SymLan: substrate-native symbolic readout, stated conservatively.

The Core Idea

SymLan is an experimental research language for a kind of computation that conventional programming languages do not describe cleanly: programs where some important values cannot be known from source text alone, because they depend on a run of a dynamical system.

In conventional software, the programmer normally writes the symbols and the machine manipulates them. In SymLan, the programmer describes a substrate model, run conditions, coupling constraints, and a readout method. The vocabulary produced by that run is not available at parse time. It is pending until the run completes and a validation procedure accepts it.

This is the core idea of SymLan. The language does not ask the reader to believe that any physical substrate already solves the problem. Instead, it makes the unresolved state explicit. A vocabulary is not treated as real simply because the source file names it. It becomes usable only after resolution, measurement, and validation.

What This Preview Does Not Claim

This preview does not claim that physical hardware currently runs SymLan, that SymLan replaces existing programming languages or transformer models, or that symbolic grounding is solved.

The narrower claim is that SymLan provides a disciplined language and validation framework for exploring whether stable attractor behavior can be registered, encoded, and replay-tested as a symbolic vocabulary under explicit constraints.

Hello World, Reframed

The smallest useful SymLan program declares a reference simulation, registers a pending vocabulary, runs the simulation, resolves the vocabulary, and sends a token only if validation succeeds.

substrate H: hopfield_sim {
  N: 1024;
  M: 54;
  alpha: 0.052734;
  W: learn(rule: projection, patterns: 54, seed: "hello");
}

vocab V = register(
  H,
  threshold: auto,
  features: PEFP(k: 9),
  method: PiAPlus,
  min_basins: 54,
  min_entropy: 5.0bits,
  max_collisions: 0
)

run discovery for 10000steps

after resolve V {
  replay V { trials: 1000; require: fidelity >= 0.90; }
  send V.top(1) -> out
}

V.top(1) is not written by the programmer as a fixed symbol. It is discovered during vocabulary resolution and only becomes usable after validation succeeds.

Claims Ladder

Physical Backends

Future experimental physical backends, including chemical oscillator arrays such as Belousov-Zhabotinsky systems, vanadium dioxide devices, photonic systems, and others, remain active research targets. They are not yet validated against the reference simulation path.

Intellectual Property

SymLan is associated with U.S. Provisional Patent Application No. 63/973,680, filed February 2026 by James Allen Clow and Melissa Ellen Clow. Selected reference materials may be released publicly while preserving appropriate patent, trademark, registry, and commercial licensing options.