Back to Projects

Gothic Grandma, LLC

Founder & Lead Engineer · 2025–Present · Private Platform

MUSE Living Worlds is an interactive storytelling platform built on the MUSE simulation ecosystem—a custom C++ runtime that models biological needs, psychological states, and emergent behaviors.

At its core is a deterministic execution engine written in C++ that runs compiled behavioral kernels. Visual tools generate kernel source code, which is compiled into binaries and executed under a unified runtime shared across research tooling and consumer applications.

Platform-first architecture: Rather than building a game, I built the infrastructure that games (and research tools, and creative tools) run on. MUSE is designed like a platform company: central runtime, role-based access control, multi-tenant permission models, real-time collaboration primitives, and observability from the start.

LLMs are used at the periphery — for authoring, interpretation, and assistance — never inside the simulation loop.
All world state, causality, and long-term consequences are produced by deterministic systems, not probabilistic generation.

MUSE is built as a full-stack platform:
execution engine → data persistence → inspection tooling → role-specific interfaces.

Gothic Grandma. Laboratories builds the platform and research layer; Gothic Grandma. Studios builds living worlds on it.

Visit Gothic Grandma →

What Problem MUSE Solves

Problem: Modern interactive fiction, educational software, and AI-driven storytelling systems are built on static scripts or probabilistic text generation. They can be immersive, but they do not simulate causality. Their characters do not truly perceive, decide, or change in response to integrated biological, psychological, and social forces. Consequences in these platforms are authored, not emergent.

Solution: MUSE enables embodied, scientifically grounded simulation where:

• Perception, decision-making, learning, and relationships arise from real models of biology and psychology
• Long-term change emerges from cause-and-effect, not narrative branching
• Users can probe systems like scientists—not just consume a story
• Worlds persist, evolve, and surprise without relying on LLM memory or hallucinated state

The result is a new medium: interactive worlds where people don't just read about perspectives—they inhabit them.

What Makes MUSE Different

Deterministic Simulation, Not Generation

• Real causality — consequences emerge from biological and psychological systems, not scripts
• Persistent state — worlds remember everything through database-backed memory, no LLM context limits
• Inspectable logic — trace any behavior back to root cause through explicit data flow
• No hallucination risk — simulation truth is ground truth; LLMs only translate at boundaries

Event-Driven Architecture

• Burst computation on prompts — processes 100,000+ entities in milliseconds when user acts
• Efficient by design — idle when paused, no wasted cycles on continuous rendering
• Scales independently — simulation complexity grows with world richness, not interaction frequency
• Infrastructure-style patterns — central state management, distributed timing, resource-aware scheduling

Visual → Compiled Pipeline

• Declarative configuration — design systems visually in CYPHER, define behavior through graphs
• Generated execution — visual models compile to optimized GPU/CPU kernel code automatically
• Database-driven contracts — all operations defined in SQLite schema, type-safe runtime dispatch
• No hardcoded logic — every system emerges from schema, enabling full traceability

Multi-Tenant Platform Design

• Role-segmented access — readers see GLYPH, engineers see CYPHER, researchers see PYTHIA
• Same runtime core — everyone uses the same simulation instance, different permission layers
• Separation of concerns — system engineering isolated from world authoring isolated from reading
• Control plane thinking — CLIO functions as ops dashboard, like modern platform companies expose to customers

Authentication, Authorization & Collaboration

• OAuth-based identity layer — designed for federated authentication across consumer and enterprise contexts
• RBAC permission model — readers, authors, researchers, engineers share runtime with different capability surfaces
• Real-time collaboration infrastructure — operational transform foundations for concurrent editing in CODEX
• Postgres-backed state layer — persistent, auditable permission hierarchies with cross-world identity

Scientifically Grounded

• Based on embodied cognition research — not game AI heuristics, actual perceptual and motor control models
• Biological timing — systems execute at realistic rates (50ms reflexes, 1000ms deliberation)
• Emergent complexity — societies, economies, cultures arise from simple rules at scale
• Transparent models — no black boxes; every system is inspectable, verifiable, traceable

The Ecosystem

Eight core tools spanning reader, creator, research, and engineering roles. Each role sees only what they need.

Consumer Layer

What readers experience. Consumer-facing tools are built in Flutter for reading comfort, animation smoothness, and cross-platform parity.

Core Runtime

Language Layer

System Engineering

Control Plane

Research & Authoring

Impact

• Built full data → execution → UI → observability → AI-assisted toolchain as a single coherent platform
• Enabled full lifecycle traceability across development, execution, and analysis
• Designed for long-horizon extensibility via modular tools and role-segmented permissions
• Separated system engineering from world authoring—creators think in semantics, not mathematics

Who It's For

• Readers — experience Living Worlds through GLYPH, interacting via natural language
• Researchers — run experiments and export data via PYTHIA without touching code
• Creative directors — author worlds through CODEX using semantics and description
• System engineers — build the perceptual, behavioral, and physical systems that power everything via CYPHER and CLIO

Why This Exists

MUSE grew out of my work in computational neuroscience—specifically research on sensorimotor control, affordance perception, and embodied cognition. For years I studied how biological agents perceive possibilities for action, coordinate motor primitives, and adapt through experience. I kept thinking: what if people could actually inhabit these models, not just read about them?

Most interactive fiction relies on branching narratives or probabilistic text generation. Both feel authored—because they are. Branching stories give you choices, but the world doesn't truly respond; it follows predetermined paths. LLM-driven narratives feel fluid, but they hallucinate consistency—characters forget, worlds contradict themselves, consequences evaporate.

Real people don't experience life as branching paths or probabilistic outputs. They perceive affordances in their environment, coordinate distributed actions across biological systems, form memories tied to spatial and emotional context, and adapt their behavior based on integrated feedback. That's the kind of experience I wanted to build.

MUSE is an attempt to create scientifically-grounded simulation you can inhabit—where characters have genuine perceptual systems that detect affordances, biological needs that drive motivation, social bonds that evolve over time, and memories that persist with causal integrity. Not to replace traditional storytelling, but to create a new medium where narrative emerges from authentic cause-and-effect.

The technical architecture—event-driven execution, visual-to-compiled pipelines, multi-tenant workbenches— emerged from the requirements of that vision. To simulate 100,000 entities with biological realism, you need infrastructure-style patterns. To let domain experts (writers, researchers, readers) interact without touching code, you need role-segmented interfaces. To ensure reproducibility and traceability, you need database-driven contracts.

Platform lesson: The same way researchers don't want to rebuild ML pipelines for every study (ePOCHE), or scientists don't want Docker chaos for every scan (MR.Flow)—storytellers and readers shouldn't have to choose between rigid branching or hallucinated consistency. MUSE provides a third path: emergent narratives from deterministic, inspectable simulation. Stories that surprise their creators because the world actually responds to what you do.

Building this required combining my research background (embodied cognition, biological timing, affordance theory) with platform engineering thinking (event-driven architecture, central state management, declarative compilation). It's a synthesis of neuroscience and infrastructure design—both focused on making complex systems work reliably at scale.

MUSE is still early. The runtime works. The core workbenches exist. The first Living Worlds are taking shape. But the vision is larger: a platform where anyone can build and experience worlds that think, remember, and evolve—not through scripts or generation, but through authentic simulation.