The §1b two-surface law landed in a11c1bd; docs/* (ist) still showed the
pre-§1b universal-pin spellings. Swept top-level docs/*.md: 12 coercion sites
rewritten to the letter form (8::u8, x::f32, [1,2,3]::List<Int,12>, …); all
genuine specializations left (path/generic-call receivers, and the []::<T>
/ []::<T,N> empty-list specialization surfaces per plan_values_02 §1b and
the coercion_letter_form golden). syntax.md's superseded "universal pin"
law paragraph re-stated as the two-surface law (locked 2026-05-28, amended
2026-07-17); list-literal element/bound prose reworded to the coercion-
target form. Verified: remaining `::<` in docs/*.md are specializations
only; no file outside docs/ touched.

Flagged for follow-up (not this sweep's scope): syntax.md's claim that
annotated-bound assignment does not type-check may be stale against
plan_values_02 §2's bound-bucket law — surfaced to owner, not edited.

Closes debt #44 for top-level docs; docs/impl/** not in scope this pass.

@@ -0,0 +99,4 @@
+/// Whether `line` contains `needle` as a whole token — the char after the
+/// match must not continue an identifier, so `SyntaxKind::Pipe` does NOT match
+/// `SyntaxKind::PipeRParen` / `SyntaxKind::PipePipe` (a real substring hazard).
+fn contains_token(line: &str, needle: &str) -> bool {

@@ -0,0 +116,4 @@
+    );
+    let body = &stdout[LINE_PREFIX.len()..stdout.len() - LINE_SUFFIX.len()];
+    let (passes_str, seconds_str) = body.split_once(';').unwrap_or_else(|| {
+        panic!("expected `<passes>;<seconds>` between the fixed ends, got `{body}`")