diff --git a/nanvm-lib/src/vm/bigint/add.rs b/nanvm-lib/src/vm/bigint/add.rs
index 281d008e1..6963c7389 100644
--- a/nanvm-lib/src/vm/bigint/add.rs
+++ b/nanvm-lib/src/vm/bigint/add.rs
@@ -1,11 +1,11 @@
 use core::ops::Add;
 
-use crate::vm::{BigInt, IContainer, IVm};
+use crate::vm::{BigInt, IVm};
 
 impl<A: IVm> Add for BigInt<A> {
     type Output = Self;
     fn add(self, rhs: Self) -> Self::Output {
-        let rhs_sign = *rhs.0.header();
+        let rhs_sign = rhs.sign();
         self.add_signed(rhs, rhs_sign)
     }
 }
diff --git a/nanvm-lib/src/vm/bigint/cmp.rs b/nanvm-lib/src/vm/bigint/cmp.rs
index 6e1d87f41..74fd506d4 100644
--- a/nanvm-lib/src/vm/bigint/cmp.rs
+++ b/nanvm-lib/src/vm/bigint/cmp.rs
@@ -1,4 +1,4 @@
-use crate::vm::{IContainer, IVm, bigint::BigInt};
+use crate::vm::{IVm, bigint::BigInt};
 use std::cmp::Ordering;
 
 impl<A: IVm> PartialOrd for BigInt<A> {
@@ -11,8 +11,8 @@ impl<A: IVm> Ord for BigInt<A> {
     fn cmp(&self, rhs: &Self) -> Ordering {
         use crate::sign::Sign;
 
-        let lhs_sign = *self.0.header();
-        let rhs_sign = *rhs.0.header();
+        let lhs_sign = self.sign();
+        let rhs_sign = rhs.sign();
 
         match (lhs_sign, rhs_sign) {
             (Sign::Positive, Sign::Negative) => Ordering::Greater,
diff --git a/nanvm-lib/src/vm/bigint/debug.rs b/nanvm-lib/src/vm/bigint/debug.rs
index 71951dad7..db4ea9891 100644
--- a/nanvm-lib/src/vm/bigint/debug.rs
+++ b/nanvm-lib/src/vm/bigint/debug.rs
@@ -10,7 +10,7 @@ impl<A: IVm> Debug for BigInt<A> {
         if self.is_zero() {
             return f.write_str("0n");
         }
-        if *self.0.header() == Sign::Negative {
+        if self.sign() == Sign::Negative {
             f.write_char('-')?;
         }
         f.write_str("0x")?;
diff --git a/nanvm-lib/src/vm/bigint/mod.rs b/nanvm-lib/src/vm/bigint/mod.rs
index 7b7ba93bf..eead2ba3d 100644
--- a/nanvm-lib/src/vm/bigint/mod.rs
+++ b/nanvm-lib/src/vm/bigint/mod.rs
@@ -46,6 +46,10 @@ impl<A: IVm> BigInt<A> {
         self.0.items().is_empty()
     }
 
+    fn sign(&self) -> Sign {
+        *self.0.header()
+    }
+
     /// The function doesn't normalize the bigint.
     fn unchecked_new(sign: Sign, items: impl IntoIterator<Item = u64>) -> Self {
         Self(A::InternalBigInt::new_ok(sign, items))
@@ -75,7 +79,7 @@ impl<A: IVm> BigInt<A> {
     /// via `Neg`, because `Neg` rebuilds the container (copying every word) —
     /// don't "simplify" `sub` back to `self + (-rhs)`.
     fn add_signed(self, rhs: Self, rhs_sign: Sign) -> Self {
-        let lhs_sign = *self.0.header();
+        let lhs_sign = self.sign();
         let (sign, vec) = if lhs_sign == rhs_sign {
             (lhs_sign, self.abs_add_vec(rhs))
         } else {
diff --git a/nanvm-lib/src/vm/bigint/mul.rs b/nanvm-lib/src/vm/bigint/mul.rs
index b487dd14b..13eae3527 100644
--- a/nanvm-lib/src/vm/bigint/mul.rs
+++ b/nanvm-lib/src/vm/bigint/mul.rs
@@ -1,7 +1,7 @@
 use crate::{
     common::{sized_index::SizedIndex, vec::with_default},
     sign::Sign,
-    vm::{BigInt, IContainer, IVm},
+    vm::{BigInt, IVm},
 };
 
 use std::ops::Mul;
@@ -34,7 +34,7 @@ impl<A: IVm> Mul for BigInt<A> {
             i += 1;
         }
 
-        let sign = if self.0.header() == rhs.0.header() {
+        let sign = if self.sign() == rhs.sign() {
             Sign::Positive
         } else {
             Sign::Negative
diff --git a/nanvm-lib/src/vm/bigint/neg.rs b/nanvm-lib/src/vm/bigint/neg.rs
index a8a5bbacd..ff0e99e6a 100644
--- a/nanvm-lib/src/vm/bigint/neg.rs
+++ b/nanvm-lib/src/vm/bigint/neg.rs
@@ -2,7 +2,7 @@ use std::ops::Neg;
 
 use crate::{
     common::sized_index::SizedIndex,
-    vm::{BigInt, IContainer, IVm},
+    vm::{BigInt, IVm},
 };
 
 impl<A: IVm> Neg for BigInt<A> {
@@ -11,7 +11,7 @@ impl<A: IVm> Neg for BigInt<A> {
         if self.is_zero() {
             self
         } else {
-            Self::unchecked_new(self.0.header().flip(), self.index_iter())
+            Self::unchecked_new(self.sign().flip(), self.index_iter())
         }
     }
 }
diff --git a/nanvm-lib/src/vm/bigint/shl.rs b/nanvm-lib/src/vm/bigint/shl.rs
index 23bbac580..cc8bcfbf0 100644
--- a/nanvm-lib/src/vm/bigint/shl.rs
+++ b/nanvm-lib/src/vm/bigint/shl.rs
@@ -2,7 +2,7 @@ use core::ops::Shl;
 
 use crate::{
     common::{div_mod::DivMod, sized_index::SizedIndex},
-    vm::{Any, BigInt, IContainer, IVm},
+    vm::{Any, BigInt, IVm},
 };
 
 const TOO_LARGE: &str = "shl: shift amount too large";
@@ -58,7 +58,7 @@ impl<A: IVm> Shl for BigInt<A> {
             "shl: result must be normalized and non-empty"
         );
 
-        Ok(Self::unchecked_new(*self.0.header(), value))
+        Ok(Self::unchecked_new(self.sign(), value))
     }
 }
 
diff --git a/nanvm-lib/src/vm/bigint/shr.rs b/nanvm-lib/src/vm/bigint/shr.rs
index 997eca560..7b463cc89 100644
--- a/nanvm-lib/src/vm/bigint/shr.rs
+++ b/nanvm-lib/src/vm/bigint/shr.rs
@@ -2,7 +2,7 @@ use core::ops::Shr;
 
 use crate::{
     common::{div_mod::DivMod, sized_index::SizedIndex},
-    vm::{BigInt, IContainer, IVm},
+    vm::{BigInt, IVm},
 };
 
 impl<A: IVm> Shr for BigInt<A> {
@@ -38,7 +38,7 @@ impl<A: IVm> Shr for BigInt<A> {
             }
         }
 
-        Self::normalize_new(*self.0.header(), value)
+        Self::normalize_new(self.sign(), value)
     }
 }
 
diff --git a/nanvm-lib/src/vm/bigint/sub.rs b/nanvm-lib/src/vm/bigint/sub.rs
index 5ba84e332..2c344dc01 100644
--- a/nanvm-lib/src/vm/bigint/sub.rs
+++ b/nanvm-lib/src/vm/bigint/sub.rs
@@ -1,11 +1,11 @@
 use core::ops::Sub;
 
-use crate::vm::{BigInt, IContainer, IVm};
+use crate::vm::{BigInt, IVm};
 
 impl<A: IVm> Sub for BigInt<A> {
     type Output = Self;
     fn sub(self, rhs: Self) -> Self::Output {
-        let rhs_sign = rhs.0.header().flip();
+        let rhs_sign = rhs.sign().flip();
         self.add_signed(rhs, rhs_sign)
     }
 }
diff --git a/nanvm-lib/todo/bigint-sign-accessor.md b/nanvm-lib/todo/bigint-sign-accessor.md
deleted file mode 100644
index 0f7cb8fc5..000000000
--- a/nanvm-lib/todo/bigint-sign-accessor.md
+++ /dev/null
@@ -1,50 +0,0 @@
-## bigint-sign-accessor. `*self.0.header()` is open-coded in eight bigint files
-
-**Priority:** P4
-**Status:** open
-
-### Problem
-
-`BigInt<A>` stores its sign as the container header but exposes no accessor,
-so every operator file reaches through the private tuple field into the
-generic container — ten sites in eight files under `src/vm/bigint/`:
-
-```rust
-// mod.rs:78      let lhs_sign = *self.0.header();
-// cmp.rs:14-15   let lhs_sign = *self.0.header();  let rhs_sign = *rhs.0.header();
-// add.rs:8       let rhs_sign = *rhs.0.header();
-// sub.rs:8       let rhs_sign = rhs.0.header().flip();
-// neg.rs:14      Self::unchecked_new(self.0.header().flip(), self.index_iter())
-// mul.rs:37      if self.0.header() == rhs.0.header() {
-// shl.rs:61      Ok(Self::unchecked_new(*self.0.header(), value))
-// shr.rs:41      Self::normalize_new(*self.0.header(), value)
-// debug.rs:13    if *self.0.header() == Sign::Negative {
-```
-
-This is also why most `bigint/*.rs` files carry a `use crate::vm::IContainer`
-import they need for nothing else. The crate already has the right shape one
-directory over: `Function<A>` exposes `name()`/`length()` instead of letting
-callers read `self.0.header().0`, and `BigInt` itself already wraps one
-container access in `is_zero()` (`mod.rs:45-47`).
-
-### Proposal
-
-```rust
-// src/vm/bigint/mod.rs, next to is_zero():
-pub(crate) fn sign(&self) -> Sign {
-    *self.0.header()
-}
-```
-
-Route the ten sites through it (`rhs.sign()`, `self.sign().flip()`, …) and
-drop the now-unneeded `IContainer` imports.
-
-### Tasks
-
-- [ ] Add `sign()`; rewrite the ten sites; prune imports.
-- [ ] `cargo test`, `cargo clippy`, `cargo fmt -- --check`.
-
-### Related
-
-- [sign-algebra](./sign-algebra.md) — gives `Sign` its algebra once the sign
-  is in hand; this issue is about how the sign is *read*.