The evaluator is the locus of meaning. In a system built from snippets, walks, and probes, semantics do not reside in prose or intention but in the evaluator that interprets forms. To understand what a construct “means,” one must read the evaluator that executes it.
The evaluator defines the user language. Forms such as CLICK, NEIGHBORS, WALK, and SOLO are language constructs whose expressive power is fixed by what the evaluator recognizes and how it transitions state. Documentation can only describe; the evaluator decides.
Evaluation is staged and stateful. The evaluator processes blocks sequentially, producing and consuming intermediate state. Later forms depend on artifacts produced earlier, enforcing an epistemic order that mirrors a pipeline rather than a script.
CLICK is the evaluator’s side-effect gate. It marks the boundary where evaluation is permitted to mutate state. This makes effects explicit and optional, allowing a narrative surface to coexist with executable behavior without collapsing one into the other.
The evaluator separates scope from meaning. Operations such as NEIGHBORS establish a working set without interpretation, while later stages (walks, probes, compilation) introduce semantic distinctions. This separation is a design decision with consequences for modularity and reuse.
Extending the evaluator extends the language. Adding new walk modes or discourse-aware selectors does not require rewriting pages, only modifying evaluation rules. Small changes to the evaluator can therefore enable entirely new styles of use.
The evaluator is the true architectural boundary. Probes should remain referentially transparent, producing stable results from given inputs, while the evaluator is allowed to be interactive and effectful. Keeping this boundary clear is what allows incremental growth without semantic drift.
pages/the-evaluator