diff --git a/Project.toml b/Project.toml
index ebe5308a..e4067b1f 100644
--- a/Project.toml
+++ b/Project.toml
@@ -1,7 +1,7 @@
 name = "BlockSparseArrays"
 uuid = "2c9a651f-6452-4ace-a6ac-809f4280fbb4"
 authors = ["ITensor developers <support@itensor.org> and contributors"]
-version = "0.3.5"
+version = "0.3.6"
 
 [deps]
 Adapt = "79e6a3ab-5dfb-504d-930d-738a2a938a0e"
@@ -33,7 +33,7 @@ Adapt = "4.1.1"
 Aqua = "0.8.9"
 ArrayLayouts = "1.10.4"
 BlockArrays = "1.2.0"
-DerivableInterfaces = "0.3.8"
+DerivableInterfaces = "0.4"
 DiagonalArrays = "0.3"
 Dictionaries = "0.4.3"
 FillArrays = "1.13.0"
diff --git a/src/abstractblocksparsearray/cat.jl b/src/abstractblocksparsearray/cat.jl
index 66626c9f..cdefdcf9 100644
--- a/src/abstractblocksparsearray/cat.jl
+++ b/src/abstractblocksparsearray/cat.jl
@@ -1,9 +1,6 @@
 using DerivableInterfaces: @interface, interface
+using DerivableInterfaces.Concatenate: concatenate
 
 function Base._cat(dims, as::AnyAbstractBlockSparseArray...)
-  # TODO: Call `DerivableInterfaces.cat_along(dims, as...)` instead,
-  # for better inferability. See:
-  # https://github.com/ITensor/DerivableInterfaces.jl/pull/13
-  # https://github.com/ITensor/DerivableInterfaces.jl/pull/17
-  return @interface interface(as...) cat(as...; dims)
+  return concatenate(dims, as...)
 end
diff --git a/src/abstractblocksparsearray/wrappedabstractblocksparsearray.jl b/src/abstractblocksparsearray/wrappedabstractblocksparsearray.jl
index 29d38e51..0e577f17 100644
--- a/src/abstractblocksparsearray/wrappedabstractblocksparsearray.jl
+++ b/src/abstractblocksparsearray/wrappedabstractblocksparsearray.jl
@@ -1,5 +1,5 @@
 using Adapt: Adapt, WrappedArray, adapt
-using ArrayLayouts: zero!
+using ArrayLayouts: ArrayLayouts
 using BlockArrays:
   BlockArrays,
   AbstractBlockVector,
@@ -9,7 +9,7 @@ using BlockArrays:
   blockedrange,
   mortar,
   unblock
-using DerivableInterfaces: DerivableInterfaces, @interface, DefaultArrayInterface
+using DerivableInterfaces: DerivableInterfaces, @interface, DefaultArrayInterface, zero!
 using GPUArraysCore: @allowscalar
 using SplitApplyCombine: groupcount
 using TypeParameterAccessors: similartype
@@ -154,9 +154,8 @@ function Base.setindex!(a::AnyAbstractBlockSparseArray{<:Any,1}, value, I::Block
   return a
 end
 
-# TODO: Use `@derive`.
 function ArrayLayouts.zero!(a::AnyAbstractBlockSparseArray)
-  return @interface interface(a) zero!(a)
+  return zero!(a)
 end
 
 # TODO: Use `@derive`.
diff --git a/src/blocksparsearrayinterface/blocksparsearrayinterface.jl b/src/blocksparsearrayinterface/blocksparsearrayinterface.jl
index 17c54c7f..e258af84 100644
--- a/src/blocksparsearrayinterface/blocksparsearrayinterface.jl
+++ b/src/blocksparsearrayinterface/blocksparsearrayinterface.jl
@@ -1,4 +1,4 @@
-using ArrayLayouts: ArrayLayouts, zero!
+using ArrayLayouts: ArrayLayouts
 using BlockArrays:
   BlockArrays,
   AbstractBlockVector,
@@ -16,7 +16,7 @@ using BlockArrays:
   blocklength,
   blocks,
   findblockindex
-using DerivableInterfaces: DerivableInterfaces, @interface, DefaultArrayInterface
+using DerivableInterfaces: DerivableInterfaces, @interface, DefaultArrayInterface, zero!
 using LinearAlgebra: Adjoint, Transpose
 using SparseArraysBase:
   AbstractSparseArrayInterface,
@@ -266,7 +266,9 @@ end
   return a
 end
 
-@interface ::AbstractBlockSparseArrayInterface function ArrayLayouts.zero!(a::AbstractArray)
+@interface ::AbstractBlockSparseArrayInterface function DerivableInterfaces.zero!(
+  a::AbstractArray
+)
   # This will try to empty the storage if possible.
   zero!(blocks(a))
   return a
diff --git a/src/blocksparsearrayinterface/cat.jl b/src/blocksparsearrayinterface/cat.jl
index 71ec2e7f..2c0178a6 100644
--- a/src/blocksparsearrayinterface/cat.jl
+++ b/src/blocksparsearrayinterface/cat.jl
@@ -1,18 +1,14 @@
-using BlockArrays: AbstractBlockedUnitRange, blockedrange, blocklengths
-using DerivableInterfaces: DerivableInterfaces, @interface, cat!
-using SparseArraysBase: SparseArraysBase
+using BlockArrays: blocks
+using DerivableInterfaces.Concatenate: Concatenated, cat!
 
-# TODO: Maybe move to `DerivableInterfacesBlockArraysExt`.
-# TODO: Handle dual graded unit ranges, for example in a new `SparseArraysBaseGradedUnitRangesExt`.
-function DerivableInterfaces.axis_cat(
-  a1::AbstractBlockedUnitRange, a2::AbstractBlockedUnitRange
+function Base.copyto!(
+  dest::AbstractArray, concat::Concatenated{<:BlockSparseArrayInterface}
 )
-  return blockedrange(vcat(blocklengths(a1), blocklengths(a2)))
-end
-
-@interface ::AbstractBlockSparseArrayInterface function DerivableInterfaces.cat!(
-  a_dest::AbstractArray, as::AbstractArray...; dims
-)
-  cat!(blocks(a_dest), blocks.(as)...; dims)
-  return a_dest
+  # TODO: This assumes the destination blocking is commensurate with
+  # the blocking of the sources, for example because it was constructed
+  # based on the input arguments. Maybe check that explicitly.
+  # This should mostly just get called from `cat` anyway and not get
+  # called explicitly.
+  cat!(blocks(dest), blocks.(concat.args)...; dims=concat.dims)
+  return dest
 end
diff --git a/src/blocksparsearrayinterface/getunstoredblock.jl b/src/blocksparsearrayinterface/getunstoredblock.jl
index db9267e5..43b58c02 100644
--- a/src/blocksparsearrayinterface/getunstoredblock.jl
+++ b/src/blocksparsearrayinterface/getunstoredblock.jl
@@ -1,5 +1,5 @@
-using ArrayLayouts: zero!
 using BlockArrays: Block
+using DerivableInterfaces: zero!
 
 struct GetUnstoredBlock{Axes}
   axes::Axes