diff --git a/.claude/CLAUDE.md b/.claude/CLAUDE.md
index b1d1c9e6..a09b2ffa 100644
--- a/.claude/CLAUDE.md
+++ b/.claude/CLAUDE.md
@@ -112,11 +112,19 @@ for the canonical statement.
 > ruling and agent behaviour had diverged: agents were being instructed to migrate
 > *away* from the estate's first-choice runtime.
 >
-> **Still unresolved, deliberately left alone:** the table below bans TypeScript in
-> favour of AffineScript, while `LANGUAGE-POLICY.adoc` §1.2 states "TypeScript is
-> *permitted under Bun*". Those two cannot both be right. The AffineScript
-> direction is a substantive architectural position rather than an oversight, so
-> it is flagged here for an owner ruling instead of being changed.
+> **RESOLVED 2026-08-25 — this file governs.** The contradiction previously
+> flagged here (this table bans TypeScript in favour of AffineScript, while
+> `LANGUAGE-POLICY.adoc` §1.2 stated "TypeScript is *permitted under Bun*") has
+> been ruled by the owner: **AffineScript governs.** `LANGUAGE-POLICY.adoc` §1.2
+> was the error and has been rewritten to match.
+>
+> The distinction that keeps both documents coherent: **Bun is the runtime, tier 1
+> and unchanged; AffineScript is the language for new application code.** Those
+> were run together in the withdrawn text. TypeScript is permitted only where
+> AffineScript cannot reach — the same narrow, transitional carve-out JavaScript
+> holds below.
+>
+> ReScript's ban is uncontested, and its migration destination is **AffineScript**.
 
 ### BANNED - Do Not Use
 
diff --git a/LANGUAGE-POLICY.adoc b/LANGUAGE-POLICY.adoc
index 8b4de0b9..6f4c4b6a 100644
--- a/LANGUAGE-POLICY.adoc
+++ b/LANGUAGE-POLICY.adoc
@@ -50,15 +50,49 @@ second-choice one. **The ordering above supersedes it.**
 
 === 1.2 TypeScript
 
-TypeScript is *permitted under Bun*, which executes `.ts` directly. The former
-"use ReScript instead" rule is *retired*: ReScript is itself no longer used in
-this estate, so the gate enforcing it referred to a dead alternative.
-
-CAUTION: `ts-blocker.yml` (also 55 repositories) purported to enforce that
-retired rule and was structurally incapable of failing — it diffed against
-`HEAD~1` on a depth-1 checkout, where that ref does not exist, and swallowed the
-error. Measured: *20 of 20 runs green; it has never once fired.* A gate that
-cannot fail is worse than no gate, because it is read as evidence.
+RULED 2026-08-25 by the owner. *AffineScript is the estate's language for new
+application code.* TypeScript is not.
+
+Separate the two questions this section previously ran together:
+
+Runtime:: *Bun* remains tier 1 and unchanged — see the ordering in §1. Where
+`.ts` is executed at all, it is executed by Bun.
+Language:: *AffineScript* is the target for new application code (AffineScript →
+typed-wasm). TypeScript is permitted *only where AffineScript cannot reach* —
+the same narrow carve-out JavaScript holds for MCP protocol glue and runtime
+APIs, and transitional by intent.
+
+ReScript remains banned (§3), uncontested. Its migration destination is
+*AffineScript*.
+
+==== Why the previous text was withdrawn
+
+Between 2026-07-29 and 2026-08-25 this section read "TypeScript is _permitted
+under Bun_" and declared the "use ReScript instead" rule _retired_. That put it
+in direct contradiction with `.claude/CLAUDE.md`, which bans TypeScript in
+favour of AffineScript and is *the file agents read first*. The contradiction was
+self-flagged there on 2026-08-07 and deliberately left for an owner ruling. This
+is that ruling.
+
+The reasoning that produced the retirement is worth naming, because it is a
+recurring error class rather than a one-off:
+
+CAUTION: `ts-blocker.yml` (55 repositories) purported to enforce the rule and was
+structurally incapable of failing — it diffed against `HEAD~1` on a depth-1
+checkout, where that ref does not exist, and swallowed the error. Measured: *20
+of 20 runs green; it has never once fired.* That measurement is correct and still
+stands. A gate that cannot fail is worse than no gate, because it is read as
+evidence.
+
+But that is evidence about *the gate*, not about *the rule*. A broken enforcement
+mechanism establishes that enforcement was broken; it establishes nothing about
+whether the policy it failed to enforce was right. Retiring the rule on the
+strength of the gate's failure read the instrument as the answer — the error
+class this estate documents in `AGENTS.md` §5.
+
+The correct remedy for a gate that cannot fail is to *repair or replace the
+gate*, and to verify it both ways: that it passes on clean input and fails on a
+deliberately planted violation. Not to withdraw the policy.
 
 == 2. Package management