Immersive Sims and the Cost of Letting Players Cheat

In Deus Ex you can stick a proximity mine to a wall. It sticks because the mine is an object with an attach behaviour and the wall is a surface, and neither of those facts was written with the other in mind. And once a mine is stuck to a wall, it’s a ledge. Stick another one above it and you have two ledges. Ion Storm shipped a game in 2000 in which the intended solution to a locked compound — find the keycode, hack the panel, locate the vent — could be replaced by climbing a staircase of live explosives over the fence.

They left it in. That’s the sentence I want you to sit with, because everything worth saying about this genre and everything that has gone wrong for it comes out of that decision.

What Ultima Underworld actually invented

Blue Sky Productions — shortly to be Looking Glass — shipped Ultima Underworld in 1992, and the thing people remember is the texture-mapped first-person view a year ahead of Doom’s corridors. That’s the least interesting part of it.

The interesting part is that the Stygian Abyss contained objects. You could pick a thing up, and the thing had a weight, and you could throw it, and the throw had a physics behaviour that didn’t care what you were throwing at. You could talk to a monster instead of hitting it. You could drop a lit torch and have it still be lit on the floor. None of this was in service of a puzzle. It was in service of the world being made of consistent parts.

That’s the doctrine, and Looking Glass spent a decade refining it: System Shock (1994), Thief (1998), System Shock 2 (1999) with Irrational. Doug Church wrote about intention as a design property — the player forms a plan, the game lets them attempt it, and the outcome follows from rules rather than from the designer’s approval. Warren Spector carried the ethos to Ion Storm Austin and built Deus Ex out of it: state a problem, hand the player a physics engine and a set of verbs, and let the systems adjudicate.

Thief is the purest small version. Sound propagates. Light is a number the game computes and shows you in a gem on your HUD. Guards have states, and they transition between them because of things you did rather than because a trigger volume fired. The immersive sim learns to listen, and everything after it is that idea with more verbs.

The GLOO cannon problem

Prey (Arkane Austin, 2017) gives you a gun that fires quick-setting foam. Its stated purpose is crowd control — glue the mimic to the floor and hit it with a wrench.

Its actual purpose, within about twenty minutes of anyone getting hold of it, is architecture. The foam sticks to any surface and it’s solid and you can stand on it. So you can build a staircase up a wall. Up the outside of a wall. Through a gap in the ceiling, into the maintenance space, over the top of a locked section, and out the far side of a chunk of Talos I that a level designer built, lit, populated and tested and which you have now declined to visit.

Arkane knew. They obviously knew — you cannot ship a foam gun in a 3D space and not know. And the studio built the best argument the genre has had in years on top of knowing, because the alternative was invisible walls, and invisible walls are the sound of a designer withdrawing the offer.

That’s the cheat in the title. It isn’t exploits or trainers. It’s the player using the systems against the intended difficulty, and the immersive sim’s whole moral position is that this is allowed — that if the physics say the foam holds your weight, then the foam holds your weight, and the locked door was only ever locked as a suggestion.

Every game in the canon has one. Deus Ex has the mines. Dishonored 2 has the Clockwork Mansion, a level that reconfigures itself at the pull of a lever and which you can get inside — into the machinery between the rooms — so Arkane’s showpiece set-piece can be completely bypassed through its own guts. They built the bypass on purpose. It’s the best level of the last decade partly because it argues with itself.

Cost one: you cannot test it

Here’s the first bill, and it’s arithmetic.

A scripted game has n scripted things and you test n things. A systemic game has n systems and the surface you’d have to test is every system against every other system in every location with every object, which is not a number QA can be resourced against. You don’t test an immersive sim. You sample it, you ship, and you find out.

Which is why every single one of these games is buggy. Not incidentally — structurally. Vampire: The Masquerade – Bloodlines (Troika, 2004) is the extreme case, a game so broken at launch that the community patch is now twenty years into development and is how anybody sane plays it. It’s a broken masterpiece and the two halves of that phrase have the same cause.

The genre’s bugs are also its best moments, which is the joke. The mine ladder is a bug. It’s also the most-told Deus Ex story in existence, and it’s why the game is remembered as the one that lets you break it. Nobody’s most-told story about a scripted shooter is a story at all; it’s a description of a cutscene.

Cost two: nobody sees the good bit

This is the one that kills the genre commercially, and it took me years to see it properly.

The immersive sim’s quality is conditional on the player’s curiosity. A player who walks up to the locked door in Prey, finds the keycode on a corpse, types it in and walks through has played a mediocre shooter with a slow start. Everything that makes Prey remarkable was available to them and required them to have the thought first. The game will never suggest it. Suggesting it is the one thing the design cannot do, because a game that tells you the foam is a staircase has converted the discovery into an instruction, and the discovery was the product.

So you cannot market it. You cannot put “you can climb the wall” in a trailer, because the value is in you working out that you could. The trailer shows a shooter. The reviews say “you can do anything”, which is what marketing says about everything and which no-one believes. And the audience that responds is the audience already fluent in the grammar — people who’ve played System Shock 2 and know to read the horror in the log files instead of waiting for the game to explain.

That audience is small, loyal, and about a tenth the size the budgets need.

Hitman is the one that solved it, and the solution is instructive: IO turned the sandbox into a puzzle box with the answers labelled. Mission Stories put the emergent solution on a checklist and walk you to it. It is, purists will tell you, a betrayal of the doctrine. It’s also the only systemic-freedom franchise of the last fifteen years that’s commercially healthy, and I don’t think those two sentences are unrelated.

The bill came in May

Microsoft closed Arkane Austin on 7 May this year, along with Tango Gameworks, Alpha Dog and Roundhouse. The studio that made Prey no longer exists.

The last thing it shipped was Redfall, in 2023, a co-op vampire shooter with service-game architecture and a reception I won’t relitigate here. What I’ll note is the shape: the studio that built the finest systemic space of its generation was put to work on a genre with a proven revenue model, the result landed badly, and the studio was closed. Looking Glass closed in May 2000, eleven days after — near enough — a similar set of arithmetic. Ion Storm Austin closed in 2005. Troika closed in 2005.

Four decades, and the genre has never once managed to be a business. It has managed to be an influence: the log file, the vent, the physics object left lying around for you to think of a use for — these are in everything now, sanded down and made safe.

The wall next to the door is still there. Somebody is still deciding, every time, whether to let you climb it, and the honest ones let you, and it costs them the level they built and the QA they can’t afford and the trailer they can’t cut. The mine ladder has been in Deus Ex for twenty-four years. It’s still the best argument anyone has made for any of this, and it’s still, technically, a mistake.