The Mech plugin defines a small user language whose meaning is given entirely by its interpreter. Blocks such as CLICK, NEIGHBORS, WALK, and SOLO are not metaphors or UI hints; they are executable forms. As in Steele and Sussman’s analysis of interpreters, the available “programming styles” for users are determined by what the evaluator actually implements, not by documentation or intention.
The Discourse Graphs snippet is a specification written in the surface syntax of its own future interpreter. Expressions like “WALK 10 questions” and “WALK 30 claims” are not currently executable semantics but design intentions embedded in executable form. This mirrors the meta-circular move in The Art of the Interpreter, where language evolution is expressed by modifying and reusing the language itself.
CLICK marks a controlled side effect in an otherwise narrative medium. A wiki page is normally descriptive and referentially transparent, but CLICK introduces an explicit execution boundary where state is committed. This corresponds to Steele and Sussman’s argument that modular systems require carefully packaged side effects, not their elimination.
NEIGHBORS establishes scope before meaning. It materializes a working set from federated sitemaps without interpreting discourse roles or argumentative intent. This separation reflects a core interpreter design principle: scoping and binding decisions precede semantic specialization.
ClaimLinkSurveyProbe externalizes discourse semantics from prose into typed edges. By interpreting pagefold structure as question, claim, support, or oppose, it extends the effective language of the wiki without changing page syntax. This is an interpreter extension in Steele and Sussman’s sense: a small semantic change that enables new modular constructions.
Discourse roles act as a scoping discipline for narrative assembly. Selecting “questions” or “claims” during a walk is not a simple filter but a rule for what contextual information is implicitly carried forward. This is analogous to dynamic scoping as a structured side effect: locally invisible, globally decisive.
Basis walks function like toy interpreters. Each basis walk is a maximal, exportable narrative path that can be reused, varied, or extended without reanalyzing the entire discourse graph. As with Steele and Sussman’s incremental interpreters, system evolution proceeds by modifying small, comprehensible artifacts rather than redesigning the whole.
See The Evaluator