Field Evolution Runtime

A language that runs

physics.

Write a .tri file. Pair two regimes. Run it.
Models neurons, climate, markets — anything that evolves continuously.

Get startedSee the syntax
562 / 562 tests passing

The Language

Readable syntax.
Real physics output.

neuro_climate.tri
// declare two physical regimes
reg neuron  : HodgkinHuxley = 4;
reg climate : ENSO_recharge = 4;

// couple them, evolve for T = 18 seconds
pair(neuron, climate) kappa=-2.5 for T=18.0;

// observe emergent structure
OBSERVE neuron  crystallinity, peak;
OBSERVE climate crystallinity, peak;
$ triad run neuro_climate.tri
neuron = {
  crystallinity  0.9423
  peak           0.3769
  atom_count     21
}

climate = {
  crystallinity  0.8607
  peak           0.5366
  atom_count     24
}

// no target was specified.
// structure emerged from coupling alone.

The Triad

Three forces.
One field.

P1 / Phase

Phase and Oscillation

Dispersion spreads structure across the grid by wavenumber content. The fractional Laplacian makes anomalous, non-local dispersion tunable by a single exponent.

alpha(-Delta)^(sigma/2) * psi

P2 / Memory

Memory and Self-Reference

Three Ornstein-Uhlenbeck channels track the field’s own history. The state at time t depends on what the field has been through: fast, medium, and slow timescales.

V_mem = sum_j lambda_j * y_j(t)

P3 / Coupling

Coupling and Open Systems

Dissipation and FDT noise keep every field thermally open. Substrates from entirely different physical regimes share one runtime and couple through their evolving densities.

-i*Gamma*psi + eta(2*Gamma*kT)

Get Started

Up and running
in 30 seconds.

$git clone https://github.com/triadlang/triadlang && cd triadlang
$pip install -e .
$triad run examples/triad/neuro_climate.tri

Requires libfftw3-dev + libgc-dev for native compilation.  CUDA optional via cupy.  562 / 562 tests passing.