diff --git a/verify/goobj_link_atomic_test.go b/verify/goobj_link_atomic_test.go
new file mode 100644
index 0000000..b838d21
--- /dev/null
+++ b/verify/goobj_link_atomic_test.go
@@ -0,0 +1,60 @@
+// Copyright (c) 2026 Petr Balvín <opensource@petrbalvin.org> (https://petrbalvin.org)
+// SPDX-License-Identifier: BSD-3-Clause
+
+package verify
+
+import (
+	"os"
+	"os/exec"
+	"testing"
+)
+
+// goobjAtomicMainSrc is the Go side of the LSE regression.  The kernel
+// drives a counter with LDADDALD and mixes the accumulated total; main
+// derives the same value in Go and panics on any divergence before
+// printing the deterministic line the baseline and the gasm-linked
+// binaries must agree on.
+const goobjAtomicMainSrc = `package main
+
+func caller(p *int64, addv int64, times int64) int64
+
+func main() {
+	addv, times := int64(3), int64(5)
+	v := int64(0)
+	got := caller(&v, addv, times)
+	oldsum, cur := int64(0), int64(0)
+	for i := int64(0); i < times; i++ {
+		oldsum += cur
+		cur += addv
+	}
+	k := uint64(0x9e3779b97f4a7c15)
+	want := int64(uint64(oldsum+cur) * k)
+	if got != want {
+		panic("atomic")
+	}
+	println("ok", caller(&v, addv, times))
+}
+`
+
+// TestGOOBJLinkAtomicKernel proves the arm64 LSE atomic families are proven
+// end to end: the kernel assembles by gasm into a GOOBJ object, substitutes
+// byte-wise into a real go build archive, relinks with cmd/link and the
+// linked binary computes the same atomic total as the toolchain-built one
+// under qemu-user, with the documented degradation on emulators without
+// LSE (the baseline run fails first and the claim drops to link-only).
+func TestGOOBJLinkAtomicKernel(t *testing.T) {
+	if testing.Short() {
+		t.Skip("builds the gasm binary and links Go programs")
+	}
+	if os.Getenv("GASM_LINK_PARITY") == "" {
+		t.Skip("deliberate verification: set GASM_LINK_PARITY=1 (just link-parity)")
+	}
+	goBin, err := exec.LookPath("go")
+	if err != nil {
+		t.Skip("no Go toolchain available")
+	}
+	gasmBin := buildLinkParityGasm(t, goBin)
+	t.Run("arm64", func(t *testing.T) {
+		goobjLinkKernel(t, goBin, gasmBin, "arm64", "linkatomic_arm64.s", goobjAtomicMainSrc)
+	})
+}
diff --git a/verify/testdata/linkatomic_arm64.s b/verify/testdata/linkatomic_arm64.s
new file mode 100644
index 0000000..88a2dfe
--- /dev/null
+++ b/verify/testdata/linkatomic_arm64.s
@@ -0,0 +1,59 @@
+// Copyright (c) 2026 Petr Balvín <opensource@petrbalvin.org> (https://petrbalvin.org)
+// SPDX-License-Identifier: BSD-3-Clause
+
+#include "textflag.h"
+
+// The link-regression kernel for the arm64 LSE atomics: atomicsum adds a
+// value to a counter with LDADDALD in a loop, accumulating every old value
+// the atomic returns, and finishes with the counter's final contents, so
+// the result is addv times times times times plus one under any schedule.
+// caller passes the pointer and the two scalars to atomicsum and mixes the
+// total through the intra-file mixa call over the ABI0 stack slots.  mixa
+// multiplies by the odd golden-ratio constant, the same arithmetic the Go
+// side of the regression mirrors bit for bit.  The run claim follows the
+// documented degradation: a qemu-user without LSE fails the baseline and
+// the claim degrades to link-only.
+
+// func atomicsum(p *int64, addv int64, times int64) int64
+TEXT ·atomicsum(SB), NOSPLIT, $0-32
+	MOVD     p+0(FP), R3
+	MOVD     addv+8(FP), R4
+	MOVD     times+16(FP), R5
+	MOVD     $0, R6
+
+loop:
+	CBZ      R5, done
+	LDADDALD R4, (R3), R7
+	ADD      R7, R6, R6
+	SUB      $1, R5, R5
+	JMP      loop
+
+done:
+	MOVD     (R3), R8
+	ADD      R8, R6, R6
+	MOVD     R6, ret+24(FP)
+	RET
+
+// func mixa(x int64) int64
+TEXT ·mixa(SB), NOSPLIT, $0-16
+	MOVD x+0(FP), R0
+	MOVD $0x9e3779b97f4a7c15, R1
+	MUL  R1, R0, R0
+	MOVD R0, ret+8(FP)
+	RET
+
+// func caller(p *int64, addv int64, times int64) int64
+TEXT ·caller(SB), NOSPLIT, $32-32
+	MOVD p+0(FP), R3
+	MOVD addv+8(FP), R4
+	MOVD times+16(FP), R5
+	MOVD R3, 8(RSP)
+	MOVD R4, 16(RSP)
+	MOVD R5, 24(RSP)
+	BL   ·atomicsum(SB)
+	MOVD 32(RSP), R0
+	MOVD R0, 8(RSP)
+	BL   ·mixa(SB)
+	MOVD 16(RSP), R0
+	MOVD R0, ret+24(FP)
+	RET