Summary
Point has one public authoring language: semantic .point source. Core exists inside the compiler so the checker and emitters have a precise representation.
Semantic source
Semantic source is what people and agents write:
module Pricing
calculation annual price
input monthly price: Int
output annual price: Int
annual price is monthly price * 12The words in source preserve intent. A calculation is not just any function; it is a pure derived value with named inputs and output.
Core IR
After parsing, Point lowers semantic source into typed core data structures in memory. Core has implementation-level concepts such as functions, values, calls, loops, assignments, and return values.
Authors do not write core syntax in .point files. If you see names like fn, let, or generated camelCase identifiers, you are looking at compiler internals or emitted target code, not public Point source.
Why keep the split
Semantic source keeps author intent visible. Core IR gives the compiler a compact, typed representation. Generated JavaScript, TypeScript, or Python is a build artifact.
That split lets Point stay friendly to agents without making the emitter guess from loose text.