Warhammer:
The Enemy Within

The central promise of Book Mode is that the same generated story can be run by different groups of players · perhaps months or years apart · and still deliver a meaningful, surprising experience. That is a hard problem. Most authored content, when played twice by people who've compared notes, collapses into a guided tour of already-known beats. The tension evaporates. The sense of discovery is gone. For Book Mode to be commercially viable, it cannot be a single-use experience. People will share stories. Players will replay with different characters. The engine needs to produce content that can survive that exposure · and still feel alive.

Before designing any schema, we needed to know whether this was even solvable through structure alone · whether a pre-generated story could be made replay-resistant through authorial discipline, or whether it required algorithmic variation at generation time. The answer to that question would determine the entire architecture of Book Mode.

The Enemy Within (TEW) is the answer. Published by Games Workshop between 1986 and 1989, revised and expanded across decades, and now in a Director's Cut edition that Graeme Davis · one of its original authors · spent three decades refining, TEW is not merely considered a great campaign book. It is the campaign book. In the tabletop RPG community, arguments about the greatest adventure ever written converge on TEW with unusual regularity. Groups have been running it since Reagan was in office. New players who know nothing about Warhammer pick it up and find it works. Groups who've run it before run it again · differently. That longevity is structural, not incidental. Understanding precisely how it achieves that is the whole point of this research.

TEW is a picaresque narrative · its characters are deliberately small people swept into events far larger than themselves, moving through a world that doesn't care whether they show up. The protagonists begin as ordinary travellers: hired bodyguards, petty merchants, wandering scholars. By the time the campaign concludes across its five volumes, they have penetrated the highest levels of conspiracy in the Old World. But the world never paused to wait for them. That restlessness, that sense of a plot in motion regardless of the players' attention, is baked into TEW's structure from the first page.

The five volumes · Enemy in Shadows, Death on the Reik, Power Behind the Throne, The Horned Rat, and Empire in Ruins · are not chapters in a single linear story. They are locations. Each book is a place: a city, a river, a carnival, a frontier. The players move through these places, and the conspiracy is what they find threaded through all of them, connecting everything they encounter into a single coherent pattern that only becomes legible late in the campaign.

That coherence is maintained by a conspiracy spine that exists entirely independently of player action. The Purple Hand cult is plotting. Karl-Heinz Wasmeier, Law Lord of Middenheim, is using blackmail, murder, and political manipulation to consolidate power for forces of Chaos. The Skaven are moving beneath the city. These facts are true whether the players discover them or not. They would be true if no players ever sat down at the table. The GM knows them. The NPCs act accordingly. The world shows their effects. The players are not discovering a story that is written for them · they are discovering a history that already exists.

This is what makes TEW structurally different from most adventure design. The typical campaign book arranges the story around the players: they arrive, the villain reacts, the investigation proceeds at their pace, the climax happens when they are ready. TEW inverts this. The conspiracy proceeds at its own pace. Players who move slowly find that events have overtaken them. Players who rush find they've stumbled into something before they understood it. Neither experience is wrong · both are consequences of the world having its own momentum.

The campaign scales across five volumes without losing this coherence because the spine · the overarching Chaos conspiracy · touches every location. When players are bouncing down the Reik on a river barge, the conspiracy is there in the passengers sharing their boat and the cargo in the hold. When they reach Middenheim for the carnival, the conspiracy is in the tax laws, the Guild structure, the carnival performers, and the legal system simultaneously. Nothing is decorative. Every location element connects upward to the spine.

The distinction that makes TEW replayable can be stated plainly: the conspiracy is a fact, not an experience. The players' discovery of the conspiracy is the experience. These two things are entirely separable. The first · what is true in the world · is the spine. The second · how any particular group of players comes to understand it · is the flesh. TEW holds the spine constant across every playthrough. It surrenders the flesh entirely to the unpredictable intersection of player choices, GM improvisation, dice, and circumstance.

The result is that two groups can run Power Behind the Throne · the third volume, set in Middenheim during carnival week · and have almost no overlap in their actual experience. One group might make allies of the city watch and work through official channels, slowly learning that the institutions they trusted are compromised. Another might go directly underground, find the Skaven tunnels first, and approach the surface conspiracy from below with information the first group never had. A third might antagonize half the city, burn bridges with every faction, and still stumble into the truth through a drunken NPC in a tavern they visited out of boredom. All three groups will have discovered the same conspiracy. None of their stories are the same.

For the Narrative Engine, this distinction maps directly onto a generative question: what must the arc generator produce once and preserve forever, versus what the rendering layer can vary on every pass? The spine is generated by Sonnet in the Arc Generation phase and locked. The flesh · the specific clue paths, the precise sequence of discovery, which NPCs the player encounters first · is where Haiku has latitude. This is not just a design preference. It is what makes Book Mode economically viable: you generate the spine once, and it serves hundreds of playthroughs without expensive regeneration.

TEW's NPC design is not a list of characters · it is a web of contradictions. Every significant NPC in the campaign has at minimum two layers: what they appear to be, and what they actually are. The apparent layer is what players observe on first contact. The actual layer is what investigation, persuasion, or accident reveals. The relationship between these layers is not random · it is always in service of the spine. Every secret connects upward to the conspiracy, even if the connection is indirect.

Power Behind the Throne · universally considered TEW's masterwork volume · makes this web explicit. Its core adversary, Karl-Heinz Wasmeier, presents publicly as the Law Lord of Middenheim: a stern but legitimate authority, someone players might initially want as an ally. His secret is that he is simultaneously the Magister Magistri of the Purple Hand cult, using his legal position to protect cultists, eliminate threats, and advance a Chaos agenda beneath the veneer of civic order. The public face is not just a cover · it is actively part of the conspiracy. His legal authority is the mechanism of his evil.

This model repeats at every level of the NPC web. A carnival performer who seems to be a harmless entertainer is in fact a cult spy. A merchant complaining about new taxes is genuinely suffering from policies designed to destabilize the city · and his grievance, real as it is, is being manipulated by forces he doesn't know exist. A guard captain who obstructs the players is not a villain · he's a functionary doing his job · but his job protects a corrupt system, and whether players treat him as an enemy or a potential ally changes their path through the investigation.

What makes this web navigable · and what makes it feel like a coherent world rather than an overwhelming tangle · is that TEW provides the GM with a diagrammatic relationship map and nine pages of NPC usage guidance for Power Behind the Throne alone. This is the analog equivalent of what our Entity schema must produce at generation time: a structured representation of every NPC's public face, secret, connections, and agenda that the rendering layer can query against when generating dialogue or interaction beats.

The NPC web model also solves a specific generation problem: it prevents NPCs from feeling interchangeable. When every NPC has a defined agenda · not just a personality · their behavior in any given scene can be derived rather than invented. The Guild Factor doesn't want the players digging too deep into guild finances. That agenda shapes every interaction he has with them, regardless of what the players ask. The rendering layer doesn't need to be told how he reacts to every possible player input · it needs to know his agenda, and then it can derive the reaction.

One of TEW's most quietly radical design decisions is that every significant location includes a calendar of events that proceeds regardless of player action. Bogenhafen has its Schaffenfest. Middenheim has its carnival week. The cult has its meeting schedule. The ritual has its window. These are not encounter prompts waiting for players to trigger them · they are facts about the world that exist in time, and the players are moving through that time whether they engage or not.

A typical event timeline entry reads something like: Day 1 · the cult meets in the Merchant Guild cellars at midnight. Day 3 · the merchant who knows too much is found dead. Day 5 · a prominent city official disappears. Day 7 · the ritual begins. Day 10 · the ritual concludes unless interrupted. Players who arrive at the start of the carnival have ten days to unravel the conspiracy. Players who waste three days pursuing a red herring arrive to find the merchant already dead · and the pressure has escalated. Players who ignore every signal find themselves present at a ritual they had no idea was coming.

There is a second function that passive timelines serve beyond dramatic tension: they provide narrative content for the rendering layer even when players are doing nothing. If the cult meets on Day 1 whether players attend or not, that meeting creates a feed item, a board notice, a rumour in the chat channel. The world is generating signal. Players who are paying attention will notice the pattern. Players who aren't will still experience a world in motion, even if they don't understand what is moving.

This is the origin of what the engine calls the passive timeline in the Arc schema: not merely a design principle borrowed from TEW, but a structural requirement that every arc define what the world looks like at each moment if the players are absent. In the Arc schema, the passive_timeline array entries each carry a visible_signs field the observable surface trace of an event that players might notice without knowing what caused it. A closed warehouse. An empty chair at the merchant's usual table. A watch patrol that's been doubled without explanation. These are the signs of a timeline advancing, rendered in the appropriate delivery channel, whether or not any player is watching.

TEW is famous for its physical handouts · and the fame is deserved, because the handouts represent a design principle that most campaign books get entirely wrong. A TEW handout is not a summary of information the GM would otherwise narrate. It is a document that exists in the world of the campaign. A wanted poster is not a convenient information delivery mechanism. It is an actual wanted poster · written in the vernacular of an Imperial magistrate, laid out with the typography of an eighteenth-century broadsheet, including institutional boilerplate that no narrator would ever include because narrators naturally strip the noise to signal. The noise is the point.

When players receive a TEW handout · an intercepted letter from a Purple Hand cultist, a legal summons from the Middenheim court, a hand-drawn map with annotations in a stranger's handwriting · they are not receiving information. They are receiving an artifact. The artifact carries information, yes, but it also carries context: who wrote this, in what circumstances, with what expectations of the reader. A cultist's letter doesn't announce "I am a cultist conspiring against the Empire." It uses the language of ordinary correspondence to discuss ordinary business, and the conspiracy is visible in the gaps, the euphemisms, and the unexplained references to people and places the players might eventually recognise.

TEW's handout design is, in effect, a delivery channel specification document built into the adventure itself. When TEW defines what a wanted poster looks like, it is not just providing props · it is establishing a grammar. A wanted poster uses declarative sentences. It includes a physical description, a named offence, a reward amount, and an authority signature. It does not editorialize. It does not explain context. It assumes the reader knows what a wanted poster is for. When the Narrative Engine generates a board notice, every one of these attributes should be present · not because someone wrote a rule that says "include reward amounts," but because the board notice is a wanted poster, and wanted posters have a form that exists in the world for functional reasons.

This is what the engine means by delivery channel rendering. A feed broadcast doesn't sound like a wanted poster. A chat message doesn't sound like a legal summons. Each channel has a form that exists in the world for its own functional reasons, and that form is not optional flavor · it is the mechanism by which the content lands as real rather than narrated.

The research produced six concrete design decisions that are now locked into the engine's architecture. Each one maps a TEW mechanism to a specific schema field or generation constraint.