Contents

The Immersive Sim Canon

Ten games built on the same stubborn bet: that a consistent world beats a designed solution

Contents

There’s a moment every game in this canon eventually hands you. You’re looking at a door. Somewhere in the building is a keycard, and a designer put it there, and you know roughly what the intended sequence is. And then you notice the vent, or the crate you can stack, or the fact that the wall is a material the game has told you is destructible — and you realise the lock was a suggestion.

That moment is the entire genre. Everything below is an attempt to make it happen reliably.

The definition, briefly

Advertisement

The term comes out of Looking Glass in the nineties and has been argued over ever since, which is what happens when a genre is named by the people making it. The workable version: an immersive sim simulates its world with consistent rules, exposes those rules to you, and then declines to check whether you’re solving problems the way anyone anticipated. The keycard exists. The keycard is also optional.

That’s expensive. Every system you add multiplies against every other system, and the testing burden goes up faster than the content does. Which is why this canon is thirty years long and about a dozen games deep, and why half the studios in it are dead.

Ultima Underworld: The Stygian Abyss (Blue Sky Productions, 1992)

Paul Neurath’s team shipped a texture-mapped first-person dungeon with full freelook, sloped floors, physical objects and an NPC population you could talk to, and they did it months before Wolfenstein 3D — which most people remember as the year zero of first-person games.

Underworld’s world has volume. You can look up. Things fall. You can drop a lit torch and it stays lit on the floor and keeps casting light. In 1992 that was near-hallucinatory, and the studio that became Looking Glass spent the following eight years working out what you build on top of a world that behaves.

System Shock (Looking Glass, 1994)

Advertisement

Doug Church’s design, Warren Spector producing. Take Underworld’s simulated space, move it to a space station, remove every other living human, and narrate the whole thing through audio logs left by the dead.

The audio log is the invention people credit it with, and that’s underselling it. The real move is that Citadel Station is a place with a plan — SHODAN reconfigures it, the security level is a number you can lower by finding cameras and shooting them, and the game’s difficulty is something you negotiate with rather than something you’re assigned.

Thief: The Dark Project (Looking Glass, 1998)

The one that changed what the simulation is for. Thief simulates light and sound and then builds the whole game inside those two variables. You’re a burglar. Shadow is literally a resource displayed on your HUD, guards hear your footsteps differently on stone and carpet and moss, and combat is a failure state you survive rather than a system you master.

The design courage here is enormous: they built a first-person game in the Quake era and made the correct verb leaving. The consequence is that Thief’s tension comes from the world’s rules rather than from scripting, and it’s why the levels still work. The full case for the immersive sim learning to listen is here.

System Shock 2 (Irrational Games and Looking Glass, 1999)

Ken Levine’s studio, built on Looking Glass’s Dark Engine, doing the original’s idea with an RPG’s stat economy bolted on. The upgrade system is the addition: cybermodules are scarce, and the game makes you commit to a build early enough that you’ll regret it.

The horror works because the fiction and the mechanics agree. Everything on the Von Braun is documented — the horror is genuinely in the log files — and the audio logs are doing the same job SHODAN’s station did five years earlier.

Thief II: The Metal Age (Looking Glass, 2000)

Shipped in March 2000. Looking Glass closed in May.

Thief II is the refinement pass: bigger levels, better mission variety, a mechanical antagonist who justifies the technology creep. It’s the most playable game the studio made and it arrived exactly as the studio ran out of money, which is an unfairly good summary of the genre’s whole commercial history. The sequel that refined everything deserves its own read.

Deus Ex (Ion Storm Austin, 2000)

Warren Spector and Harvey Smith, same year, and the genre’s most-cited game for good reason: Deus Ex is where the systems stop being a level-design tool and become a political one. The game has opinions about surveillance, biotech and authority, and it argues them by letting you take positions and then live in the consequences.

It’s also gloriously breakable. The skill system, the augmentations and the physics interact in ways nobody tested, and the game is better for it — the famous LAM-stacking wall-climb is an emergent bug that everyone agrees should stay. A game that lets you break it is the point.

Arx Fatalis (Arkane, 2002)

Raphaël Colantonio founded Arkane in Lyon specifically to make an Ultima Underworld successor when nobody else would, and Arx Fatalis is exactly that: an underground world, freelook, physical objects, and a spell system where you draw runes with the mouse in real time while things try to kill you.

The rune-drawing is the tell. It’s a bad idea in usability terms and a great one in simulation terms, because casting becomes a physical act with a failure mode. Arkane spent the next twenty years being the studio that kept this genre alive.

Dishonored 2 (Arkane, 2016)

The series’ best level design, and the Clockwork Mansion is the exhibit: a house that physically reconfigures at the pull of a lever, with a hidden between-walls space you can navigate if you work out that the mansion’s machinery has an inside.

It’s a thesis in a building. The mansion says: this world has rules, the rules have a back, and finding the back is the game. The clockwork mansion as thesis holds up as an argument.

Prey (Arkane Austin, 2017)

Saddled with a title belonging to a different game, sold badly, and it’s the most complete argument the genre has produced. Talos I is a single continuous station you learn as geography. The GLOO cannon turns level design into a material you author. The Mimics weaponise the object simulation itself — anything in the world might be an enemy, which means the game’s clutter is a mechanic.

And it has the neuromod system, which lets you take the aliens’ powers at the cost of the station’s own defences treating you as one. That’s a systemic consequence for a build choice, which is the sort of thing this genre spent twenty-five years reaching for. The immersive sim’s best argument in years is a fair description.

Hitman: World of Assassination (IO Interactive, 2016–2023)

The heretical entry, and the one I’d defend hardest. IO’s trilogy — reissued as World of Assassination — takes the genre’s simulation logic and narrows it viciously: one verb, six or so enormous levels, and a world of NPCs running schedules dense enough to be read as clockwork.

It’s the only immersive sim that’s a commercial success on its own terms, and the reason is scope discipline. Rather than simulating everything shallowly, IO simulates social legibility very deeply. Disguises, suspicion, sightlines and routine are the whole system. The puzzle box perfected earns its place.

What they share, and what it costs

The common thread is a refusal to check your work. Every game here has an intended path and treats it as one option among several, and the design labour goes into making the world consistent enough that your improvised path also holds together.

The price is legible in the studio list. Looking Glass: closed, 2000. Ion Storm Austin: closed, 2005. Irrational: gutted, 2014. Arkane Austin: closed, 2024. This is the most critically decorated lineage in games and it has a mortality rate like a Napoleonic regiment, because the thing that makes these games good — systems that multiply — is the thing that makes them impossible to schedule. The cost of letting players cheat is a real number on a spreadsheet somewhere.

Which is why the genre’s future is currently in smaller hands. Weird West came from Colantonio’s post-Arkane studio. Gloomwood is a Thief tribute from a tiny team. Cruelty Squad is one man with a grudge against taste. The scope has come down to where the economics work, and the design ideas have survived intact.

Where to play them

The Looking Glass catalogue is on GOG, patched to run on modern machines — System Shock, both Thiefs, and Ultima Underworld all work with community fixes that ship in the installers. System Shock 2 has an enhanced edition from Nightdive. Deus Ex is on GOG and Steam, and wants the community renderer to look correct. Prey, Dishonored 2 and Hitman are all on current storefronts and current hardware.

Start with Thief. It’s the one where the design idea is cleanest and the least dependent on tolerating a 1994 interface. Then Prey, and notice how much of Talos I is Citadel Station with twenty-three years of budget.

Advertisement
Advertisement
Jay
Written by Jay

vo.rs's games critic. Jay covers the medium as a system rather than a spectacle — this month's release, the indie nobody bought, and the Amiga game it's quietly descended from — asking what a mechanic makes you feel and why the loop holds. Learned to wait through a C64 tape load, never stopped playing since, and still finishes the odd 60-hour RPG out of spite. Expect argued verdicts, no score ever, spoilers below the line, and a running list of older games worth your weekend.