can open and close their path through the circuit in response to an electrical signal. Because these doors can be controlled electronically, software, rather than a human, can decide where in the device current flows. And software can make many more decisions per second than a person can, making computers the masterminds they are.

The transistors we work with in ChipWizard are "BJTs": Bipolar Junction Transistors. As we see in the play, BJTs come in two types: NPN and PNP. To avoid overloading you, I'm going to start by only describing NPN transistors. Bipolar Junction Transistors connect to the rest of the circuit at three points: the base, the collector, and the emitter. As depicted in the game, if there is no charge coming into the base, no electricity will flow between the collector and emitter. If, however, you route an electrical signal into the base, current will push from the collector to the emitter. You can think of the collector as the entrance, the emitter as the exit, and the base as the lever that opens up the flow.^[1]

To have us construct transistors, ChipWizard needs to provide us with the basic materials that can compose them: p-type and n-type silicon. When designing a technology, the materials we choose match what we want that tech to do. When dreaming up our transistor, we know we can't make them out of insulating materials because little charge would flow through them. However, we also can't use anything too greedily conductive because then these parts couldn't dam a circuit. Therefore, engineers make transistors from semiconductors: materials that fall halfway between conductors and insulators on the conductivity spectrum. You've probably heard that silicon is the category killer in this field. What you might not know is that you never use pure silicon for this purpose.^[2]

We'll need to go back to high school science to understand why pure semiconductors don't make good transistors. Inside atoms, electrons are arranged into "shells", which we can think of as concentric circles extending outwards from the atom's nucleus.^[3]

In reality, electrons aren't little balls that exist at pinpoint locations; they're more like fields, but this illustration of the atom (the Bohr model) reduces friction in our understanding.^[3] Electrons are negatively charged, while the protons in the atom's nucleus are positively charged, and those opposites attract, holding the electrons in orbit.^[4] The electrons in the shell closest to the nucleus are most strongly attracted to it, while the electrons in the shell furthest from the nucleus (the valence shell) have a higher energy level and drift further from the nucleus than others. Sometimes, the connection between an electron in the valence shell and the nucleus is weak enough that outside forces can knock the electron off the atom.^[3]

An electron detached from an atom is a free electron. A free electron can push a loose electron off of another atom, which can displace more electrons, and so on. The potential for outer shell electrons to be knocked free is what we use to define conductors and insulators. Conductors have loosely bound outer electrons that can easily be jailbroken, while insulators refuse to give up their outer electrons. These free electrons moving through the material constitute an electrical current and can also enter the outer shell of other atoms in the material ready to accept them.^[4]

Many elements follow the "octet rule", in which they accept or eject electrons until they have eight of them in their valence shell. Atoms can reach eight electrons through importing electrons from or exporting electrons to other atoms. For example, if an atom has nine electrons in its outer shell and it gives one away, it can drop down to eight electrons. But an atom can also become a magic eight ball by forming covalent bonds with other atoms. Covalent bonds are chemical bonds in which atoms attach themselves to each other by sharing electrons. The same electron or electrons will exist in the valence shells of both atoms. It's like two people being stuck in Chinese finger traps and the electrons are the traps.

A silicon atom has four electrons in its valence shell.^[5] So, if you surround it with four other silicon atoms, and each shares one electron with our central silicon atom, the central atom can reach its goal of eight valence atoms and obey the octet rule. If we surround each of those edge atoms with three more silicon atoms which each donate one electron, it's possible for them to have eight electrons in their outer shell too. Because of the octet rule, silicon will covalently bind together into a neat grid, letting us create the wafers of silicon that we can use in transistors.^[6] But we were meant to be making a semiconductor: a material in which electrons could be knocked off of atoms and passed onto other atoms, and that's not going to happen often in this lattice in which every particle has already fulfilled its allotted quota of electrons. These silicon atoms have already reached their target eight electrons and will rarely capture or release those charged particles.

This is one of the great lies of pop science: silicon is not a particularly good semiconductor. Even electrons in this silicon lattice can occasionally be severed from their atoms by outside forces, but their flow is still not free enough for a transistor. But there is some chemistry we can feasibly perform on our element to make it a decent semiconductor.

To reiterate: lone silicon atoms have four electrons in their outer shell, and therefore, can share two electrons with other silicon atoms on four sides. But if you inserted a few atoms with five outer electrons among those silicon atoms, those five-electron atoms would form four covalent bonds with the surrounding silicon. Then, they'd have nine outer electrons, and they'd want to give that extra electron away to drop down to eight outer electrons and close on the octet rule. That freed electron is going to get pushed into the outer shell of a silicon atom in the lattice, which will then try to offload one of its electrons to drop down to eight, which will affect another neighbouring silicon atom, and now, we have an electron being passed through our matrix. Electrons are negatively charged particles, so you have a negative electrical charge propagating through the material.

We could, alternatively, seed the silicon with an atom that has three electrons in its outer shell: one down from silicon's four. These three-electron atoms would form covalent bonds with the silicon on three sides, but on the fourth side, they wouldn't have an electron to share with the neighbouring silicon atom. Both the three-electron atom and the fourth adjacent silicon particle will only have seven electrons in their outer shells and will steal an electron from the surrounding lattice to make eight.^[6] There is a "hole" (the official term) in the shell that is positively charged, attracting negative particles. Stealing the electron to fill that hole will create a hole in the outer shell of another silicon atom though, which will cause that atom to steal an electron, creating another hole. Now, we have a positive charge bouncing through the wafer.

The process of introducing these atoms with five or three outer electrons to the silicon is called "doping". That silicon that has five-electron atoms in it and a negative charge running through it is called n-type silicon. The silicon that has some three-electron atoms and is batting about the positive charges is p-type silicon.^[6] N-type "dopants", the ones with five electrons to share, include arsenic, antimony, and phosphorus. P-type dopants, with three valence electrons, include boron, gallium, and aluminium.^[7]

We can observe in ChipWizard that we need both p-type and n-type silicon to make a single bipolar junction transistor. In the game, we fabricate an NPN bipolar junction transistor by drawing a line of blue n-type silicon and then having a line of red p-type silicon overlap it at a 90-degree angle. It's important to understand that in the real transistor, the p-type silicon is not sitting atop the n-type. Instead, you have a bank of p-type silicon with two n-type regions on either side.

These transistors are "NPN" because if you orient them so the collector is on the left and the emitter is on the right (or vice-versa), reading across, you have some n-type silicon, then p-type, then n-type again. One of those n-type chunks is the collector (the entrance for the electrical current), the smaller p-type stripe in the middle is the base (the gate for the current) and the n-type region on the other side is the emitter (the exit for the current). Remember, n-type has the atoms with excess electrons, and p-type has the atoms with excess holes.

If we push some electrons into the emitter, it will start that long chain of silicon and n-type atoms offloading their electrons. Eventually, the emitter will push its electrons out into the holes in the p-layer that makes up the base. Now, if we have some holes in the atoms outside the base (which we can get by positively charging that portion of the circuit), those electrons will flow out of the base and into the circuit. So, we have a current flowing into the base and out of the emitter. This current is what the arrow on the NPN transistor's circuit symbol indicates.

It might sound like a contradiction that the electrons are flowing from the emitter to the base, but the charge is flowing from the base to the emitter. It takes a minute to wrap your head around, but charges in a circuit proceed in the opposite direction to the electron movement. A charge is not the electrons moving through a circuit; it's the "holes" that those electrons are going to fill propagating. You'll also notice we're not all the way to ungating the circuit. We have electricity moving from base to emitter, but we also want it moving from collector to emitter.

This will happen naturally because the atoms in the p-type base will steal electrons from the n-type collector. This creates a "depletion region" of atoms with holes to fill on the edge of the collector.

The emitter is more liberally doped with n-type atoms than the collector, meaning that the emitter has more electrons to give away, and the collector is more prone to developing holes. With its heaving bag of electrons, the emitter can fill many of the holes in the base. The depletion region in the collector will then grab those electrons from the base. At the same time, a power source can pull electrons away from the side of the collector opposite the base, creating holes for the collector's electrons to fall into. The electrons leaving the collector then create holes deeper into the collector for more of the electrons from the base to fill. Now, we have electrons forcing their way into the emitter, across the base, through the collector, and out of the other side. Or, in terms of conventional current flow, there is a charge entering the collector, moving through the base, and then leaving the emitter. There is a consistent flow of electricity through the entire transistor, and the circuit has been ungated.^[1]

Let's go over the whole journey one more time. The n-type emitter is chock full of five-electron atoms that will form covalent bonds with the surrounding silicon to get nine atoms in their outer shell. This will leave each with an extra atom that it wants to push out to the surrounding silicon to fulfil the octet rule. A chain of electron pushes through the emitter eventually results in electrons gushing into the p-type base. The base is full of three-electron atoms, which will each bond with four silicon atoms to get seven outer electrons. The seven-electron atoms will then steal electrons from the surrounding lattice to achieve their eight valence electrons. The push of electrons from the emitter into the base and the absence of electrons directly outside of the base will then force electrons out of the base into the circuit. We now have quite a strong flow of electrons through the N and P of the NPN transistor, and those particles can force their way into the n-type collector, which will pass the electrons between its atoms until it can push them out into the holes of the circuit on the other side.

Notice that the circuit cannot be completed if there aren't atoms which need electrons outside the base. If the valence shells of the atoms outside the base are full, electrons would just pile up in the centre of the part and have nowhere to go, creating a traffic jam of negative particles. It's in this way that applying the positive charge to the base (making the holes below it) ungates the transistor.

From our comprehension of an NPN BJT, it's easy to learn how the PNP transistor works. The NPN's collector and emitter are n-type, but its base is p-type. For the PNP, it's reversed: the collector and emitter are p-type, and the base is n-type. Inverting the polarity of the parts inverts the current flow through it. Instead of flowing from collector to emitter, the electricity moves from emitter to collector. Instead of applying a positive charge to the base to open up the transistor, you must apply a negative charge, and instead of the charge starting by flowing from base to emitter, it flows from emitter to base.^[1]

Again, the direction of the charge is displayed on the circuit symbol for the transistor. Compare the NPN icon to the PNP:

ChipWizard, however, uses a fictionalised version of the PNP. Where the game's NPN blocks a charge from flowing through it unless you electrify the base, its PNP allows a charge to flow unless the base receives a signal. That is, the NPN transistor is closed by default, while the PNP is open by default. Not that fiction = bad and non-fiction = good, but if I'm educating on this topic, you should know that that's not how a PNP BJT acts. Definitionally, all transistors only allow a current to flow only once the base is electrified.^[2] From a design standpoint, ChipWizard can't model the real difference between an NPN and a PNP transistor because it doesn't implement positive and negative charges or directional current flow. To it, electricity is just electricity and fills up any space it has access to, like water filling pipes. This also means that even ChipWizard's NPN transistors don't have collectors or emitters; a charge will flow from emitter to collector as easily as it flows from collector to emitter.

Now we have a solid understanding of the transistor, we can see how we can implement them in conjunction to form logic gates. Working out how to make logic gates out of transistors is the challenge in the opening stretch of ChipWizard puzzles. Here are a few gates from the game:

ChipWizard also demonstrates more direct applications for transistors. In one stage, we can connect some transistors' collectors to a power source and its emitters to an alarm. Then, we connect the bases of these transistors to inputs which likely stand in for signals from sensors. When both of the probable sensor signals are "on", or "1", an electrical signal can flow to the alarm.

In a couple of levels, I implemented the AND-OR latch from the previous article: the circuit that lets us store or delete a single bit. In the stage shown below, the bit represents whether a motor is "on" or "off". When the "On" signal is electrified, a charge is sent to the motor until the "off" signal is electrified.

Examples like these show the gargantuan practical potential of the transistor. It's a utility born of its adaptability. Now, I could end the article here and leave you with a litany of misconceptions about electronics, but I don't want to do that. So, let's do some cleanup.

Firstly, we've been discussing how semiconductors work under the free electron theory, but once you get down to a small enough scale, that theory stops making correct predictions. Modern computers pack mucho transistors into very little space to maximise memory capacity and processing power. The modern smartphone transistor is just seven to ten nanometers across, and IBM has been able to fabricate a two-nanometer transistor. For comparison, a typical atom is 0.1 to 0.5 nanometers across. At that zoom level, only quantum mechanics can correctly describe interactions. But, dear god, we can't get any further into quantum physics here.

Secondly, transistors aren't the only non-human-operated switches in circuits. The most famous alternative to the transistor is the vacuum tube which was the standard before the widespread production of transistors. However, the microscopic size, miniscule power consumption, and high speeds of the transistor mean that they beat out opposing technologies.^[2]

Thirdly, while we use silicon as our medium in ChipWizard, there are other semiconductors. For example, some transistors are made of germanium, which also has four valence electrons, or the compound gallium arsenide. But silicon has monopolised the business because it's the second most abundant element in the Earth's crust after oxygen, making up more than a quarter of it. Therefore, it's cheap and not going to run out any time soon.

Fourthly, while we're strapping together BJTs in ChipWizard, 99.9% of transistors today are MOSFETs or Metal Oxide Semiconductor Field Effect Transistors. As transistors, MOSFETs are still gates in circuits controlled by an electronic signal, but we use different terminology for their parts, and they have a distinct construction from BJTs. There are many types of MOSFETs, and these aren't the only types of transistors either.

Fifthly, don't get confused into thinking that n-type semiconductors are negatively charged, and p-type semiconductors are positively charged. There are shifting areas of the materials that will carry negative or positive charges when you run a current through them, but when taken as a whole, they all have the same amount of net positive and negative energy, making them electrically neutral.

Sixthly, transistors are not just used to gate sections of circuits; they can also serve as amplifiers or control the amount of power running through part of a circuit.^[2] Then there are one or two details that the game is straight-up taking artistic license with. We've discussed that there is no direction of current flow or polarity of charges in the game. We also never have to worry about resistors or electromagnetic interference, or many other challenges a real engineer faces.

Despite its pixelation of electronics, ChipWizard Professional captures the instrumentality of the transistor. Something I appreciate about Zachtronics' library is that it doesn't just consist of detailed descriptions of computer science. The games that comprise it also fit together to form a more comprehensive image of a computer's workings than you could get from any one title. Puzzlers like Exapunks and TIS-100 teach you the rough logic of programming, 20th Century Food Court shows you how that code could be executed by a circuit, and then ChipWizard Professional demonstrates how you might build some of the parts in that circuit. Rather than choosing depth or breadth, Zachtronics games incorporate both, taking you from understanding components in a system to understanding the system as a whole. Thanks for reading.

Notes

1. How a transistor works by Ben Eater (Mar 20, 2015), YouTube.
2. Definition: Transistor by Rahul Awati (June 2022), TechTarget: What Is?
3. 2.5: Arrangement of Electron (Shell Model) by John Hutchinson (Jun 5, 2019), LibreTexts.
4. Electricity Explained: The Science of Electricity by US Energy Information Administration Staff (December 17, 2021), US Energy Information Administration.
5. Ions of silicon will have more or less than four electrons in their valence shell, but we're talking about non-ionised silicon here.
6. How semiconductors work by Ben Eater (Mar 19, 2015), YouTube.
7. The Doping of Semiconductors by Carl R. Nave (2017), HyperPhysics.

8. Image adapted from diagrams by Michel Bakni, licensed under Creative Commons Attribution-ShareAlike 4.0 International.

All other sources linked at relevant points in article.

Note: The following article contains moderate spoilers for 20th Century Food Court and SHENZHEN I/O.

Computers took over the world because they are the polymorphic machine. Other appliances are built to carry out a limited number of functions, usually one. They separate cotton from its seeds, they wash and dry clothes, they ventilate a building, etc. A computer is a calculator and a typewriter and a simulator for the stressors on a plane and a regulator for an industrial mixer and a camera and a hundred other things. While upgrading machines to perform new jobs is often a lengthy and involved process, many computers can be taught to take on new tasks in a few minutes through software installation.

The computer's adaptability is also why so much of computer science and programming is highly abstract. There are task-specific development tools for computers, but you'll be hard-pressed to find a programming language just for writing word processors or just for simulating plane physics. They're more broadly applicable than that, largely because software tools can operate on a wide range of data. Take conditional statements: operations that tell you if some variable meets some criteria. You could use them to program a microwave so that its alarm goes off if, and only if, the timer on it has reached "0". Or, there are loops, which run the same set of code repeatedly, usually until some condition is or is not met. You may notice loops implement conditionals and build on top of them. Imagine you're playing an MMORPG in which you cast a spell that buffs your party. The software will use a loop to cycle through every adventurer in your group and run the code to apply the boost to each one. The loop terminates when it has passed every party member. These operations don't care about the purpose of the data they're operating on, which is why they're flexible.

On the hardware layer, consider that you can't crack open a processor and find the part of the computer that specifically calculates wind speeds or the memory address that will always hold the first character of your document. Like the building blocks of software, the constituents of hardware are more use-agnostic than that. Take, for example, logic gates. Logic gates are elements of a circuit that take some signal/s as input, perform an operation on that data, and output it. If this sounds like a foreign concept, remember that addition and subtraction are also operations that digest some input and spit out a result. When I say "signal" in this context, I'm just talking about whether a wire or copper trace has electricity running through it; whether it's charged or not charged.^[1] Typically, we take an electrical charge to represent a "1" or "on" state and no charge to mean a "0" or "off" state.

The simplest logic gate is the NOT gate which gives an output opposite to its input. If its input wire is on, its output wire will be off. If its input is off, its output will be on. Another common player is the AND gate, which outputs an "on" or "1" signal only if all its inputs are "on" or "1". Otherwise, it outputs a "0". Here is a table showing what output the AND gate will produce based on what data is fed in:

For one last example, let's examine the OR gate. An OR gate outputs a 1 if any of its inputs are 1. It will only output a 0 if all its inputs are 0. Here is the OR "truth table":^[1]

These logic gates might look so abstract that you can't imagine how they could perform constructive operations but consider this AND-OR latch. The AND-OR latch is a collection of logic gates that lets you store a 1 or 0 and read that value back. It looks like this:

The circuit will default to storing a 0, but sending a charge through its upper input wire will set it to a 1. Sending a charge through the lower input wire will change it back to a 0. The stored bit is sent out of the output wire. The circuit below is a "full adder": an assemblage of logic gates that can add up numbers:

Understanding these schematics in full is not important. What is important is internalising that elements that can do many different jobs are integral to computers. The machine's output must change depending on the software and data fed to it. Therefore, it's full of parts, both soft and hard, that change their output based on the input. Many of Zachtronics' video games are about engineering machines out of instructions or parts that are ambivalent to their functions, and so, can be applied to a wide range of tasks.

The commands to route particles in SpaceChem don't care about the kind of particle they're routing, and the mechanical arms in Opus Magnum can pick and place any atom. In SHENZHEN I/O, the studio even has you build and program single circuits that must be able to take on a variety of different tasks, embodying the shapeshifting nature of real computers. One stage has you develop a smart window that must open or close based on pollution levels. Another challenges you to build a vending machine that calculates and dispenses correct change, even when the inputs and item prices are in constant flux. In 20th Century Food Court, one character in Zachtronics' Last Call BBS anthology, all the machines we build must be similarly adaptable. Both through the circuitry components that Food Court includes and the ones it's missing, the game shows why the parts we've discussed, like conditionals and logic gates, are crucial to adaptive machines.

20th Century Food Court is a game in which we build hardware that takes food orders as inputs and produces a tray matching that order as output. If you've played it before, feel free to skip the remainder of this paragraph and the one after it. During puzzles, Food Court's screen is spliced in two. On the right side, we can place food dispensers, conveyor belts, sensors, and other items, but the logic to run this circuit, we dictate via the gubbins on the left side of the screen. On the left, you'll find racks where you can install gizmos with input and output ports, and you can use those ports to wire parts into each other. "Out" connectors generally send a charge when some event in the circuit has occurred, while any charge coming into an "in" port will cause a part to perform a function.

For example, you might wire the "Start" output on your circuit to the "Ice" input for a freezer. This would make it so that when the machines start up, a signal will be sent to the freezer to make it dispense ice. Or, there are machines called Sorters which can push items that land on them in different directions. You could route the "Sense" output for a Sorter to its "Left" input, meaning that when it senses an item sitting on it, like a food tray, it will push it to the left.

A lot of us think of a circuit as a bunch of straight copper lines connecting parts on a flat wafer, but there's no reason the connections can't be slack or that you can't mount the widgets on a rack instead of a board. However, just as it can be difficult to see how you get from logic gates to a whole computer, much of your fight in 20th Century Food Court is straining to cobble together simplistic input/output devices in a configuration that would make something as complex as a chef. More than anything else, the complicating factor is that your mechanical cook needs to fill multiple possible orders without hardware modification. You won't be furrowing your brow figuring out how to hook the pretzel order output to the pretzel dispenser input. Yet, problems quickly arise when there are extra processing steps between reading input and sending output.

In the "Wine O'Clock" puzzle, we get a dispenser that can pour either red or white wine and must use it to fill a glass based on the customer's stated preference. The headscratcher is that you need two squirts from the wine container to fill the vessel, but conveyer belts move continuously outside of your control. You can't just sense when the glass falls under the tap and dispense wine then because you'll only have half a glass of alcohol heading to the exit slot.

A couple of levels involve constructing a burger. A tall hurdle in these stages is ensuring that all the ingredients get stacked onto the bottom bun before you place the top bun on because the top bun is effectively dispensed directly behind the bottom bun.

If both halves of the bun reach the Stacker in the same form they're dispensed, it will place the top bun on the bottom bun, and you'll end up with an empty sandwich. Part of your challenge is to route the top bun elsewhere and time the machine's operations so that the top arrives at the Stacker after all other ingredients. Yet, it's not easy to pre-program a time for the top bun to be delivered because, based on how many toppings the customer ordered, it may take more or less time to construct the rest of the burger.

There are obvious solutions. On the wine puzzle, you could send the glass to the correct dispenser and then drag it around a loop that whisks it back to the station for a second pour. For the burger trials, you could install bins to throw away redundant items. You wouldn't have to worry about holding back that top bun for the right moment if you routed it into the trash. When it's time to add the top bun, you dispense a whole bun again but drop the bottom half of that one into a bin. Yet, these answers are exorbitant and still don't tell our appliances when to dispense their foodstuffs. Ideally, this would be where conditionals step in to save the day. Time in 20th Century Food Court advances in steps. Imagine that we've built that wine vendor with the loop in it and know that the glass falls under the spout on step 5 and step 10. We could write code like:

"IF order = red wine AND (step = 5 OR step = 10), dispense red wine".

That would lend a specificity of operation to the wine machine that it lacks. If we could write a function that looped through every item on our burger and counted them, we could deduce the right time to place the top bun. But with no coding in Food Court, there are no conditional functions and no ANDs or ORs. We feel the absence of elements in proportion to their usefulness and the lack of these components in Food Court hurts. Yet, because computer hardware is made to execute software, it can perform the same functions as software. So, we know there must be physical parts that can do the equivalent work to conditionals or loops or boolean operators. Our reliance purely on physical circuitry lets the game emphasise the instrumentality of hardware like logic gates. The logic gates in Food Court are not called "logic gates", but they're in there.

We have the basic Multimixer, which takes up to four inputs and routes their signals into up to four outputs. In the Rosie's Doughnuts level, we must cover all chocolate or berry doughnuts in sprinkles, but not any plain doughnuts. So, we could set up a gate that either funnels the doughnut towards the sprinkle machine or doesn't. Then, we could connect the "chocolate" and "berry" outputs into a Multimixer and one of that mixer's outputs to the gate exit that moves in the direction of the sprinkler. Like so:

In other words, the Multimixer allows us to say, "IF the doughnut is chocolate OR berry, activate the part of the circuit that covers it in sprinkles". When we attach our interfaces with wires, we effectively build a conditional, and the base Multimixer serves as an OR gate.

There is a slightly more elaborate Multimixer that connects up to three inputs to up to three outputs. However, it only makes that connection if its "enable" port has an active signal entering it. This is useful in the Chaz Cheddar puzzle, in which we need to figure out whether or not to place meat on pizzas. We can create one circuit that detects when it's time to dispense the meat and another that saves whether the "meat" option was selected by the customer. We channel the part of the circuit that remembers whether we should dispense meat to one of the inputs of the Advanced Multimixer. Then, we connect the module that calculates when the pizza falls under the dispenser to the Multimixer's enable wire. So, if the part receives the "meat selected" signal AND the "meat time" signal, it will produce meat. The advanced Multimixer works as an AND gate.

There is no way to fabricate the sprinkle or meat system we just discussed without these gates, so you can see why they'd be essential to building computers. You might have heard that the invention of the transistor was a breakthrough in modern computing. If you've wondered why, consider that transistors let you build these logic gates on a miniature scale. When we route signals in 20th Century Food Court and SHENZHEN I/O, we also get a feel for how engineers use electrical charges to stand in for concepts like numbers, characters, or the need to pour cola. It's not that circuits consciously understand these concepts or that there is an innate signal that represents any one of these ideas. Electricity is always just electricity. It's that by making the signal pulse in different patterns or enter circuits with specific functions, it contextually works as the signal for "7" or "A" or "cola". There's no mechanical distinction between the electrons in the wire leading to the Freezer and the one leading to the Sorter; you assign the meaning to them through the fact that they lead to the Freezer or to the Sorter.

Yet, AND and OR are the only two gates Food Court gives you to define those purposes. Especially hampering is the lack of a NOT gate. In the level Sushi Yeah, we must serve every dish on the menu on a plate, apart from the soup, which we must serve in a bowl. If we could connect the soup input to the plate dispenser and put a NOT gate between them, the machine would place a plate on the belt whenever the order is NOT soup. Alas, we get no such luxury. Or there are Counters in the game that output a signal when they reach 0, but not on any other number. If we could combine Counters with NOT gates, they would be useful for, say, telling a machine to dispense 6 roast beef slices. You could have a Counter that's set to 6 if the order for 6 slices comes in. Every time a piece of beef is placed on the belt, you subtract 1 from the Counter. Then, you wire the Counter's output to a NOT gate and that gate to the dispenser. As long as the Counter's number is "NOT 0", it will keep sending the dispenser signal.

Three other logic gates: NAND, NOR, and ENOR, effectively incorporate the NOT gate, and so we're actually down four gates rather than one through its exclusion. It's possible to build the EOR gate in this game, out of AND and OR gates, but it's resource intensive and often too obscure to be useful.^[4]

There are workarounds to implement an analogue to NOT logic. For example, there is a more expensive Counter that outputs a signal when its variable is positive. Positive is almost "NOT 0". Or you can set up a bunch of OR gates to route any of the rice orders to the plate dispenser. However, these solutions consume more space and money while making your circuit harder to understand. Frequently, they also don't cover all the bases the NOT gate can. "Positive" is not the same thing as "NOT 0"; you've also got minus numbers.

Counters are your best friend as they're the only store you have for data entering the device. There's no RAM or on-chip memory, and you can't build latches. Memory storage circuits require gates beginning with "N". But all Counters start at 0, and you can only add to or subtract from that count using the numbers set via 2-4 pre-programmed inputs. As stated, you can also only get them to output a signal on 0 or any positive figure, so you can't freely write numbers into or read numbers out of these registers unless you want to work in binary.

You don't need me to explain to you why a computer requires memory, but again, the game makes the requirement abundantly clear by showing you what it's like to work without proper storage. As the only form of memory, Counters also serve as your timers. You can't count up to or down from a number if you have no way to remember what figure you're at. Therefore, you have minimal means to decide the "when" of an operation in the circuit without timers. In both Zachtronics titles and real hardware, the "when" is everything. This is one side of adaptivity in computers: circuits need to be able to bend not just to accommodate different inputs but also different timings with which those inputs arrive and outputs send.

A recurring theme in Zachtronics games is synchronising components: taking a screen of parts that don't innately know what the others are doing and getting them to cooperate on the same operation, doing the right job at the right time. However, 20th Century Food Court demands you exchange information between components with the correct timing to a greater extent than any other Zachtronics game.

Networking is important in SHENZHEN I/O. As the output of its circuits must change to match the input, some chips in its devices must inform other chips of the relevant data they'll need to produce that new output. For example, in the aquarium cleaner level, maybe you have one microcontroller that works out what operations the computer must do based on inputs and another that moves the cleaning head. That first chip would need to send data to the second chip. However, with SHENZHEN's modules acting as canvases for your code, there is a greater focus on setting the internal logic of parts over trading data between them. They can each remember and process a fair bit, and therefore, don't need to send data to other components to be stored and crunched. TIS-100 gives you fewer lines to write each screen of code with and forces you to pass information across its level's grids. The game thereby splits the difference between components processing data for themselves and exchanging it with their neighbours.

In 20th Century Food Court, your components must adapt to the changing inputs. However, you can set only a little of their internal logic at most, and many have only the practical power to relay signals to other components. With less processing capability per component, making a functional circuit means splitting all the processing between more parts. And that thin slicing of all the logic is what defines Food Court as a Zachtronics creation. If you play TIS or SHENZHEN, you'll discover that portioning out the program to different brains isn't just a case of performing the same logic in a few scattered text boxes. Instead, you need to design the hardware or software for the circuit's fundamental purpose and do the extra work of implementing an infrastructure for communication. For example, in SHENZHEN, you often need to come up with codes that chips can send to one another that represent modes they should operate in. Maybe for your aquarium robot, you use "0" to tell the chip to move the cleaner head left and "1" to move it right. Or you may have to ensure that one chip is listening for incoming data on the same tick another is transmitting it.

As Food Court is spreading the tasks thinner, it demands you implement more networking than ever. It's how it becomes a game about circuitry as much as about the components in the circuit. It's why you must tie cat's cradles of cables over the faces of your machines. It's the reason you're placing down plenty of Multimixers so that multiple components can speak to one particular input of another component. It's having one element reduce the value on a counter each time it does its job so that the timer can activate another element at the right moment. As Katie Salen and Eric Zimmerman observed, scaling up a problem often does not mean just scaling up a solution. Instead, the new solution must often include features the previous did not. In Food Court, it's normal to implement a feature, then add another feature to manage that first, and then perhaps add another feature to manage the management. But that also doesn't mean that you can expand your circuit as an alternative to engineering it efficiently.

As in all these system-builder games, you can throw more hardware at a problem instead of thinking through how to better optimise the hardware you already have. But implement enough components in Food Court, and you can confuse yourself with the new complexity of your machine. Even if you don't, if you're careless about what parts you install in your creation, you will run out of space. You might also create a circuit that works but is slow and costly, preventing you from competing seriously on the leaderboards. Limited resources and leaderboard metrics are another Zachtronics staple. In this sense, the studio's games don't act like puzzle titles as much as strategy games or management sims. In other puzzle games, you can often run up against hard limits that stop you from implementing a suboptimal solution. In Zachtronics titles, you usually can execute a flawed solution. Yet, you pay a resource or score cost for it, and when you have limited resources, it's often a lot of little problems that prevent you from implementing your solutions, not just one being flat-out "wrong".

Understanding everything going on in your circuit and being able to creatively realise different approaches is how you avoid these traps. Arguably, Food Court should be giving you more robust management tools, considering the number of parts you have to juggle and their abstraction. Opus Magnum, SpaceChem, ChipWizard Professional, and TIS levels don't contain too many actors to manage, and SHENZHEN lets you label elements in your circuit. By comparison, Food Court has shelves full of plastic and no labels to hand. There's no better way to relearn how your machine works than by turning it on and tracing the route that charges take through it.

But even if you could circle regions of the circuit and attach descriptors, it's unclear how you'd draw all the boundaries of these circles. The processing is crushed into a fine powder, and efficiency dictates that multiple components share the shame Multimixers, Counters, or other parts. Therefore, you can't have, say, a complete coffee machine module in one corner of the rack and a water boiler module in another. Instead, each module blends into one or more other modules, forming a logic soup. These gradients also make modifying a production line an involved process in which you have to be sharply conscious of its mechanics. Because each part is heavily reliant on those around it, when you take one piece out of the puzzle, you often need to change a lot about the neurons skirting it to compensate.

Again, recall that that network wouldn't be as complicated if you weren't building a machine to react to any possible order. The minimalist components and the spaghetti plate of wires that you hook between them are both the joints that let you flex into so many tasks and the complications that create hardship in solving problems. That's the life of the circuit engineer. But one of the great joys in these systems-building games is seeing some complex construct you've made acting under its own will. Because 20th Century Food Court has you making more system to do a job than any other Zachtronics title, there's more to be proud of at the end. There's nothing quite like seeing a screen of wires light up, and counters blink down to 0, somehow thinking without consciousness, rising to any challenge you can throw at them. They might even be able to get a burger out on time. Thanks for reading.

Notes

1. Boolean Logic & Logic Gates: Crash Course Computer Science #3 by Amy Ogan and Chris Harrison (March 8, 2017), YouTube.
2. Diagram by A-disciple, licensed under Creative Commons Attribution-ShareAlike 4.0.
3. Diagram by Trex4321.
4. Basic Gates and Functions by Wale Sangosanya, David Belton, and Richard Bigwood (1997–2005), University of Surrey.

All other sources linked at relevant points in the article.

Note: The following article spoils puzzle solutions for these games: Baba Is You, Braid, Broken Age, Cogs, Contradiction: Spot the Liar!, Death Squared, Deponia, Gabriel Knight 3, Genesis Noir, King's Quest, The Legend of Zelda: Breath of the Wild, Nancy Drew: Stay Tuned for Danger, Opus Magnum, A Plague Tale: Innocence, Portal 1 and 2, Q.U.B.E 2, Resident Evil 2 (2019), Slayaway Camp, The Turing Test, Unravel, and The Witness.

Video games still have a reputation as unyielding assaults of action and violence. Yet, as people who play games, we know that even mainstream titles aren't all kicking down doors and mulching enemies with chainguns. They often involve a lot of sitting around and thinking because puzzles are a mainstay of the format. You can't "get" video games without getting puzzles. But like so many other elements of the medium, you can read endlessly about puzzles without understanding how they tick. So, what makes a satisfying puzzle? For that matter, what makes a puzzle?

Defining Puzzles

Like "art" or "game", "puzzle" has a lot of definitions in circulation with no method for determining a definitive one. However, we can find a definition that works for our purposes by looking for commonalities between the things we refer to as "puzzles". I'd argue that most players identify puzzles as challenges that significantly test their reasoning. This would be true of a gameplay section in which the player is working out how to escape their home on a space station, how to transit an orb to a receptacle, or which of a subject's statements contradict each other. It's the working out that makes a puzzle a puzzle.

The Skills Puzzles Test

In puzzles, players employ an IF-THEN mode of thinking. "If I have a screwdriver, then I can unscrew the grate", "If I freeze the orb in time, then I can store up momentum in it", "If the interviewee was at their parent's home at the time in question, then they can't also have been at the market". This approach to problems employs inductive reasoning, in which we generalise from our previous experiences. For example, if we saw a mirror deflect a beam of light at a 45-degree angle before, we can safely assume another mirror will do the same. We also use deductive reasoning, in which we apply general knowledge to specific challenges. For example, if you know how a mirror deflects light and where each mirror in the room is situated, you can work out how to route the light from the source to the receiver.

This is not to say that puzzles only involve logical thinking and speculation. Many also test observation. Formulating an exit plan from the space station in Broken Age requires working out that you could use your screwdriver to open the vent. Yet, to do that, you must spot the vent in the first place. A Plague Tale: Innocence has many puzzles that task you with hitting a ring to loosen a chain, but that ring's location isn't always obvious.

Puzzles also often test your powers of memorisation and recall. A head for saving and retrieving information can help you when patterns from one puzzle game appear in another. There are popular tropes of the genre, like the weighted seesaw conundrum or the panel on which each switch inverts some combination of the other switches. Exact designs differ, but all puzzles of these types implement some of the same logic. Even a sense of when and how puzzle games will try to trick you, regardless of mechanics, can be learned. Specific games will also teach you the functions of mechanical entities, as well as techniques that will be indispensable during your quest, which you must mentally save and load. Rewards in puzzle games don't have to be resources or aesthetic feedback; they can be knowledge to use going ahead. Many puzzles are interlinked and cannot be randomly reordered within their game; they are scenes in a directed journey of extracting and applying information.

In the time-travel game Braid, certain items will remain in your hand even when you reverse time, and certain pits are too deep to climb back out of. In an early level, you learn that if you jump into a chasm with a valuable in it, collect the item, and then reverse time, you can be standing back where you started with the object you need. This is of use in future stages. In Q.U.B.E. 2, green panels let you spawn moveable cubes, and orange panels let you extrude platforms from walls. An essential technique in the game is spawning a green block in front of one of those orange panels, and then pulling the orange panel out of the wall to push the green block across the floor. You cannot place these tutorial levels from Braid or Q.U.B.E. 2 after those in which we apply their techniques. Alone, these dynamics from Braid and Q.U.B.E. might sound trivial to record and remember, but when the player is constantly juggling concepts, and under cognitive strain, memorisation and recall take effort. Some puzzles are also large enough that you can't keep all of them on screen at once. So, you must remember at least some of what's outside the frame.

Reasoning, observation, memorisation, and recall are easier to apply when we aren't distracted or pressured by threats. Therefore, the puzzle games that really want to grill our logic and imagination mostly don't involve enemies that can chase us down or a target to race against. They often have no explicit fail state, and in the case that they do, won't set us back far upon a loss. No punishment or minimal punishments for failure also mean that we have less trouble with our work being spontaneously erased, which is always annoying. Re-doing a hand-eye coordination challenge after we've just done it can still be taxing; every headshot requires accuracy, every jump requires timing. With a puzzle, once you know the solution, the challenge is nullified, so immediately repeating a puzzle, or part of one, feels like busy work.

Conversely, games that want to integrate time limits or combat into puzzles can survive by not placing too much strain on the player's reasoning, observation, etc. Working out what blocks to swap in a match-three grid or the procedure for a puzzle boss in an action-adventure game would be straightforward if you didn't also need to dodge attacks or fret about the clock. However, with those constraints involved, a little cognitive challenge can go a long way.

Good Puzzle Mechanics

So, what constitutes a successful puzzle mechanic? Why do some ideas fade into the wallpaper while others stick in your head? In short, objects and abilities in puzzle games have a high potential when they can interact with a wide range of other objects and abilities in many different fashions. Having mechanics that can synergise with others in countless distinct ways makes a game's design elegant and uncluttered while freeing the designer up to create a wide diversity of experiences. This is true of game mechanics in general. But there are also reasons to make highly adaptable play elements that pertain specifically to the puzzle genre.

For one, it adds to the mystery of the puzzle. If a cube's only use is to weigh down pressure plates, then we can perform dumb pattern recognition every time we encounter one. If you see a block, all you have to do is find the switch it goes on; a rote kind of square peg goes in the square hole logic. However, if a cube can depress a plate, deflect a ball of plasma, block oncoming fire, knock down a turret, act as a platform, or move through a portal, we can't assume the function of any one cube. We must work out the purpose of each cube on a case-by-case basis. To identify this principle in a different genre, note that in many block puzzle games, any tile can stack on or against any other. Therefore, there are numerous shapes you can build out the board in, some useful, some not so useful. Not every object needs dizzying depth, but memorable mechanics and ones with longevity usually have it.

Another good reason to make entities in puzzles flexible is to reduce the amount of tutorialising the game has to do. A lot of games want us to be clear on the functions of the verbs we're using. But if an object or ability only has a few purposes, the designer will teach us how to use it and quickly exhaust all its possible interactions with other elements. Then, they will have to introduce another in its place, again doing the necessary bootstrapping. That object will also get burned up fast, and then the game must do more education as it introduces another. The player is spending more time in Mechanics 101 and less time in advanced applications of those mechanics. The items and powers in puzzle games are more liable to this kind of exhaustion than those in action games because, again, game feel and hand-eye coordination have longevity that logic problems don't. The items and powers in action games depend more on the former concepts, while those in puzzle games are weighted towards the latter.

Finally, a puzzle game could want components with high interactive potential because we would be able to make many revelatory discoveries about those components. Those discoveries are what we love about plenty of puzzles. Slayaway Camp is a sliding maze game in which you can only move in straight lines and only stop once you hit an object or character. The goal in each level is to collide with every one of your "victims", killing them, and removing them from the map, before escaping to the exit tile. Stage 5 teaches us that when we murder an NPC standing adjacent to another, that witness NPC will run away from the victim, in the direction they are facing, until they hit something solid.

That rule means that you can force enemies to flee to a certain square and then use them as a buffer to align yourself with places in the level you might otherwise be unable to reach. You can use them to build out the maze. This technique wouldn't be possible if the NPCs only interacted with you and not with each other. Plus, you wouldn't get that little moment of confusion upon seeing you have no exit and feeling like a genius when you figure out how to open one up.

In the puzzle game Death Squared, you control both a red and blue robot and must move them to their respective goals. Level 16 of the game is a multi-storey environment in which the blue bot can stand on red blocks, and the red robot can stand on blue blocks, but not vice-versa. Despite the fact that the red bot falls through the red blocks, you must get it to the other side of a gap with red blocks in it.

Each bot can also stand on a button of their colour to activate an elevator elsewhere in the level.

Blue starts the level in a basement area, cut off from red, but by getting red to stand on a button, we can lift blue up to be just below red. We then get blue to stand on the red bridge and use them to ferry red across the chasm.

This puzzle is only possible if the bots work in conjunction with the buttons and lifts, can stack atop each other comfortably, and if they have unique relationships with the coloured blocks. If you take away any of those possible interactions, the puzzle, and the pleasant moments of realisation are lost.

The lack of depth in traditional point-and-click puzzle games is likely one reason that they were overtaken by other types of puzzle experiences. If there's only one use for the developer fluid or the yellow petal, they tend to feel flatter and less fully realised than, say, the ability to search words in a database or the power to copy yourself, which can find many more applications. There is one other reason that you might want mechanics that have lots of points of contact with each other, but that's a topic for later.

Good Puzzle Design

In addition to mechanics offering a wide variety of options, the levels or puzzles themselves must also do, for the same reasons: you want player experience to be varied, to make solutions non-obvious so that players have to think hard, to avoid having to introduce new concepts constantly, and to facilitate pleasant realisations. Mechanical design and level design are inseparable. A caveat, however: "Aha" moments are generally less common in games that place a stronger emphasis on action than working out. Revelations are also rarer in casual puzzle games, which are meant to be more accessible and relaxing than they are confrontational.

As in all games, the quantity of dynamics between elements must be balanced against limitations in what the player can do. Audiences and critics often see it as a virtue when an experience lets the player find multiple solutions to the same problem. When games have alternative routes through a level, allow us to select characters with distinct abilities, or let us choose whether we hack, stealth, or shoot our way around a problem, we tend to say that's inherently good. More valid options heighten player expression and accommodate players with different abilities. They also strengthen the sense of agency, the capacity for experimentation, replayability, and potentially, realism.

Puzzles break this common design wisdom as they frequently only allow one correct solution to a problem while remaining gratifying for audiences. There is a single path through most of the panels in The Witness. There's only one "win" state for the mechanisms in the original Myst. In Critical Mass, you can't insert new blocks anywhere in the grid; you can only stack them atop the current layers. By boxing us in, puzzles force us to focus on a goal.

Cogs is a collection of sliding tile puzzles, and some of those tiles have gears on them. The objective is to move the gears into the proper position to transfer the rotational force from a cog on one side of the board to a cog on the other. If we could drop these components in many different positions and still transfer that rotation, we wouldn't have to think as acutely when we place them. It's because we have to build a specific machine that we consider every cog's role in the overall mechanism. Or take the riddles in Nancy Drew: Stay Tuned for Danger. We must input our answers to these riddles as text, and every one of them is a closed question: a query with one correct answer. When the clerk says, "The more you make, the more you leave behind", we can only answer "Footsteps". When she asks, "What is full of holes but holds water?" she will only accept the response "Sponge". If these were questions with many correct answers like "Name a US president" or "What's your favourite flower?", you'd need to think about it a bit, but there are so many valid responses you could afford to be lazy. When the game is less flexible towards you, you must bend further towards it to meet your target.

This isn't to say that puzzles can't account for different skill levels between audiences or produce new experiences for returning players. However, the puzzle genre leans into it less than others. When puzzle games do cater to these interests, they are also less likely than other games to do it by offering a platter of diverse solutions to challenges. Although, there are exceptions in block puzzle games or systems construction games like Mini Metro, Bridge Constructor, or the Zachtronics library. As discussed, block puzzle games can get by making you think fast instead of hard, and systems construction games have characteristics that only some of their peers want to replicate. For example, Mini Metro doesn't have an explicit success condition which most puzzles are meant to have. However, just because you can't find many correct answers for one puzzle also does not mean that there aren't a lot of approaches available. On the contrary, good puzzles are often baffling because you could try many different actions, even if only one or two will work.

Categorising Puzzles

Drilling down a little further, we can divide puzzles (or at least sections of puzzles) into two categories. Our first category is puzzles in which we don't understand the relationship between the involved entities, and the challenge is in figuring it out. We'll call these Relationship Puzzles. Relationship Puzzles are common in adventure games and in tutorial areas of other puzzler types. Examples include deducing that you can foil the fire-breathing dragon by throwing a bucket of water at it or working out, unprompted, that the lines you draw on panels must always separate black and white dots.

The other category is puzzles in which we understand the relationships between the gameplay entities but are unsure about the configuration they should exist it. We'll call these Configuration Puzzles. This one takes a minute to wrap your head around because you'd think that if we know the rules a system abides by, it should be a breeze to arrange the parts of that system to meet any goal. However, that's not necessarily true. In Sokoban games, you can use your avatar to push a crate one square in any direction unless a push would cause it to intersect with another crate. By moving crates, you must clear yourself a route to the exit. The rules are incredibly minimalist, but working from those rules to a solution can still be a nightmare. Or there's a puzzle in Professor Layton and the Unwound Future in which we must find the working pen in a set of four. We are given the rules to select the correct pen, e.g. "All pens currently have the wrong colour caps" or "The working pen is to the left of the one that should have the green cap.". However, we have to do some hard graft to get from abstract guides to a specific idea of which pen is the answer.

It is also possible for a single puzzle to be a Relationship Puzzle and a Configuration Puzzle. For example, in the Resident Evil 2 remake, there's a puzzle in which you must transfer liquids between vials and fill one beaker to a target level. The catch is that when you pour the chemical from one container to another, you must move all of it into that second container, and not just some. Completing this puzzle involves figuring out how the vial machine operates and divining your objective from the clues, but also concluding what vials you must switch liquids between in what order. Genesis Noir contains a section in which you're destined to make a breakthrough advancement using a particle accelerator. You do this by turning knobs on various pieces of equipment to change a monitor's graphical output, trying to recreate an image drawn on a blackboard. However, you aren't taught that's the objective or how the controls affect the output.

Good Puzzle Solutions

Whatever kind of puzzle the designer is producing, they need to come up with a satisfying solution for it, not an arbitrary one. For the remainder of the article, we're going to put block and casual puzzle games back in the drawer and keep our mind on puzzles in which designers are trying to generate "aha" moments. Broadly speaking, players enjoy it when solutions follow rules that they can reasonably identify. In Configuration Puzzles, objects and powers should generally function as they have previously. What acts as a rope in one level should generally not become a solid wall in the next. In Relationship Puzzles, you should be able to intuit the function of objects and powers reasonably. "The numbers on this notepad are codes to a safe" qualifies, "I have to spray a cat with a hose to make a fake moustache" does not.

Intrinsically rewarding puzzles also usually force us to change our perspective on something to reach the correct conclusion. Maybe that chewing gum that you previously saw as a piece of candy, you now consider as an adhesive, or you might realise that the objective of a section is not to carry the item across the stream. It's to place the item on a platform in the middle of the stream, swim across, and retrieve it from the other side. Many of the well-received puzzles in contemporary games also lure you into an incorrect solution by appealing to your current perspective. This creates the starkest contrast between your "Before" and "After" viewpoint on a puzzle or mechanic.

For example, in Coldwood's Unravel, you play a character made of yarn who leaves a trail of string behind them when they walk. You can grapple objects towards you with this thread. In the game's Sea stage, you pass below a couple of beams and come upon a grappleable hatch hinged at the top; the goal is to pass through this hatch. The intuitive solution is to grab the door, pull it open, and then run through, but if you're the one holding the door open, moving towards it will close it.

The solution comes when you realise that you can wrap your yarn around objects in the environment to change which direction the thread pulls in. If you tie your string onto the door, then pass up over that log at the top and run back towards the door, you create a pulley system which can open the hatch as you proceed in its direction.

In Bulkhead's The Turing Test, we must power locks to make progress. We can do this either by inserting a power cube into them or firing power orbs into them. Our gun allows us to sap or insert power orbs at range. Level 1-7 runs in a loop and starts us on the ground floor in front of a locked door. Near this locked door is a power cube.

On the other side of the locked door is another door kept open by a power orb, then a staircase up to the exit, which has another lock.

From this raised platform by the exit, you can also see the starting door and its lock.

So, there are three locks but only two keys. How do we get all three doors open? The revelation comes in realising that cubes and orbs aren't equivalent. That orb can be reused and so isn't "expended" when inserted into a door. The solution is to insert the cube into the first door, collect the power orb from behind it, and then switch the orb for the cube in the first lock.

Then, you place the cube in the second lock and run up the stairs.

Now, you can use your gun to collect the orb from the first lock and shoot it into the third lock, opening the exit.

Returning to Q.U.B.E. 2, by the late game, you've learned not only that trick to push cubes we discussed earlier. You've also taken on board that you can light green blocks on fire and use orange blocks to slide them across the floor. You can even transfer fire from one item to another. In one puzzle, you have a green block generator and an orange block generator facing a door. You also have a means to set a block alight, but each block can only burn for a limited time. Your objective is to crash a burning block into the door.

The problem is, if you set that green cube alight and then push it towards the door, the fire goes out before it reaches the door.

The false perception we must let go of to succeed is that we need one cube to make it all the way across the floor. We actually need to spawn a cube and push it halfway. Then, we can spawn a second cube, light that one, and push it towards the first cube. That second cube will transfer its flame to the first cube, as well as its momentum, and the first cube will be able to make it to the door, aflame.

I've simplified this solution a little for clarity, but this is the gist.

How Puzzles Mislead

We know that these games facilitate ample perspective shifts by letting the involved items interact in many different ways, but how do games bait us into false solutions? A ubiquitous technique is to make wrong solutions the easiest to implement, and the right one take more steps. In Q.U.B.E. 2, pushing one block towards the door was easier than orchestrating this more complex mechanism for transferring fire from one cube to another. In The Turing Test, the simplest approach was to open the first lock, and then forget about it and just worry about the last two, but a more complex conception of how the three locks are connected is required to resolve the puzzle.

A closely related concept is that players often find it unintuitive to move backwards to advance; it feels like the opposite of progress. Designers can exploit that. In The Turing Test, we see that concept employed as the designer requires the player to re-lock some doors before they can unlock them again. In Unravel, the player had to move away from the door before they could approach it again.

You can also mislead the player through visual and auditory presentation. In Unravel, the door was right in front of Yarny, while the beam you needed to use was out of their direct line of sight. In The Turing Test, after you pass through the first two doors, the orb you need to unlock the first and third door will then be out of sight, and potentially, out of mind. Players also come to games with mental shortcuts they've adopted from outside, like "You can't move a key at range" or "If you want to get to a goal, you should move towards it". These puzzles exploit those assumptions.

A designer can also have you use the same mechanic in the same manner up to a certain point, so you, often unconsciously, assume that mechanic must always be used that way. You'd previously used one burning cube for each fire door, so you might not speculate that there's another method. You'd not encountered puzzles before where you could use the same orb to open doors on different levels, so why would it be possible this time? Devious puzzle designers can even create a puzzle with a solution the player has already used but still have it stump them. They often give us time to forget approaches we've previously employed, make us use different approaches before returning to the former technique, or create a new spatial or narrative context around the same problem.

Zach Barth's alchemical puzzle game, Opus Magnum, has a habit of introducing new components only to quickly swipe them off the table. Because we have a lot of intervening puzzles in between first transmuting lead into gold and doing it on another occasion, we can forget something simple like the need for salt to catalyse the reaction. Arvi Teikari's Baba Is You is a puzzle game in which you can push together blocks with different words on them to change the rules of the game. For example, we could create the sentence "Rock is Push" to let us push rocks around or "Flag is Win" so that we win if we reach the flag. Touching the flag is the typical win condition, but in Stage 1-5, we create the phrase "Baba is Win" to skip the middle man and complete the level without reaching the goal. In Stage 3-9, we are meant to implement the same technique, creating the phrases "Baba is Rock" and "Rock is Win" to make ourselves the goal object. It's the same idea, but because there have been two intervening worlds, flipping through a small Rolodex of mechanics, it's difficult to recognise the solution, at first. Further confusion may be created by the reframing of the problem, having us write "Rock Is Win" instead of "Baba Is Win", even if we are the rock.

This is the secret final reason that it's useful for gameplay entities to be able to interact with many others in many ways. It allows for a barrage of different gameplay dynamics between a concept making its debut and being used again, allowing the player time to forget an item or ability's earlier functions.

Conclusion

Keep in mind this is just a cursory overview of puzzles, and there's much more to say on the subject. When you break them down, the solutions to the puzzles I mentioned in Unravel, The Turing Test, and Q.U.B.E. 2 are simple. Still, any substantial puzzle can be split into smaller parts, and the player will segment them when solving. A puzzle probably won't be challenging if it relies on the player just taking many steps and if those steps are similar to the ones they've made before. What makes a puzzle demanding is when they force us to change their methodology for each step substantially. Thus, we shift our perspective on the world just that little bit. Thanks for reading.

Note: The following article makes brief mentions of sexual assault. It also contains moderate spoilers for Neo Cab.

For a long time, the automated car has felt so close to becoming a reality, yet so far. On the one hand, we're flush with personal AI assistants and image recognition algorithms. Our factories have been automated for decades, and our phones can use adaptive computer code to erase objects from photos. With these advances, it feels like the driverless automobile is running late. Nonetheless, it's not arrived, and we've been told it's "Just around the corner" since at least 2015, which has cemented the technology as the quintessential example of Silicon Valley overpromising. So, what will it take to transport us to the world of AI taxi drivers? And what might life look like for the everyday worker if we get there? Chance Agency's 2019 game, Neo Cab, has thorough answers.

Neo Cab's protagonist, Lina, is caught in the crossfire of the AI revolution. She is a former employee of the shrewd megacorp Capra, which, in a previous life, operated as an Uberish ride-hailing service. The second Capra rolled out its automated taxis, it spontaneously fired all its drivers. Lina now eeks out a living in the near-future city of Los Ojos, driving for Neo Cab. Her new employer is a lesser economic competitor offering the boutique experience of a taxi with a flesh and blood operator inside. Even as we touch down on Neo Cab's backstory, the game is challenging our preconceptions about automation.

Common sense dictates that if we want to stem the tide of automation, then we need more people in jobs. Yet, some companies use human labour to accrue the capital to implement a computerised workforce. This creates a paradox in which getting more people into certain jobs can mean their work will be taken by AI in the long term. In Los Ojos, that bait-and-switch occurred under Capra; in our world, Uber and Amazon look eager to do something similar. Still, it's obviously not a shortage of cash keeping tech empires from rolling out the driverless car. So, let's talk about the real barriers.

One issue is that even if you invent the self-driving car, or whatever other technology, that doesn't necessarily mean it will be ubiquitous. A lot of our tech now is controlled by relatively few puppeteers due to being tied up in intellectual property law. In Neo Cab, Lina's job exists because there is one company that was able to patent a driverless car system, and every other taxi firm has to try and drive without it. Although, that is one more company than has developed a reliable self-driving car in our time.

As Neo Cab teaches us, we aren't being shunted from a manually-operated society to an automated one overnight. Some tasks it's more straightforward for a computer to perform than others. The game doesn't expound on what makes a task "easier" or "harder" for an AI, but the general trend is that automation does well in predictable environments that aren't going to throw a lot of curveballs at it. Almost all of those environments are tightly controlled for the sake of making them fit the narrow-minded AI. Think factory floors, video game rulesets, etc. But in messy, organic systems where the rules and contexts are constantly changing, computers flounder. Much of the driving done in the world happens in those chaotic contexts.

An autonomous car in flat lighting on a well-maintained road with no weather visibility problems has nothing to fear. But if it's snowing, street signs are damaged, pedestrians appear in dappled lighting, the vehicle has to negotiate with other motorists, or if you run into any number of edge cases, a self-driving car's senses can fail it. Advocates for driverless cars are quick to say that only the outliers trip up this tech, but if you can't account for outliers, you can't account for real streets.

Getting a bundle of electronics to understand human dialogue and respond appropriately is probably harder. Sentence construction is devilishly complicated, with words morphing their meaning as they are placed within different grammatical and semantic holders. Cues for tone and meaning are subtle and inconsistent, as is body language. And you have to pull from complexly interlocking knowledge bases every time you form an opinion and express what you think. The current work in conversational AI is at once breathtaking and far short of the real thing. What's more, the appropriate margin for error varies between tasks. If a computer program can't identify a flower species from a photograph, that's not the end of the world. If it can't work out which shape in the road is a child, you're in big trouble.

AI's hike to absolute competence is slow, and the most obvious conclusion to draw from that is that some jobs are getting automated before others. In Neo Cab, "taxi driver" is an occupation that digital minds can slip into easily but something like "quantum statistician" is beyond their capabilities. Less immediately evident is that when some tasks can be automated, the pressure on workers to perform non-automatable tasks, and the standard at which they're expected to perform those tasks, dramatically increases.

In isolation, Lina's driving skills and car are no longer profitable assets. It's only the "authenticity" of manually driving a car and her social skills that she can sell to a consumer. That means that her job hinges more than ever on her ability to converse with customers in a manner they appreciate. Not every client cares about cross-seat patter; some are happy enough to lounge back and play on their phone, or can recognise Lina as a person just trying to get by. But from the rest, there is more push to conform to their idealised image of a big city cabbie. To some, that means a folksy streetwise caricature; to others, she's an object of fascination like a zoo animal; to others still, she's a punching back they can say and do whatever they want to. Neo Cab suggests that as people have become more used to interacting with computers to meet their everyday needs, their social skills have atrophied. They have forgotten how to speak to service workers, although there are some people who have never known.

And if Lina doesn't play into her customers' expectations, she might be left without an income and without a home. She exists in that precarity partly because of how the taxi industry was run even pre-Capra. Firms in the Uber model don't have formal hiring or firing processes, and while other jobs might review an employee once every few months or once a year, gig work employees are rated after every job. It is that much faster to eject them from their post, and individual customers have a direct position of power over them. Throwing employees overboard strips no skin off the nose of the platforms because of the simplicity of recruiting new workers and the abundance of their labour.

Furthermore, gig work employees are operating on razor-thin margins. Many gig work companies are effective at extracting value from their workers because they've figured out how to make them pay for their own tools. A traditional taxi driver didn't have to fork out for their fuel, but a ride-hail driver like Lina does. And with that drain on our income, missing one or two fees a night could be the difference between breaking even and operating at a loss. Then you have to factor in spontaneous interruptions like customers changing the pick-up point or police pulling you over.

As we juggle Lina's fuel and accommodation expenses, we experience that the problem for the gig worker is not just a consistently low standard of living. It's also that their financial conditions are subject to fluctuations. The number or length of rides we might be offered each night is fluid, and fuel prices yo-yo based on the markets or where in the city we find ourselves. Neo Cab could be harsher in forcing you to balance your books, but it effectively communicates the dreadful uncertainty of gig work. It also conveys that when you are making scraps, savings aren't just a safety net but also offer some freedom and dignity. When you're not starving for cash, you can afford to be pickier about which riders you pick up or how much disrespect you take from them. But even freedom is relative, and with the fickleness of the job, you must save some money in case of a sudden famine of work.

There's the sense in Neo Cab that Lina's occupation might not exist for much longer. Maybe, just for now, the human touch is enough to build a taxi business on, but it's a speciality interest, and all the little gaps in the market are getting filled in by the megacorps. The irony is that Lina can't sell many of her customers an authentic conversation anyway. It's not that you never say something true to a passenger or can't ever stand up for yourself when they're trying to push your boundaries. However, you'll find many prompts for which you know your earnest response might be disappointing for the rider, and you have to decide whether to zhoosh up the truth for a superior "customer experience". Some customers can't buy an authentic conversation because they're so rude that you'd never be speaking to them if they hadn't paid.

Most public-facing writing on driverless cars is either about its technological or economic aspects. There's not nearly as much ink spilt on the dimension Neo Cab is covering: the interpersonal impacts. Even when trying to understand how autonomous vehicles enter our world, we have to be conscious of the social component. This is not comfortable to think about, but you're probably never going to reach 0 deaths from driverless cars. Even if you coded the perfect motorist, pedestrians could put drivers in circumstances where they have no choice but to collide with someone. It's why researchers are asking questions like, "If it came to it, should a self-driving car hit three adults or one child?" or "Is it better to kill two people violating traffic laws or one innocent?".

Therefore, you won't get autonomous cars on the road just by raising capital and enhancing their brains. An essential step in the process is normalising deaths by driverless car. In our age, manually driven cars are associated with rigorous safety checks, while the autonomous car still feels fairly dangerous and liable to lose control. In Neo Cab's future, the autonomous vehicle has been improved to the point where the human-driven car now claims more lives. Accordingly, the stigmas associated with each mode of transportation have flipped.

Of course, just because Capra's driverless cars are safer than their non-autonomous competitors doesn't mean they're as safe as they could or should be. And what makes this world scarily believable is that, for many countries, it's not a stretch to think they'd accept hundreds or even thousands of deaths a year from self-driving cars because they already tolerate an equivalent quantity of deaths from other sources. Lack of medical care and regular car crashes are enormous killers worldwide. In the US specifically, thousands die of gun deaths every year. The line between the world now and a world which embraces the driverless car isn't the green light for random, inescapable violence. That already exists. What must happen is the normalisation of a new type of violence. Although, I do suspect that high-regulation, low-violent crimes regions, like Western Europe or Japan, would demand a high standard for autonomous automobile safety.

Given that over 46,000 people die from vehicular collisions every year in the US, a less dangerous mode of transportation could do monumental good. The citizens in Neo Cab are well aware of the benefits, with many taking to the picket line in support of a ban on manual cars. It's a ban that, predictably, has the backing of Capra. Capra churns out tearful media about the human cost of manually-driven automobiles and even fakes grassroots support for their corporation among pro-autonomous protestors. Yet, as a taxi driver ourself, we know what it would mean to lose our manual car. Neo Cab depicts a distasteful catch-22 that we may have to stomach one day. You can stick up against big business and keep the drivers on the economic ropes from getting fired, or you can reduce the number of road deaths. But you can't do both. Capitalism will have its blood one way or another.

While a fictional company, Capra's legal meddling has them following in the footsteps of Uber. Uber wrote the book for the Silicon Valley taxi industry, and it had to change the law to do it. When the company first mosied into cities, it was somewhere from a legally grey entity to an outright criminal organisation.^[1] To quote one Uber executive, "We're just fucking illegal". You had an unregulated taxi firm that did not extend employee rights to their drivers, and one of the reasons the field was regulated beforehand was for customer safety.^[2] States can identify licensed taxis in the case that the driver decides to kidnap or assault a passenger, but Uber didn't have to report to the state.

If you want to look into it, there was this wild period in which Uber tried to sabotage law enforcement's investigations into it with ghost cars and held secret behind-the-scenes meetings with world leaders.^[2] But there are two particular areas of this history that Neo Cab riffs on. Firstly, in Neo Cab, sometimes you have to pick between following the law and providing a worse taxi service or bending the rules and potentially getting ticketed. Ride-hail drivers face that exact dilemma today. But in addition, when Uber was still establishing itself, it was instructing its drivers to keep working even as the police were fining and towing drivers. Those workers also had to choose between what was profitable and what was legal.^[1]

Secondly, sometimes Lina's interactions during her shift threaten to spill over into violence. A few times, I wondered if saying the wrong thing to a customer might trigger a physical response from them. Lina is a woman of colour driving around a city alone at night. Uber was banned from London, England, for failing to report sexual assaults in their vehicles and for performing lax background checks, among other incompetencies. According to their records, in 2020 alone, there were 998 sexual assaults reported in Uber vehicles.

Other times, Lina and her customers are caught in crowds of activists that become so ferocious there's a possibility of someone getting hurt. Like Capra, Uber extorted their legal rights through exploiting tragedy and throwing their weight behind protests that endangered workers. While taxi drivers worldwide protested willfully in favour of their livelihood, Uber encouraged their employees to attend counterprotests. Assaults against their drivers became a recurring problem, with some even being murdered.^[3]

When then-CEO, Travis Kalanick, was warned that violent far-right agitators had entered the ranks of the anti-Uber protestors and that he should withdraw his own workers, he responded that this turn of events would be favourable for Uber. In his words, "violence guarantees success". Talking about the issue more generally, one manager commented, "We'll keep the violence narrative going for a few more days before offering a solution".^[3] The sordid details of Uber's rise mostly came to light after Neo Cab released. Still, in retrospect, the game underestimated how much of a moral void Silicon Valley is.^[2]

There are still a few starstruck techno-optimists who think we can automate our way out of crisis. Neo Cab is a dystopian rejoinder to that view. It depicts the driverless car as being ushered in by a new wave of corporate opinion-laundering and as displacing and increasing pressures on already desperate workers. Even then, it edits out the most disgusting tendencies of the industries it satirises. In a lot of ways, it feels like a mercy that the self-driving car is still science fiction. Thanks for reading.

Notes

1. Today in Focus: The Uber files: the unicorn (part 1) by Michael Safi, Johana Bhuiyan, and Paul Lewis (July 11, 2022), The Guardian.
2. Uber broke laws, duped police and secretly lobbied governments, leak reveals by Harry Davies, Simon Goodley, Felicity Lawrence, Paul Lewis, and Lisa O'Carroll (July 11, 2022), The Guardian.
3. 'Violence guarantees success': how Uber exploited taxi protests by Felicity Lawrence and Jon Henley (July 10, 2022), The Guardian.

All other sources linked at relevant points in the article.