One thesis.
Four expressions.
Every language here starts from the same place: structured declaration over imperative code. The result is always a sovereign artifact — something complete in itself, owned by whoever holds it, executable without infrastructure.
Four domains. Four syntaxes. One underlying conviction.
CAL
StrategyCascade Analysis Language
A DSL where the cascade is the program — 11 keywords, 3 formulas, structured intelligence source code.
Traces how failure and success propagate across six organizational dimensions, scores the result, and produces a deterministic decision.
EMBER
AgentsEmergent Behavior Expression Runtime
A language for AI agents — structured intent declaration for agentic pipelines.
Declares agent behavior, intent, and orchestration as readable markup. Powers Project Phoenix — the agentic greenfield pipeline.
RECALL
PublishingStructured Publishing Language
COBOL-inspired publishing language for the web — declare pages, compile to self-contained HTML.
Separates content from presentation. Produces sovereign, portable documents. What COBOL would have built for the web.
Mere
AppsWorkbook Format for Apps
A file format for reactive applications — declarative screens, state, computed values, and behavior.
A .mp.html file is a complete, self-contained, executable artifact. Open it — it runs. Send it — it travels. The file is the app.
What they share
These aren't four separate projects. They're four dialects of the same philosophy.
Declarative over imperative
Every language describes what something is, not how to execute it. The computer follows instructions. These languages give the instructions shape.
Sovereign artifacts
Output is always a self-contained artifact — a CAL analysis, an EMBER pipeline, a RECALL document, a Mere workbook. Each one owns its own meaning.
AI as the primary author
Each language is designed as a generation target. Bounded vocabulary, unambiguous semantics, no escape hatches. A model produces consistent, valid output reliably.
Human-readable by constraint
The same constraints that make AI generation reliable make the output readable to humans. Minimalism serves both audiences simultaneously — by accident, not by design.
CAL — how do we reason about cascades? EMBER — how do AI agents declare intent? RECALL — how do documents become artifacts? Mere — how do files become applications?
The pattern continues. The next one is already implied.