diff --git a/CHANGELOG.md b/CHANGELOG.md
index 4982794..0ff98db 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -7,6 +7,14 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.4.0] - 2026-08-10
+
+A reach release. The library itself is unchanged — no type, member or behaviour
+differs from 0.3.0, and package validation against it reports no difference —
+and what is new is where it runs: the packages now carry Android and iOS
+alongside the six desktop runtimes. That is a minor version rather than a patch
+because a platform is a capability, and nothing here is a fix.
+
 ### Added
 
 - **Android and iOS.** The packages now carry native binaries for
@@ -40,11 +48,23 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   Both are verified in CI against the packed `.nupkg` rather than the
   repository: a real Android application is built and its `.apk` opened to
   confirm `libbox3d.so` is inside it, and a real iOS application is built and
-  checked for evidence that the package handed Box3D's archive to the linker.
-  A clean build proves nothing on iOS by itself — a P/Invoke to `__Internal` is
-  resolved at run time, so an application the archive never reached builds and
-  launches like a correct one. Neither runs a simulation on a device, and the
-  platform table in the README says so rather than implying otherwise.
+  its executable read with `nm` to confirm Box3D's own native symbols are
+  defined in it — 415 of them, in the run this release was cut from. A clean
+  build proves nothing on iOS by itself: a P/Invoke to `__Internal` is resolved
+  at run time, so an application the archive never reached builds and launches
+  like a correct one.
+
+  Only symbols named `_b3…` are counted as evidence of the link. An AOT-compiled
+  method carries its parameter types in its mangled name, so `b3BodyId` and
+  `b3ShapeDef` occur inside managed symbols such as
+  `_Box3D_NET_Box3D_Body__ctor_Box3D_Native_b3BodyId`, which are in the binary
+  whether or not the archive survived; counting those would have counted the
+  wrong thing. They are counted separately instead, because they answer the
+  other question — whether the trimmer kept Box3D's C# at all — and that
+  distinction is what tells the two failures apart.
+
+  Neither platform runs a simulation on a device, and the platform table in the
+  README says so rather than implying otherwise.
 
 ### Changed
 
@@ -450,6 +470,8 @@ rather than left standing:
   follow Box3D and are documented rather than corrected.
 - Single precision only. Box3D's large-world mode changes the ABI and would need
   a separate package.
-[Unreleased]: https://github.com/Miguel249/Box3D.NET/compare/v0.2.0...HEAD
+[Unreleased]: https://github.com/Miguel249/Box3D.NET/compare/v0.4.0...HEAD
+[0.4.0]: https://github.com/Miguel249/Box3D.NET/compare/v0.3.0...v0.4.0
+[0.3.0]: https://github.com/Miguel249/Box3D.NET/compare/v0.2.0...v0.3.0
 [0.2.0]: https://github.com/Miguel249/Box3D.NET/compare/v0.1.0...v0.2.0
 [0.1.0]: https://github.com/Miguel249/Box3D.NET/releases/tag/v0.1.0
diff --git a/Directory.Build.props b/Directory.Build.props
index fe02413..3a22a93 100644
--- a/Directory.Build.props
+++ b/Directory.Build.props
@@ -71,7 +71,7 @@
 
   <!-- Package metadata. -->
   <PropertyGroup>
-    <VersionPrefix>0.3.0</VersionPrefix>
+    <VersionPrefix>0.4.0</VersionPrefix>
     <Authors>Box3D.NET contributors</Authors>
     <Company>Box3D.NET</Company>
     <Product>Box3D.NET</Product>
diff --git a/Directory.Build.targets b/Directory.Build.targets
index 87b0955..42ed680 100644
--- a/Directory.Build.targets
+++ b/Directory.Build.targets
@@ -56,10 +56,13 @@
       This was still 0.1.0 while 0.3.0 was being prepared, which is a gap rather
       than extra strictness: comparing against 0.1.0 says nothing about whether
       a member 0.2.0 added has survived, so anything introduced and then removed
-      inside the 0.2.x line would have passed unremarked.
+      inside the 0.2.x line would have passed unremarked. It was then left at
+      0.2.0 while 0.3.0 shipped, which is the same gap again, so the instruction
+      above is worth taking literally: this line moves in the commit that moves
+      VersionPrefix, not in the one after it.
     -->
     <EnablePackageValidation>true</EnablePackageValidation>
-    <PackageValidationBaselineVersion>0.2.0</PackageValidationBaselineVersion>
+    <PackageValidationBaselineVersion>0.3.0</PackageValidationBaselineVersion>
   </PropertyGroup>
 
   <ItemGroup Condition="'$(IsPackable)' == 'true'">
diff --git a/README.md b/README.md
index 3a26f54..e69453c 100644
--- a/README.md
+++ b/README.md
@@ -734,17 +734,27 @@ start an application on a phone, so those two words are the honest ceiling:
   it for `arm64-v8a`. That covers the failure mode specific to Android: an
   unresolved runtime asset produces a perfectly valid application that dies on
   its first physics call.
-- **linked** — CI builds a real iOS application against the packed `.nupkg` and
-  confirms the package handed Box3D's archive to the linker. That is the failure
-  mode specific to iOS: the library is linked into the application rather than
-  loaded, and a P/Invoke to `__Internal` is resolved by Mono at run time, so an
-  application the archive never reached builds and launches exactly like a
-  correct one and fails on its first physics call.
-
-  The stronger check — Box3D's symbols in the finished executable — is attempted
-  and normally cannot run, because a release build for iOS is stripped and
-  defines no symbols at all. Where the binary does keep a symbol table and Box3D
-  is missing from it, that is a failure rather than a shrug.
+- **linked** — CI builds a real iOS application against the packed `.nupkg`,
+  then reads the executable it produced with `nm` and requires Box3D's own
+  native symbols to be defined in it. That is the failure mode specific to iOS:
+  the library is linked into the application rather than loaded, and a P/Invoke
+  to `__Internal` is resolved by Mono at run time, so an application the archive
+  never reached builds and launches exactly like a correct one and fails on its
+  first physics call.
+
+  Only symbols named `_b3…` count, since those can only have come out of
+  `libbox3d.a`. An AOT-compiled method carries its parameter types in its
+  mangled name, so `b3BodyId` and `b3ShapeDef` turn up inside managed symbols
+  like `_Box3D_NET_Box3D_Body__ctor_Box3D_Native_b3BodyId`, which are in the
+  binary whether or not the archive survived the link. Those are counted too,
+  separately, because they answer the other question — whether the trimmer kept
+  Box3D's C# at all — and between them they say which half to go and look at
+  when the check fails.
+
+  A stripped executable would leave nothing to read, and the check falls back to
+  the build log's evidence that the linker was handed the archive, reporting
+  that it did so rather than claiming the stronger result. That has not happened
+  yet: the release simulator build keeps its symbol table.
 
 Neither runs a simulation on a device. If you ship Box3D.NET on a phone, test on
 a phone.