diff --git a/selfhost/cmd/w6c/main.combined.ww b/selfhost/cmd/w6c/main.combined.ww
index 3d3ccf87..f83e4bea 100644
--- a/selfhost/cmd/w6c/main.combined.ww
+++ b/selfhost/cmd/w6c/main.combined.ww
@@ -13426,7 +13426,9 @@ fn cgindex(c: *cgen, n: *node) void = {
 			emitline("\tMOVQ\t(BX), AX\n");
 			return;
 		};
-		if (esz == 16) {
+		// #43: str element-stride routes through primtypesize so the
+		// 16-vs-24 dispatch tracks ty_str.size for #1.
+		if (esz == primtypesize("str"): i32) {
 			emitline("\tMOVQ\t8(BX), CX\n");
 			emitline("\tMOVQ\t(BX), AX\n");
 			emitline("\tMOVQ\tCX, BX\n");
@@ -13465,9 +13467,10 @@ fn cgindex(c: *cgen, n: *node) void = {
 			emitline("\tMOVQ\t(BX), AX\n");
 			return;
 		};
-		// str element (16B): load (ptr, len) into (AX, BX) so
+		// str element (16B today): load (ptr, len) into (AX, BX) so
 		// the value flows through the str-rhs convention.
-		if (esz == 16) {
+		// #43: route via primtypesize so the stride tracks #1.
+		if (esz == primtypesize("str"): i32) {
 			emitline("\tMOVQ\t8(BX), CX\n");
 			emitline("\tMOVQ\t(BX), AX\n");
 			emitline("\tMOVQ\tCX, BX\n");
@@ -13497,7 +13500,9 @@ fn cgindex(c: *cgen, n: *node) void = {
 		emitline("\tMOVQ\t(BX), AX\n");
 		return;
 	};
-	if (esz == 16) {
+	// #43: str element-stride routes through primtypesize so the
+	// 16-vs-24 dispatch tracks ty_str.size for #1.
+	if (esz == primtypesize("str"): i32) {
 		emitline("\tMOVQ\t8(AX), BX\n");
 		emitline("\tMOVQ\t(AX), AX\n");
 		return;
@@ -16391,7 +16396,10 @@ fn cgassign(c: *cgen, n: *node) void = {
 					};
 				};
 				cgexpr(c, n.rhs);            // value → AX
-					if (esz == 16) { emitline("\tPUSHQ\tBX\n"); };
+					// #43: spill BX (str.len) before computing
+					// the index so the post-index store can pop
+					// it; the stride gate tracks ty_str.size.
+					if (esz == primtypesize("str"): i32) { emitline("\tPUSHQ\tBX\n"); };
 				emitline("\tPUSHQ\tAX\n");
 				cgexpr(c, idx);              // idx → AX
 				if (esz > 1) {
@@ -16429,7 +16437,10 @@ fn cgassign(c: *cgen, n: *node) void = {
 				emitline("\tPOPQ\tAX\n");   // scaled idx
 				emitline("\tADDQ\tAX, BX\n");
 				emitline("\tPOPQ\tAX\n");   // value
-					if (esz == 16) {
+					// #43: str-element write — pop the saved
+					// .len and store both halves. Stride gate
+					// routes through primtypesize for #1.
+					if (esz == primtypesize("str"): i32) {
 						emitline("\tMOVQ\tAX, (BX)\n");
 						emitline("\tPOPQ\tCX\n");
 						emitline("\tMOVQ\tCX, 8(BX)\n");
@@ -19001,8 +19012,13 @@ fn cglet(c: *cgen, n: *node) void = {
 				};
 			}; };
 			let shapeok: bool = false;
+			// #43: route the slice-shape size guard through SSoT.
+			// The N_TSLICE kind gate already discriminates here, so
+			// this is belt-and-suspenders, but the literal would
+			// silently miss after #1 if check.ww's astsize ever
+			// drifted from this dispatch.
 			if (scall != nil && tn != nil
-			    && tn.kind == nkind.N_TSLICE && sz == 24) {
+			    && tn.kind == nkind.N_TSLICE && sz == tyslicesize(): i32) {
 				let callee: *node = scall.lhs;
 				let a0: *node = scall.list;
 				let a1: *node = nil;
@@ -19417,13 +19433,20 @@ fn cglet(c: *cgen, n: *node) void = {
 		emitoff(off: i64);
 		emitline("(BP)\n");
 		// str init: cgexpr also leaves len in BX; store both.
-		if (sz == 16) {
+		// #43: route through primtypesize so #1's str-size bump
+		// surfaces this site (today sz==16 picks str; once str=24,
+		// the in-flight str ABI (#34) must propagate cap and the
+		// dispatch needs reshaping by kind, not raw size).
+		if (sz == primtypesize("str"): i32) {
 			emitline("\tMOVQ\tBX, ");
 			emitoff((off + 8): i64);
 			emitline("(BP)\n");
 		};
 		// slice init: ptr/len/cap in AX/BX/CX.
-		if (sz == 24) {
+		// #43: same routing — bare `sz == 24` would collide with a
+		// 24B str slot under #1; the kind disambiguation lives at
+		// the caller (cglet's tn.kind == N_TSLICE shape check).
+		if (sz == tyslicesize(): i32) {
 			emitline("\tMOVQ\tBX, ");
 			emitoff((off + 8): i64);
 			emitline("(BP)\n");
@@ -21678,7 +21701,10 @@ export fn letpreintern(c: *cgen, file: *node) void = {
 	for (d != nil) {
 		if (d.kind == nkind.N_LET) {
 			let sz: i32 = letemitsize(c, d);
-			if (sz == 16) {
+			// #43: route the str-let gate through primtypesize so
+			// #1 doesn't desync this with emitletdataw's matching
+			// `sz == primtypesize("str"): i32` strlit-init branch.
+			if (sz == primtypesize("str"): i32) {
 				let r: *node = d.rhs;
 				for (r != nil) {
 					if (r.kind != nkind.N_CAST) { break; };
diff --git a/selfhost/cmd/wcc/cgen.ww b/selfhost/cmd/wcc/cgen.ww
index 8838493a..79f97a1d 100644
--- a/selfhost/cmd/wcc/cgen.ww
+++ b/selfhost/cmd/wcc/cgen.ww
@@ -1199,7 +1199,10 @@ export fn letpreintern(c: *cgen, file: *node) void = {
 	for (d != nil) {
 		if (d.kind == nkind.N_LET) {
 			let sz: i32 = letemitsize(c, d);
-			if (sz == 16) {
+			// #43: route the str-let gate through primtypesize so
+			// #1 doesn't desync this with emitletdataw's matching
+			// `sz == primtypesize("str"): i32` strlit-init branch.
+			if (sz == primtypesize("str"): i32) {
 				let r: *node = d.rhs;
 				for (r != nil) {
 					if (r.kind != nkind.N_CAST) { break; };
diff --git a/selfhost/cmd/wcc/cgenexpr.ww b/selfhost/cmd/wcc/cgenexpr.ww
index 0db26d0a..b1ba8aaa 100644
--- a/selfhost/cmd/wcc/cgenexpr.ww
+++ b/selfhost/cmd/wcc/cgenexpr.ww
@@ -799,7 +799,9 @@ fn cgindex(c: *cgen, n: *node) void = {
 			emitline("\tMOVQ\t(BX), AX\n");
 			return;
 		};
-		if (esz == 16) {
+		// #43: str element-stride routes through primtypesize so the
+		// 16-vs-24 dispatch tracks ty_str.size for #1.
+		if (esz == primtypesize("str"): i32) {
 			emitline("\tMOVQ\t8(BX), CX\n");
 			emitline("\tMOVQ\t(BX), AX\n");
 			emitline("\tMOVQ\tCX, BX\n");
@@ -838,9 +840,10 @@ fn cgindex(c: *cgen, n: *node) void = {
 			emitline("\tMOVQ\t(BX), AX\n");
 			return;
 		};
-		// str element (16B): load (ptr, len) into (AX, BX) so
+		// str element (16B today): load (ptr, len) into (AX, BX) so
 		// the value flows through the str-rhs convention.
-		if (esz == 16) {
+		// #43: route via primtypesize so the stride tracks #1.
+		if (esz == primtypesize("str"): i32) {
 			emitline("\tMOVQ\t8(BX), CX\n");
 			emitline("\tMOVQ\t(BX), AX\n");
 			emitline("\tMOVQ\tCX, BX\n");
@@ -870,7 +873,9 @@ fn cgindex(c: *cgen, n: *node) void = {
 		emitline("\tMOVQ\t(BX), AX\n");
 		return;
 	};
-	if (esz == 16) {
+	// #43: str element-stride routes through primtypesize so the
+	// 16-vs-24 dispatch tracks ty_str.size for #1.
+	if (esz == primtypesize("str"): i32) {
 		emitline("\tMOVQ\t8(AX), BX\n");
 		emitline("\tMOVQ\t(AX), AX\n");
 		return;
@@ -3764,7 +3769,10 @@ fn cgassign(c: *cgen, n: *node) void = {
 					};
 				};
 				cgexpr(c, n.rhs);            // value → AX
-					if (esz == 16) { emitline("\tPUSHQ\tBX\n"); };
+					// #43: spill BX (str.len) before computing
+					// the index so the post-index store can pop
+					// it; the stride gate tracks ty_str.size.
+					if (esz == primtypesize("str"): i32) { emitline("\tPUSHQ\tBX\n"); };
 				emitline("\tPUSHQ\tAX\n");
 				cgexpr(c, idx);              // idx → AX
 				if (esz > 1) {
@@ -3802,7 +3810,10 @@ fn cgassign(c: *cgen, n: *node) void = {
 				emitline("\tPOPQ\tAX\n");   // scaled idx
 				emitline("\tADDQ\tAX, BX\n");
 				emitline("\tPOPQ\tAX\n");   // value
-					if (esz == 16) {
+					// #43: str-element write — pop the saved
+					// .len and store both halves. Stride gate
+					// routes through primtypesize for #1.
+					if (esz == primtypesize("str"): i32) {
 						emitline("\tMOVQ\tAX, (BX)\n");
 						emitline("\tPOPQ\tCX\n");
 						emitline("\tMOVQ\tCX, 8(BX)\n");
diff --git a/selfhost/cmd/wcc/cgenstmt.ww b/selfhost/cmd/wcc/cgenstmt.ww
index 7f53443b..3c207542 100644
--- a/selfhost/cmd/wcc/cgenstmt.ww
+++ b/selfhost/cmd/wcc/cgenstmt.ww
@@ -603,8 +603,13 @@ fn cglet(c: *cgen, n: *node) void = {
 				};
 			}; };
 			let shapeok: bool = false;
+			// #43: route the slice-shape size guard through SSoT.
+			// The N_TSLICE kind gate already discriminates here, so
+			// this is belt-and-suspenders, but the literal would
+			// silently miss after #1 if check.ww's astsize ever
+			// drifted from this dispatch.
 			if (scall != nil && tn != nil
-			    && tn.kind == nkind.N_TSLICE && sz == 24) {
+			    && tn.kind == nkind.N_TSLICE && sz == tyslicesize(): i32) {
 				let callee: *node = scall.lhs;
 				let a0: *node = scall.list;
 				let a1: *node = nil;
@@ -1019,13 +1024,20 @@ fn cglet(c: *cgen, n: *node) void = {
 		emitoff(off: i64);
 		emitline("(BP)\n");
 		// str init: cgexpr also leaves len in BX; store both.
-		if (sz == 16) {
+		// #43: route through primtypesize so #1's str-size bump
+		// surfaces this site (today sz==16 picks str; once str=24,
+		// the in-flight str ABI (#34) must propagate cap and the
+		// dispatch needs reshaping by kind, not raw size).
+		if (sz == primtypesize("str"): i32) {
 			emitline("\tMOVQ\tBX, ");
 			emitoff((off + 8): i64);
 			emitline("(BP)\n");
 		};
 		// slice init: ptr/len/cap in AX/BX/CX.
-		if (sz == 24) {
+		// #43: same routing — bare `sz == 24` would collide with a
+		// 24B str slot under #1; the kind disambiguation lives at
+		// the caller (cglet's tn.kind == N_TSLICE shape check).
+		if (sz == tyslicesize(): i32) {
 			emitline("\tMOVQ\tBX, ");
 			emitoff((off + 8): i64);
 			emitline("(BP)\n");
diff --git a/selfhost/cmd/wwdump/main.combined.ww b/selfhost/cmd/wwdump/main.combined.ww
index b7bd0c72..88f6afcc 100644
--- a/selfhost/cmd/wwdump/main.combined.ww
+++ b/selfhost/cmd/wwdump/main.combined.ww
@@ -13426,7 +13426,9 @@ fn cgindex(c: *cgen, n: *node) void = {
 			emitline("\tMOVQ\t(BX), AX\n");
 			return;
 		};
-		if (esz == 16) {
+		// #43: str element-stride routes through primtypesize so the
+		// 16-vs-24 dispatch tracks ty_str.size for #1.
+		if (esz == primtypesize("str"): i32) {
 			emitline("\tMOVQ\t8(BX), CX\n");
 			emitline("\tMOVQ\t(BX), AX\n");
 			emitline("\tMOVQ\tCX, BX\n");
@@ -13465,9 +13467,10 @@ fn cgindex(c: *cgen, n: *node) void = {
 			emitline("\tMOVQ\t(BX), AX\n");
 			return;
 		};
-		// str element (16B): load (ptr, len) into (AX, BX) so
+		// str element (16B today): load (ptr, len) into (AX, BX) so
 		// the value flows through the str-rhs convention.
-		if (esz == 16) {
+		// #43: route via primtypesize so the stride tracks #1.
+		if (esz == primtypesize("str"): i32) {
 			emitline("\tMOVQ\t8(BX), CX\n");
 			emitline("\tMOVQ\t(BX), AX\n");
 			emitline("\tMOVQ\tCX, BX\n");
@@ -13497,7 +13500,9 @@ fn cgindex(c: *cgen, n: *node) void = {
 		emitline("\tMOVQ\t(BX), AX\n");
 		return;
 	};
-	if (esz == 16) {
+	// #43: str element-stride routes through primtypesize so the
+	// 16-vs-24 dispatch tracks ty_str.size for #1.
+	if (esz == primtypesize("str"): i32) {
 		emitline("\tMOVQ\t8(AX), BX\n");
 		emitline("\tMOVQ\t(AX), AX\n");
 		return;
@@ -16391,7 +16396,10 @@ fn cgassign(c: *cgen, n: *node) void = {
 					};
 				};
 				cgexpr(c, n.rhs);            // value → AX
-					if (esz == 16) { emitline("\tPUSHQ\tBX\n"); };
+					// #43: spill BX (str.len) before computing
+					// the index so the post-index store can pop
+					// it; the stride gate tracks ty_str.size.
+					if (esz == primtypesize("str"): i32) { emitline("\tPUSHQ\tBX\n"); };
 				emitline("\tPUSHQ\tAX\n");
 				cgexpr(c, idx);              // idx → AX
 				if (esz > 1) {
@@ -16429,7 +16437,10 @@ fn cgassign(c: *cgen, n: *node) void = {
 				emitline("\tPOPQ\tAX\n");   // scaled idx
 				emitline("\tADDQ\tAX, BX\n");
 				emitline("\tPOPQ\tAX\n");   // value
-					if (esz == 16) {
+					// #43: str-element write — pop the saved
+					// .len and store both halves. Stride gate
+					// routes through primtypesize for #1.
+					if (esz == primtypesize("str"): i32) {
 						emitline("\tMOVQ\tAX, (BX)\n");
 						emitline("\tPOPQ\tCX\n");
 						emitline("\tMOVQ\tCX, 8(BX)\n");
@@ -19001,8 +19012,13 @@ fn cglet(c: *cgen, n: *node) void = {
 				};
 			}; };
 			let shapeok: bool = false;
+			// #43: route the slice-shape size guard through SSoT.
+			// The N_TSLICE kind gate already discriminates here, so
+			// this is belt-and-suspenders, but the literal would
+			// silently miss after #1 if check.ww's astsize ever
+			// drifted from this dispatch.
 			if (scall != nil && tn != nil
-			    && tn.kind == nkind.N_TSLICE && sz == 24) {
+			    && tn.kind == nkind.N_TSLICE && sz == tyslicesize(): i32) {
 				let callee: *node = scall.lhs;
 				let a0: *node = scall.list;
 				let a1: *node = nil;
@@ -19417,13 +19433,20 @@ fn cglet(c: *cgen, n: *node) void = {
 		emitoff(off: i64);
 		emitline("(BP)\n");
 		// str init: cgexpr also leaves len in BX; store both.
-		if (sz == 16) {
+		// #43: route through primtypesize so #1's str-size bump
+		// surfaces this site (today sz==16 picks str; once str=24,
+		// the in-flight str ABI (#34) must propagate cap and the
+		// dispatch needs reshaping by kind, not raw size).
+		if (sz == primtypesize("str"): i32) {
 			emitline("\tMOVQ\tBX, ");
 			emitoff((off + 8): i64);
 			emitline("(BP)\n");
 		};
 		// slice init: ptr/len/cap in AX/BX/CX.
-		if (sz == 24) {
+		// #43: same routing — bare `sz == 24` would collide with a
+		// 24B str slot under #1; the kind disambiguation lives at
+		// the caller (cglet's tn.kind == N_TSLICE shape check).
+		if (sz == tyslicesize(): i32) {
 			emitline("\tMOVQ\tBX, ");
 			emitoff((off + 8): i64);
 			emitline("(BP)\n");
@@ -21678,7 +21701,10 @@ export fn letpreintern(c: *cgen, file: *node) void = {
 	for (d != nil) {
 		if (d.kind == nkind.N_LET) {
 			let sz: i32 = letemitsize(c, d);
-			if (sz == 16) {
+			// #43: route the str-let gate through primtypesize so
+			// #1 doesn't desync this with emitletdataw's matching
+			// `sz == primtypesize("str"): i32` strlit-init branch.
+			if (sz == primtypesize("str"): i32) {
 				let r: *node = d.rhs;
 				for (r != nil) {
 					if (r.kind != nkind.N_CAST) { break; };