Publication

An LLM can decide your fate—literally, though only in a game for now ❤

Fabula Rasa Dead Man Talking is a VR experience in which you face execution in a medieval square and must convince a witch, bard, executioner, and other characters to speak in your defense.

An LLM can decide your fate—literally, though only in a game for now ❤

Fabula Rasa: Dead Man Talking is a VR experience in which you are locked in a cage in a medieval square and prepared for execution. The king is about to pass judgment, and you must persuade a witch, bard, executioner, and other characters to speak in your defense—rather like that childhood coin game millennials may remember 🫤. Instead of choosing one of three prepared lines, you have to argue your case aloud, talk your way out, improvise, lie convincingly, or appeal to sympathy. The project does not look like a basement demo made by a lone enthusiast: it has already won the Audience Award at the SXSW 2026 XR Experience.

The key is that its NPCs do not simply recite pre-written lines. They listen, remember who you claim to be, and react to your profession, accusation, and speaking style. Heise reports that conversations developed differently across two 30-minute playthroughs, and the experience felt more like improvisational theatre than a conventional game 🎭

What interests me here is not that “AI can now write dialogue”—that is the least interesting part. It is that a game can suddenly test your real ability to communicate instead of putting you through the usual stress test of reaction time, aim, or character grinding. Persuade them and you live. Fail and you are fed to the monster. True to life, in a way.

This is an important shift for the industry. A writer used to build a dialogue tree. Now they design the framework: the world, roles, conflict, behavioral boundaries, and dramatic structure of the situation. The LLM under the hood becomes an engine for live communication and, to some extent, a co-author of the product ✏️

If we set token costs aside, the business potential is compelling too. These mechanics could be useful not only in games, but in negotiation and sales training, soft-skills development, legal simulations, corporate training, and VR onboarding. I can imagine a future without dull “choose the correct answer” tests: instead, a client is angry, an official applies pressure, an investor has doubts, and the employee must talk their way through the scenario.

I am honestly surprised by how few games have an LLM under the hood. We have already written about different VR experiences, and the next great leap in VR realism may not come from graphics. It may come when virtual characters finally stop being furniture and new interaction mechanics emerge.

If you are building a game, simulator, or corporate training product—or simply wondering how to add AI to an interactive experience without creating AI slop—this is exactly the kind of case I can help you analyze. Get in touch for a consultation ✨