
AI-narrated story where the rules live outside the LLM — does this count as a gamebook to you?
Hi, I'm Oleh - software engineer building a side project. AI text generation is great at prose but a terrible referee. It drops your inventory, retcons your choices, decides you've already met someone you haven't. So I built a browser thing where the rules live outside the LLM - and I'm not sure if what came out is a gamebook or something else. Looking for this sub's honest read.
How the split works: the LLM narrates each section and proposes state changes (move, set_time, add_item, set_objective) as structured calls. A separate engine validates them and renders the changes as small annotations inline - "Moved to Castra Regina, Raetia", "Winter, 179 CE" - plus persistent chips at the top of the screen showing where and when you are. The book remembers, even if the AI forgets.
Screenshot below: a section from a Marcus Aurelius / Roman frontier scenario. You can see the location and time chips, the inline state markers, and the free-text action input.
Mobile-first, no desktop layout yet, cover art is placeholder until I can afford to commission real art.
Around for the next few hours. The question: if you read 10 sections of this, would it feel like a gamebook to you, or a chatbot with stage directions sprinkled in? What's the specific thing that'd flip it?