Public
CLCA
Cross-Linguistic Compositional Analysis
The evidence base. A reproducible pipeline that asks sixteen genetically unrelated languages how they build claims about truth, and recovers the shared constraint geometry the rest of the architecture rests on.
What it is
CLCA elicits how sixteen unrelated languages (Georgian, Yoruba, Basque, Quechua, and twelve more, spanning ten families and three isolates) build claims about truth, without leading the witnesses: the prompts carry no theoretical vocabulary at all. Two independent eight-language sets and two protocol variants act as mutual controls, and a dual backend (Claude and GPT) provides cross-model validation.
Role in the program
CLCA supplies the empirical content the rest of the architecture formalizes. Where AOML states what the constraints are, CLCA is how they were found: from the data itself rather than from theory. Papers I and II report and build on it, and every prompt and run is public so the result can be re-derived and diffed against ours.