diff --git a/design/models/0001-authoring-surface.md b/design/models/0001-authoring-surface.md
index 92b6fb7..29fd3bb 100644
--- a/design/models/0001-authoring-surface.md
+++ b/design/models/0001-authoring-surface.md
@@ -58,7 +58,18 @@ Every other maximal token is classified post-hoc:
 Consequence: arithmetic operators like `+` / `-` / `*` / `/` / `%`,
 qualified names like `io/print_str`, and cross-module references
 like `std_list.map` are all single ident tokens with no special
-lex rule. The only reserved tokens are `(`, `)`, and whitespace.
+lex rule. The only token *delimiters* are `(`, `)`, and whitespace.
+
+One character is *reserved within* a token: `$`. Any identifier
+token containing `$` is rejected by the lexer
+(`LexError::ReservedDollar` in `crates/ailang-surface/src/lex.rs`),
+because `$` is reserved for the compiler's synthetic namespace
+(match-scrutinee, lifted-let-rec, and shadow-rename binders).
+Making `$` unauthorable keeps that synthetic namespace
+collision-free by construction. The reservation is on the
+Form-(A) authoring surface only — `$` inside a string literal or
+a comment is unaffected, and the canonical `.ail.json`
+deserialization path is intentionally not guarded.
 Bool literals (`true`, `false`) and unit (`(lit-unit)`) are
 disambiguated by parser context, not by lex. Comparison and
 equality are class methods (`eq` / `compare`) and named fns