Consciousness is not a thing inside the brain. It is a recursively stabilized perspective.
A theory proposing that conscious experience emerges when a system recursively models itself, its world, and its own modeling across time — under conditions of uncertainty, salience, and bounded control.
Most theories of consciousness ask where it lives or what it integrates. RTC asks how it stabilizes.
Consciousness is not a substance, location, or computational output. It is not produced by any single brain area or any specific quantity of integrated information.
Consciousness is not a single process either. It is a dynamically maintained architecture — many processes coordinated within bounded ranges to sustain a point of view.
What stabilizes is perspective itself: a system continuously modeling its own relation to a world, governed against runaway recursion, reconstituted across time.
Seven ways into the theory.
If you only have a few minutes, start with the Engine. If you want to read, start with the Manuscript. The other sections can be visited in any order.
The Perspective Engine
Move seven parameters and watch a recursive architecture stabilize, fragment, or run away. The five stability states emerge from interactions, not thresholds. The fastest way to feel the shape of the theory.
The Seven-Layer Architecture
Six generative layers — Signal, Distinction, Salience, Self-in-World, Meta-Governance, Diachronic Reconstitution — produce the seventh: Stabilized Perspective. Each layer expands into plain language, technical frame, neural substrate, and failure mode.
Theory Comparison
A positional map of RTC against Global Workspace, Integrated Information, Predictive Processing, Higher-Order Thought, Attention Schema, and Active Inference. What each captures, what each leaves unaddressed, what RTC adds.
Falsifiable Predictions
Five predictions with experimental designs and visualized data shapes: meta-cognitive dissociation, temporal fragmentation, salience gating, self-world dissociation, and the bounded-recursion U-curve. Designed to be falsified.
AI Consciousness Lab
Five reference architectures — Basic LLM, Agentic LLM, Memory-Augmented, Self-Monitoring, and a hypothetical RTC-Aligned system — scored qualitatively across the seven layers. Where current AI sits, and what would have to be added next.
The Evaluation Rubric
The instrument the AI Lab presupposes. Each of the seven layers specifies an architectural question, diagnostic signals, and concrete level criteria — including admissions of where the criteria themselves remain open research questions.
The Living Manuscript
The theory in long form: origin, core thesis, definitions, architecture, neuroscience alignment, AI implications, thermodynamic constraints, predictions, open questions, future work. With timeline.
Ryan Erbe
December 2024 — present
The Recurse Theory of Consciousness began in December 2024 as an attempt to explain conscious experience through recursive self-reference, emotional salience, and the stabilization of distinctions. Through 2025 it expanded — first into recursive self-in-world modeling, then into questions of human-AI relational emergence, then into thermodynamic and temporal constraints on bounded recursion.
This site is the theory's translation into something explorable. It does not replace the manuscript. It is a second surface — a way to feel the shape of an idea that has mostly lived in essays, sketches, and conversations.