diff --git a/src/context.rs b/src/context.rs
index 75c2840..cf0ea86 100644
--- a/src/context.rs
+++ b/src/context.rs
@@ -24,18 +24,26 @@ pub fn set_actor_sp(v: usize)       { ACTOR_SP.with(|c| c.set(v)) }
 // ---------------------------------------------------------------------------
 // Initial stack layout
 //
-// After alignment, sp = top & ~15 - 8. Then we push (downward) six callee-
-// saved register slots and a return address. The first `switch_to_actor`
-// pops r15..rbx and `ret`s — landing in `entry` with rsp % 16 == 8.
+// We start from aligned_top = top & ~15, then bias by 8 and push seven 8-byte
+// slots (downward). The first `switch_to_actor` pops r15..rbx and `ret`s —
+// landing in `entry` with rsp % 16 == 8 (the x86-64 ABI state at the point
+// just after a `call`, which is what `entry` is compiled to expect).
 //
-// Layout (high → low), relative to aligned_top = top & ~15:
-//   aligned_top - 8  : entry ptr       ← `ret` target. Post-ret: rsp % 16 == 8.
-//   aligned_top - 16 : rbx = 0
-//   aligned_top - 24 : rbp = 0
-//   aligned_top - 32 : r12 = 0
-//   aligned_top - 40 : r13 = 0
-//   aligned_top - 48 : r14 = 0
-//   aligned_top - 56 : r15 = 0         ← initial rsp
+// Layout (high → low), relative to aligned_top = top & ~15. Note the entry
+// slot is at aligned_top - 16, NOT aligned_top - 8: the function does
+// `(top & ~15) - 8` and *then* a `-= 8` before the first write, so the first
+// stored word lands at aligned_top - 16. Verified by single-stepping the
+// `ret` under llmdbg: entry sits at an address with %16 == 0, so the post-ret
+// rsp is %16 == 8.
+//
+//   aligned_top - 8  : (unused padding; keeps entry's slot %16 == 0)
+//   aligned_top - 16 : entry ptr       ← `ret` target. Post-ret: rsp % 16 == 8.
+//   aligned_top - 24 : rbx = 0
+//   aligned_top - 32 : rbp = 0
+//   aligned_top - 40 : r12 = 0
+//   aligned_top - 48 : r13 = 0
+//   aligned_top - 56 : r14 = 0
+//   aligned_top - 64 : r15 = 0         ← initial rsp
 // ---------------------------------------------------------------------------
 
 pub fn init_actor_stack(top: *mut u8, entry: extern "C-unwind" fn()) -> usize {