diff --git a/tree_hash/Cargo.toml b/tree_hash/Cargo.toml
index f3e6d35..9c23ea2 100644
--- a/tree_hash/Cargo.toml
+++ b/tree_hash/Cargo.toml
@@ -21,6 +21,7 @@ typenum = "1.12.0"
 rand = "0.8.5"
 tree_hash_derive = { path = "../tree_hash_derive", version = "0.10.0" }
 ethereum_ssz_derive = "0.9.0"
+ssz_types = "0.11.0"
 
 [features]
 arbitrary = ["alloy-primitives/arbitrary"]
diff --git a/tree_hash/src/lib.rs b/tree_hash/src/lib.rs
index e5ac66f..20f0ef8 100644
--- a/tree_hash/src/lib.rs
+++ b/tree_hash/src/lib.rs
@@ -2,6 +2,7 @@ pub mod impls;
 mod merkle_hasher;
 mod merkleize_padded;
 mod merkleize_standard;
+pub mod prototype;
 
 pub use merkle_hasher::{Error, MerkleHasher};
 pub use merkleize_padded::merkleize_padded;
diff --git a/tree_hash/src/prototype.rs b/tree_hash/src/prototype.rs
new file mode 100644
index 0000000..4f89e43
--- /dev/null
+++ b/tree_hash/src/prototype.rs
@@ -0,0 +1,90 @@
+use crate::{Hash256, TreeHash, TreeHashType, BYTES_PER_CHUNK};
+use std::marker::PhantomData;
+
+pub enum Error {
+    Oops,
+}
+
+pub trait MerkleProof: TreeHash {
+    fn compute_proof<F>(&self) -> Result<Vec<Hash256>, Error>
+    where
+        Self: Resolve<F>,
+    {
+        let gindex = <Self as Resolve<F>>::gindex(1);
+        self.compute_proof_for_gindex(gindex)
+    }
+
+    fn compute_proof_for_gindex(&self, gindex: usize) -> Result<Vec<Hash256>, Error>;
+}
+
+// A path is a sequence of field accesses like `self.foo.bar`.
+//
+// We can resolve these at compile time, because we're sickos.
+pub struct Path<First, Rest>(PhantomData<(First, Rest)>);
+
+// The field trait is implemented for all struct fields.
+//
+// It provides conversion from a named field type to a numeric field index.
+pub trait Field {
+    const NUM_FIELDS: usize;
+    const INDEX: usize;
+}
+
+// If T implements Resolve<Field> it means T has a field named Field of type Output.
+pub trait Resolve<Field> {
+    type Output;
+
+    fn gindex(parent_index: usize) -> usize;
+}
+
+// This impl defines how paths are resolved and converted to gindices.
+impl<T, First, Rest> Resolve<Path<First, Rest>> for T
+where
+    Self: Resolve<First>,
+    First: Field,
+    <Self as Resolve<First>>::Output: Resolve<Rest>,
+{
+    type Output = <<Self as Resolve<First>>::Output as Resolve<Rest>>::Output;
+
+    fn gindex(parent_index: usize) -> usize {
+        // From `get_generalized_index`:
+        // https://github.com/ethereum/consensus-specs/blob/dev/ssz/merkle-proofs.md#ssz-object-to-index
+        let new_parent_index = <Self as Resolve<First>>::gindex(parent_index);
+        <Self as Resolve<First>>::Output::gindex(new_parent_index)
+    }
+}
+
+// FIXME: we don't currently enforce I < N at compile-time
+pub struct VecIndex<const I: usize, const N: usize>;
+
+impl<const I: usize, const N: usize> Field for VecIndex<I, N> {
+    const NUM_FIELDS: usize = N;
+    const INDEX: usize = I;
+}
+
+pub fn item_length<T: TreeHash>() -> usize {
+    if T::tree_hash_type() == TreeHashType::Basic {
+        BYTES_PER_CHUNK / T::tree_hash_packing_factor()
+    } else {
+        BYTES_PER_CHUNK
+    }
+}
+
+pub fn vector_chunk_count<T: TreeHash>(length: usize) -> usize {
+    (length * item_length::<T>()).div_ceil(BYTES_PER_CHUNK)
+}
+
+pub fn get_vector_item_position<T: TreeHash>(index: usize) -> usize {
+    let start = index * item_length::<T>();
+    start / BYTES_PER_CHUNK
+}
+
+/*
+impl TreeHash for u64 {
+    fn get_field(&self, _: usize) -> Result<&dyn TreeHash, Error> {
+        Err(Error::Oops)
+    }
+
+    fn merkle_proof(&self)
+}
+*/