diff --git a/README.md b/README.md
index 57b1261..71e47db 100644
--- a/README.md
+++ b/README.md
@@ -771,9 +771,9 @@ Serves [Goals](#goals) 1, 4 and 5.
- An image's title becomes its caption when the image stays an image, and otherwise the title of
the link to the image.
- And this, for another reason: a link whose text holds an autolink keeps the inner link and leaves
- the outer brackets literal text, which the spec requires and the reference itself breaks, nesting
- one `` in the other.
+ And one more, for another reason: a link whose text holds an autolink keeps the inner link and
+ leaves the outer brackets literal text. The spec requires that; the reference implementation
+ nests one `` in the other.
- Not every document converts back: `adfToLosslessMarkdown` refuses a document nesting past 500
levels, a document meeting an `unsupported-node-shape` cause listed for it, for a writer or for
every call, and a conversion `engine-limit` stops. Show the refusal and keep the document read-only; saving
diff --git a/docs/decisions.md b/docs/decisions.md
index 464f5b2..a4020be 100644
--- a/docs/decisions.md
+++ b/docs/decisions.md
@@ -558,7 +558,8 @@ and HTML parsers are written in this repo.
## Standards ship as data
-2026-08-30, the CommonMark suite 2026-09-05 and ADF's schemas 2026-09-13, the maintainer. Goals 1,
+2026-08-30, the CommonMark suite 2026-09-05 and ADF's schemas 2026-09-13, the maintainer;
+Unicode's case folding 2026-11-15, the maintainer's plan. Goals 1,
5 and 8. Valid while each table is fixed data a dependency would only wrap.
A table a standard fixes is data rather than a dependency: HTML5's 2125 semicolon-terminated
@@ -570,9 +571,10 @@ renumbered suite. A spec bump is a deliberate re-pin, exceptions re-derived by h
Atlassian's ADF JSON Schemas ship vendored the same way, at `spec/adf-schema/`, rather than as the
`@atlaskit/adf-schema` dev dependency — CommonJS-only, some fifty packages with React among them,
and a release most days for Renovate to automerge — re-pinned by hand when a payload or a report
-shows the need. Unicode's `CaseFolding.txt` (18.0.0) is vendored at `spec/unicode/` and packed in
-`src/markdown/commonmark/case-folding.ts` the same way; its test holds every C and F mapping to the
-file, so a re-pin is a new file and a green test.
+shows the need. Unicode's `CaseFolding.txt` (18.0.0), whose common (C) and full (F) mappings
+CommonMark folds a link label by, is vendored at `spec/unicode/` and packed in
+`src/markdown/commonmark/case-folding.ts`, like the character references. Its test holds every
+mapping to the file. A re-pin replaces the file, repacks the data and updates the test's count.
## fast-check
diff --git a/src/markdown/commonmark/case-folding.ts b/src/markdown/commonmark/case-folding.ts
index 4f8ec97..965911b 100644
--- a/src/markdown/commonmark/case-folding.ts
+++ b/src/markdown/commonmark/case-folding.ts
@@ -1,7 +1,9 @@
-// Unicode's `CaseFolding.txt` (vendored at `spec/unicode/`), the common (C) and full (F) mappings that CommonMark
-// folds a link label by. The Turkic (T) and simple (S) mappings stay out: the full ones supersede the simple.
+// Unicode's `CaseFolding.txt`, vendored at `spec/unicode/`, which CommonMark folds a link label by. Only its common
+// (C) and full (F) mappings are here: the simple (S) ones are superseded by the full ones, and the Turkic (T) ones
+// apply to a language CommonMark does not name.
// A common run reads `first:count:stride:delta` in hex: `count` code points from `first`, `stride` apart, each
-// folding to its own plus `delta`. A full folding reads `code point=code points` in hex.
+// folding to its own code point plus `delta` (`+` or `-`). A full folding reads `code point=code point,code point…`
+// in hex.
const commonRuns = [
'41:1a:1:+20 b5:1:1:+307 c0:17:1:+20 d8:7:1:+20 100:18:2:+1 132:3:2:+1 139:8:2:+1 14a:17:2:+1 178:1:1:-79',
'179:3:2:+1 17f:1:1:-10c 181:1:1:+d2 182:2:2:+1 186:1:1:+ce 187:1:1:+1 189:2:1:+cd 18b:1:1:+1 18e:1:1:+4f',