diff --git a/binrep.cabal b/binrep.cabal
index 9950649..692a1ba 100644
--- a/binrep.cabal
+++ b/binrep.cabal
@@ -34,7 +34,7 @@ flag icu
 library
   exposed-modules:
       Binrep
-      Binrep.BLen.Simple
+      Binrep.BLen
       Binrep.CBLen
       Binrep.CBLen.Generic
       Binrep.Extra.HexByteString
@@ -64,6 +64,7 @@ library
       Binrep.Via.Prim
       Data.Aeson.Extra.SizedVector
       Raehik.Compat.FlatParse.Basic.Prim
+      Raehik.Compat.FlatParse.Basic.WithLength
       Util.TypeNats
   other-modules:
       Paths_binrep
diff --git a/flake.lock b/flake.lock
index 636ce2c..22ddcc3 100644
--- a/flake.lock
+++ b/flake.lock
@@ -3,11 +3,11 @@
     "bytezap": {
       "flake": false,
       "locked": {
-        "lastModified": 1710555689,
-        "narHash": "sha256-G1IFB1HAmxwySLhkU/bkJsOexgYDcLFXFGM2sG8i+7o=",
+        "lastModified": 1710619263,
+        "narHash": "sha256-Mz1w2wKhN/WagMZQno9g0yNGzL/j7t0gxmiju6g37pc=",
         "owner": "raehik",
         "repo": "bytezap",
-        "rev": "08ced1055d8bc52c703a96899695496c8f3511cb",
+        "rev": "8bb9222e16a2a86135326a88c613e0be17c2916b",
         "type": "github"
       },
       "original": {
diff --git a/src/Binrep.hs b/src/Binrep.hs
index 00f1a38..8d2d256 100644
--- a/src/Binrep.hs
+++ b/src/Binrep.hs
@@ -1,12 +1,12 @@
 module Binrep
-  ( module Binrep.CBLen
-  , module Binrep.BLen.Simple
+  ( module Binrep.BLen
+  , module Binrep.CBLen
   , module Binrep.Put
   , module Binrep.Get
   ) where
 
+import Binrep.BLen
 import Binrep.CBLen
-import Binrep.BLen.Simple
 import Binrep.Put
 import Binrep.Get
 
diff --git a/src/Binrep/BLen/Simple.hs b/src/Binrep/BLen.hs
similarity index 59%
rename from src/Binrep/BLen/Simple.hs
rename to src/Binrep/BLen.hs
index dc75008..4da06f4 100644
--- a/src/Binrep/BLen/Simple.hs
+++ b/src/Binrep/BLen.hs
@@ -10,9 +10,18 @@ byte length *before* serializing. This is that.
 It should be very efficient to calculate serialized byte length for most
 binrep-compatible Haskell types. If it isn't, consider whether the
 representation is appropriate for binrep.
+
+Note that you _may_ encode this inside the serializer type (whatever the @Put@
+class stores). I went back and forth on this a couple times. But some binrep
+code seems to make more sense when byte length is standalone. And I don't mind
+the extra explicitness. So it's here to stay :)
 -}
 
-module Binrep.BLen.Simple where
+module Binrep.BLen
+  ( BLen(blen)
+  , blenGenericNonSum, blenGenericSum
+  , ViaCBLen(..), cblen
+  ) where
 
 import Binrep.CBLen
 import GHC.TypeNats
@@ -32,7 +41,14 @@ import Generic.Data.Function.FoldMap
 import Generic.Data.Rep.Assert
 import Generic.Data.Function.Common
 
-class BLen a where blen :: a -> Int
+-- | Class for types with easily-calculated length in bytes.
+--
+-- If it appears hard to calculate byte length for a given type (e.g. without
+-- first serializing it, then measuring serialized byte length), consider
+-- whether this type is a good fit for binrep.
+class BLen a where
+    -- | Calculate the serialized byte length of the given value.
+    blen :: a -> Int
 
 -- newtype sum monoid for generic foldMap
 newtype BLen' a = BLen' { getBLen' :: a }
@@ -67,49 +83,38 @@ blenGenericSum f = getBLen' . genericFoldMapSum @'SumOnly @asserts (BLen' <$> f)
 instance TypeError ENoEmpty => BLen Void where blen = undefined
 instance TypeError ENoSum => BLen (Either a b) where blen = undefined
 
--- | Unit type has length 0.
-instance BLen () where
-    {-# INLINE blen #-}
-    blen () = 0
+-- | _O(1)_ Unit type has length 0.
+instance BLen () where blen () = 0
 
--- | Sum tuples.
-instance (BLen l, BLen r) => BLen (l, r) where
-    {-# INLINE blen #-}
-    blen (l, r) = blen l + blen r
+-- | _O(1)_ Sum tuples.
+instance (BLen l, BLen r) => BLen (l, r) where blen (l, r) = blen l + blen r
 
 -- | _O(n)_ Sum the length of each element of a list.
-instance BLen a => BLen [a] where
-    {-# INLINE blen #-}
-    blen = sum . map blen
-
--- | Length of a bytestring is fairly obvious.
-instance BLen B.ByteString where
-    {-# INLINE blen #-}
-    blen = B.length
-
--- Machine integers have a constant byte length.
-deriving via CBLenly Word8  instance BLen Word8
-deriving via CBLenly  Int8  instance BLen  Int8
-deriving via CBLenly Word16 instance BLen Word16
-deriving via CBLenly  Int16 instance BLen  Int16
-deriving via CBLenly Word32 instance BLen Word32
-deriving via CBLenly  Int32 instance BLen  Int32
-deriving via CBLenly Word64 instance BLen Word64
-deriving via CBLenly  Int64 instance BLen  Int64
+instance BLen a => BLen [a] where blen = sum . map blen
+
+-- | _O(1)_ 'B.ByteString's store their own length.
+instance BLen B.ByteString where blen = B.length
+
+-- All words have a constant byte length-- including host-size words, mind you!
+deriving via ViaCBLen Word8  instance BLen Word8
+deriving via ViaCBLen  Int8  instance BLen  Int8
+deriving via ViaCBLen Word16 instance BLen Word16
+deriving via ViaCBLen  Int16 instance BLen  Int16
+deriving via ViaCBLen Word32 instance BLen Word32
+deriving via ViaCBLen  Int32 instance BLen  Int32
+deriving via ViaCBLen Word64 instance BLen Word64
+deriving via ViaCBLen  Int64 instance BLen  Int64
 
 --------------------------------------------------------------------------------
 
--- | Deriving via wrapper for types which may derive a 'BLen' instance through
---   an existing 'IsCBLen' instance.
+-- | DerivingVia wrapper for types which may derive a 'BLen' instance through
+--   an existing 'IsCBLen' instance (i.e. it is known at compile time)
 --
 -- Examples of such types include machine integers, and explicitly-sized types
 -- (e.g. "Binrep.Type.Sized").
-newtype CBLenly a = CBLenly { unCBLenly :: a }
-instance KnownNat (CBLen a) => BLen (CBLenly a) where
-    {-# INLINE blen #-}
-    blen _ = cblen @a
+newtype ViaCBLen a = ViaCBLen { unCBLenly :: a }
+instance KnownNat (CBLen a) => BLen (ViaCBLen a) where blen _ = cblen @a
 
 -- | Reify a type's constant byte length to the term level.
 cblen :: forall a n. (n ~ CBLen a, KnownNat n) => Int
 cblen = natValInt @n
-{-# INLINE cblen #-}
diff --git a/src/Binrep/CBLen/Generic.hs b/src/Binrep/CBLen/Generic.hs
index 8d2c063..299fbef 100644
--- a/src/Binrep/CBLen/Generic.hs
+++ b/src/Binrep/CBLen/Generic.hs
@@ -14,7 +14,7 @@ generics excel, and the one this module targets.
 You can (attempt to) derive a 'CBLen' type family instance generically for a
 type via
 
-    instance BLen a where type CBLen a = CBLenGeneric w a
+    instance CBLen a where type CBLen a = CBLenGeneric w a
 
 As with deriving @BLen@ generically, you must provide the type used to store the
 sum tag for sum types.
diff --git a/src/Binrep/Put.hs b/src/Binrep/Put.hs
index b64b5bd..3a9aadc 100644
--- a/src/Binrep/Put.hs
+++ b/src/Binrep/Put.hs
@@ -1,20 +1,11 @@
 {-# LANGUAGE UndecidableInstances #-} -- required below GHC 9.6
 {-# OPTIONS_GHC -fno-warn-orphans #-} -- for generic data op instance
 
-{- | Serialization using the bytezap library.
-
-bytezap serializers ("pokes") work by writing bytes into a pointer, which is
-assumed to have _precisely_ the space required. The user must determine the
-post-serialize length before the fact. For that reason, this module requires
-that types to be serialized have a 'BLen' instance. In general, we are happy
-about this, because a binrep type should always have an efficient and preferably
-simple 'BLen' instance (and if not, it shouldn't be a binrep type).
--}
-
 module Binrep.Put where
 
+import Binrep.BLen ( BLen(blen) )
 import Data.Functor.Identity
-import Bytezap.Write
+import Bytezap.Poke
 import Raehik.Compat.Data.Primitive.Types ( Prim' )
 import Binrep.Util.ByteOrder
 import Raehik.Compat.Data.Primitive.Types.Endian ( ByteSwap )
@@ -36,23 +27,23 @@ import Generic.Data.Rep.Assert
 
 import Control.Monad.ST ( RealWorld )
 
-type Write' = Write RealWorld
+type Putter = Poke RealWorld
 
-class Put a where put :: a -> Write'
+class Put a where put :: a -> Putter
 
-runPut :: Put a => a -> B.ByteString
-runPut = runWriteBS . put
+runPut :: (BLen a, Put a) => a -> B.ByteString
+runPut a = unsafeRunPokeBS (blen a) (put a)
 
-instance GenericFoldMap Write' where
-    type GenericFoldMapC Write' a = Put a
+instance GenericFoldMap Putter where
+    type GenericFoldMapC Putter a = Put a
     genericFoldMapF = put
 
 -- | Serialize a term of the non-sum type @a@ via its 'Generic' instance.
 putGenericNonSum
     :: forall {cd} {f} {asserts} a
-    .  ( Generic a, Rep a ~ D1 cd f, GFoldMapNonSum Write' f
+    .  ( Generic a, Rep a ~ D1 cd f, GFoldMapNonSum Putter f
        , asserts ~ '[ 'NoEmpty, 'NoSum], ApplyGCAsserts asserts f)
-    => a -> Write'
+    => a -> Putter
 putGenericNonSum = genericFoldMapNonSum @asserts
 
 -- | Serialize a term of the sum type @a@ via its 'Generic' instance.
@@ -62,9 +53,9 @@ putGenericNonSum = genericFoldMapNonSum @asserts
 -- if you want better performance!
 putGenericSum
     :: forall {cd} {f} {asserts} a
-    .  (Generic a, Rep a ~ D1 cd f, GFoldMapSum 'SumOnly Write' f
+    .  (Generic a, Rep a ~ D1 cd f, GFoldMapSum 'SumOnly Putter f
        , asserts ~ '[ 'NoEmpty, 'NeedSum], ApplyGCAsserts asserts f)
-    => (String -> Write') -> a -> Write'
+    => (String -> Putter) -> a -> Putter
 putGenericSum = genericFoldMapSum @'SumOnly @asserts
 
 instance Prim' a => Put (ViaPrim a) where put = prim . unViaPrim
@@ -74,7 +65,7 @@ instance TypeError ENoSum => Put (Either a b) where put = undefined
 
 instance Put a => Put (Identity a) where put = put . runIdentity
 
-instance Put Write' where put = id
+instance Put Putter where put = id
 
 -- | Unit type serializes to nothing. How zen.
 instance Put () where
diff --git a/src/Binrep/Type/Magic.hs b/src/Binrep/Type/Magic.hs
index 6dae103..d5d436b 100644
--- a/src/Binrep/Type/Magic.hs
+++ b/src/Binrep/Type/Magic.hs
@@ -24,10 +24,9 @@ module Binrep.Type.Magic where
 
 import Binrep
 import Binrep.Type.Byte
-import Bytezap.Write.Internal ( Write(Write) )
 import FlatParse.Basic qualified as FP
 import Data.ByteString qualified as B
-import Util.TypeNats ( type Length, natValInt )
+import Util.TypeNats ( natValInt )
 
 import GHC.TypeLits
 
@@ -36,50 +35,53 @@ import Data.Data ( Data )
 
 import Strongweak
 
--- | A singleton data type representing a "magic number" (a constant bytestring)
---   via a phantom type.
+-- | A singleton data type representing a "magic number" via a phantom type.
 --
 -- The phantom type variable unambiguously defines a constant bytestring.
 -- A handful of types are supported for using magics conveniently, e.g. for pure
 -- ASCII magics, you may use a 'Symbol' type-level string.
-data Magic (a :: k) = Magic
-    deriving stock (Generic, Data, Show, Eq)
+data Magic (a :: k) = Magic deriving stock (Generic, Data, Show, Eq)
 
--- | Weaken a 'Magic a' to the unit. Perhaps you prefer pattern matching on @()@
---   over @Magic@, or wish a weak type to be fully divorced from its binrep
---   origins.
+-- | Weaken a @'Magic' a@ to the unit.
 instance Weaken (Magic a) where
     type Weak (Magic a) = ()
     weaken Magic = ()
 
--- | Strengthen the unit to some 'Magic a'.
-instance Strengthen (Magic a) where
-    strengthen () = pure Magic
+-- | Strengthen the unit to some @'Magic' a@.
+instance Strengthen (Magic a) where strengthen () = pure Magic
 
+-- | The byte length of a magic is known at compile time.
 instance IsCBLen (Magic a) where type CBLen (Magic a) = Length (MagicBytes a)
-deriving via CBLenly (Magic a) instance
+
+-- | The byte length of a magic is obtained via reifying.
+deriving via ViaCBLen (Magic a) instance
     KnownNat (Length (MagicBytes a)) => BLen (Magic a)
 
-instance (bs ~ MagicBytes a, ReifyBytes bs, KnownNat (Length bs))
-  => Put (Magic a) where
-    put Magic = Write (natValInt @(Length bs)) (reifyBytes @bs)
+instance (bs ~ MagicBytes a, ReifyBytes bs) => Put (Magic a) where
+    put Magic = reifyBytes @bs
 
 instance (bs ~ MagicBytes a, ReifyBytes bs, KnownNat (Length bs))
   => Get (Magic a) where
-    -- TODO silly optimization: we _could_ skip comparing BS lengths because we
-    -- know they have to be the same. lmao
     get = do
         -- Nice case where we _want_ flatparse's no-copy behaviour, because
         -- 'actual' is only in scope for this parser. Except, of course, if we
         -- error, in which case _now_ we copy. Efficient!
-        actual <- FP.take (blen magic)
+        actual <- FP.take (natValInt @(Length bs))
+        -- silly optimization: we could skip comparing lengths because we know
+        -- they must be the same. very silly though
         if   actual == expected
-        then pure Magic
+        then pure magic
         else eBase $ EExpected expected (B.copy actual)
       where
         expected = runPut magic
         magic = Magic :: Magic a
 
+-- TODO might wanna move this
+-- | The length of a type-level list.
+type family Length (a :: [k]) :: Natural where
+    Length '[]       = 0
+    Length (a ': as) = 1 + Length as
+
 {-
 I do lots of functions on lists, because they're structurally simple. But you
 can't pass type-level functions as arguments between type families. singletons
@@ -115,8 +117,7 @@ class Magical (a :: k) where
     type MagicBytes a :: [Natural]
 
 -- | Type-level naturals go as-is. (Make sure you don't go over 255, though!)
-instance Magical (ns :: [Natural]) where
-    type MagicBytes ns = ns
+instance Magical (ns :: [Natural]) where type MagicBytes ns = ns
 
 -- | Type-level symbols are turned into their Unicode codepoints - but
 --   multibyte characters aren't handled, so they'll simply be overlarge bytes,
diff --git a/src/Binrep/Type/NullPadded.hs b/src/Binrep/Type/NullPadded.hs
index f294f73..0da9437 100644
--- a/src/Binrep/Type/NullPadded.hs
+++ b/src/Binrep/Type/NullPadded.hs
@@ -3,8 +3,9 @@
 module Binrep.Type.NullPadded where
 
 import Binrep
-import Bytezap.Write qualified as BZ
+import Bytezap.Poke qualified as BZ
 import FlatParse.Basic qualified as FP
+import Raehik.Compat.FlatParse.Basic.WithLength qualified as FP
 import Control.Monad.Combinators ( skipCount )
 
 import Binrep.Util ( tshow )
@@ -55,10 +56,10 @@ instance (BLen a, Put a, KnownNat n) => Put (NullPadded n a) where
         paddingLen = natValInt @n - blen a
         -- ^ refinement guarantees >=0
 
-instance (BLen a, Get a, KnownNat n) => Get (NullPadded n a) where
+instance (Get a, KnownNat n) => Get (NullPadded n a) where
     get = do
-        a <- get
-        let paddingLen = natValInt @n - blen a
+        (a, len) <- FP.parseWithLength get
+        let paddingLen = natValInt @n - len
         if   paddingLen < 0
         then eBase $ EFailNamed "TODO used to be EOverlong, cba"
         else do
diff --git a/src/Binrep/Type/Prefix/Count.hs b/src/Binrep/Type/Prefix/Count.hs
index c09a507..3908739 100644
--- a/src/Binrep/Type/Prefix/Count.hs
+++ b/src/Binrep/Type/Prefix/Count.hs
@@ -34,10 +34,12 @@ instance (KnownNat (Max pfx), Foldable f, Typeable pfx)
 instance IsCBLen (CountPrefixed pfx f a) where
     type CBLen (CountPrefixed pfx f a) = CBLen pfx + CBLen (f a)
 
+{-
 instance (Prefix pfx, Foldable f, BLen pfx, BLen (f a))
   => BLen (CountPrefixed pfx f a) where
     blen rfa = blen (lenToPfx @pfx (Foldable.length fa)) + blen fa
       where fa = unrefine1 rfa
+-}
 
 instance (Prefix pfx, Foldable f, Put pfx, Put (f a))
   => Put (CountPrefixed pfx f a) where
diff --git a/src/Binrep/Type/Sized.hs b/src/Binrep/Type/Sized.hs
index 9192ba3..675156e 100644
--- a/src/Binrep/Type/Sized.hs
+++ b/src/Binrep/Type/Sized.hs
@@ -16,7 +16,7 @@ import Data.Typeable ( typeRep )
 import GHC.TypeNats
 import Util.TypeNats ( natValInt )
 
--- | Essentially reflects a 'BLen' type to 'CBLen'.
+-- | Essentially runtime reflection of a 'BLen' type to 'CBLen'.
 data Size (n :: Natural)
 type Sized n = Refined (Size n)
 
@@ -31,13 +31,14 @@ instance (BLen a, KnownNat n) => Predicate (Size n) a where
         len = blen a
 
 instance IsCBLen (Sized n a) where type CBLen (Sized n a) = n
-deriving via CBLenly (Sized n a) instance KnownNat n => BLen (Sized n a)
+deriving via ViaCBLen (Sized n a) instance KnownNat n => BLen (Sized n a)
 
-instance Put a => Put (Sized n a) where
+instance (Put a, KnownNat n) => Put (Sized n a) where
     put = put . unrefine
 
--- TODO safety: isolate consumes all bytes if succeeds
 instance (Get a, KnownNat n) => Get (Sized n a) where
     get = do
         a <- FP.isolate (natValInt @n) get
         pure $ reallyUnsafeRefine a
+        -- ^ REFINE SAFETY: 'FP.isolate' consumes precisely the number of bytes
+        -- requested when it succeeds
diff --git a/src/Raehik/Compat/FlatParse/Basic/WithLength.hs b/src/Raehik/Compat/FlatParse/Basic/WithLength.hs
new file mode 100644
index 0000000..f1d7a73
--- /dev/null
+++ b/src/Raehik/Compat/FlatParse/Basic/WithLength.hs
@@ -0,0 +1,14 @@
+module Raehik.Compat.FlatParse.Basic.WithLength where
+
+import FlatParse.Basic.Parser
+import GHC.Exts
+
+-- | Run a parser, and return the result as well as the number of bytes it
+--   consumed.
+parseWithLength :: ParserT st e a -> ParserT st e (a, Int)
+parseWithLength (ParserT f) = ParserT $ \fp eob s st -> do
+    case f fp eob s st of
+      Fail# st'      -> Fail# st'
+      Err#  st' e    -> Err#  st' e
+      OK#   st' a s' -> OK#   st' (a, I# (s' `minusAddr#` s)) s'
+{-# inline parseWithLength #-}