Active family
Local windows select regions across nested state spaces and determine which dynamics can participate.
Virgil / Cognitive systems program
AGI is not a text predictor.
Virgil investigates cognition as a contained dynamical system whose evolving internal geometry can settle into stable, discrete expression.
01 / Thesis
State first. Expression second.
The architecture separates the dynamics that form a thought from the interface that renders it communicable.
Instead of treating intelligence as a single pass through a token sequence, Virgil maintains nested coordinate spaces whose local regions evolve, interact, and converge through time.
Only after the active system becomes sufficiently stable is its state projected onto Soapstone, a compact topological surface that supports discrete readout.
02 / Field
A working geometry.
The field below is an interpretive instrument for the program’s central idea: global expression emerging from coordinated local motion.
03 / Architecture
From evolving state to stable expression.
Gaze dynamics
Gradient, coupling, and advection terms move the active system toward coherent attractor assignments.
Commit gate
Velocity, ambiguity, and drift must remain below explicit thresholds before the state is frozen.
Expression surface
Soapstone projects the settled state onto a compact field and reads its topological cell structure.
04 / Instruments
Inspect the system directly.
Public tools expose different parts of the architecture rather than reducing the program to a chatbot interface.
Architecture walkthrough
Follow the active family, gaze law, commit gate, projection, and expression stages.
Expression surface selector
Inspect field geometry, cell decomposition, topological charge, and discrete readout.
Decoder dashboard
Submit text to the available endpoint and inspect its rendered expression surface.
05 / Record