diff --git a/bench/Main.hs b/bench/Main.hs
index a77a65e..28ef227 100644
--- a/bench/Main.hs
+++ b/bench/Main.hs
@@ -2,7 +2,7 @@ import Gauge (bgroup,bench,whnf)
 import Gauge.Main (defaultMain)
 import Data.Word (Word64)
 import Data.Primitive (ByteArray)
-import qualified Data.ByteArray.Builder.Small.Unsafe as U
+import qualified Data.ByteArray.Builder.Bounded as U
 
 import qualified HexWord64
 
diff --git a/cabal.project b/cabal.project
new file mode 100644
index 0000000..079db62
--- /dev/null
+++ b/cabal.project
@@ -0,0 +1,5 @@
+packages: .
+source-repository-package
+  type: git
+  location: https://github.com/andrewthad/natural-arithmetic
+  tag: 68868c96b58ddaf71bb865b247d2c14c3668f4c2
diff --git a/common/HexWord64.hs b/common/HexWord64.hs
index 68ea963..65c5dfa 100644
--- a/common/HexWord64.hs
+++ b/common/HexWord64.hs
@@ -17,7 +17,7 @@ module HexWord64
 
 import GHC.ST (ST(ST))
 import Data.Bits
-import Data.ByteArray.Builder.Small.Unsafe (Builder,construct)
+import Data.ByteArray.Builder.Bounded.Unsafe (Builder,construct)
 import Data.Primitive
 import Data.Word
 import GHC.Exts
diff --git a/small-bytearray-builder.cabal b/small-bytearray-builder.cabal
index f5e511f..9d613bd 100644
--- a/small-bytearray-builder.cabal
+++ b/small-bytearray-builder.cabal
@@ -36,8 +36,10 @@ flag checked
 
 library
   exposed-modules:
-    Data.ByteArray.Builder.Small
-    Data.ByteArray.Builder.Small.Bounded
+    Data.ByteArray.Builder
+    Data.ByteArray.Builder.Unsafe
+    Data.ByteArray.Builder.Bounded
+    Data.ByteArray.Builder.Bounded.Unsafe
   build-depends:
     , base >=4.12.0.0 && <5
     , byteslice >=0.1 && <0.2
@@ -46,6 +48,7 @@ library
     , vector >=0.12.0.3 && <0.13
     , bytestring >=0.10.8.2 && <0.11
     , text-short >=0.1.3 && <0.2
+    , natural-arithmetic >=0.1 && <0.2
   if flag(checked)
     build-depends: primitive-checked >= 0.7 && <0.8
   else
@@ -59,6 +62,7 @@ test-suite test
   type: exitcode-stdio-1.0
   hs-source-dirs: test, common
   main-is: Main.hs
+  ghc-options: -O2 -Wall
   other-modules:
     HexWord64
   build-depends:
@@ -70,6 +74,7 @@ test-suite test
     , tasty-quickcheck >=0.10.1 && <0.11
     , tasty-hunit >=0.10.0.2 && <0.11
     , tasty >=1.2.3 && <1.3
+    , text >=1.2 && <1.3
     , primitive
     , vector
 
diff --git a/src/Data/ByteArray/Builder/Small.hs b/src/Data/ByteArray/Builder.hs
similarity index 93%
rename from src/Data/ByteArray/Builder/Small.hs
rename to src/Data/ByteArray/Builder.hs
index a9cfabe..ce49b26 100644
--- a/src/Data/ByteArray/Builder/Small.hs
+++ b/src/Data/ByteArray/Builder.hs
@@ -6,7 +6,7 @@
 {-# language ScopedTypeVariables #-}
 {-# language UnboxedTuples #-}
 
-module Data.ByteArray.Builder.Small
+module Data.ByteArray.Builder
   ( -- * Bounded Primitives
     Builder(..)
   , construct
@@ -24,6 +24,8 @@ module Data.ByteArray.Builder.Small
   , bytearray
   , shortTextUtf8
   , shortTextJsonString
+  , cstring
+  , stringUtf8
     -- * Encode Integral Types
     -- ** Human-Readable
   , word64Dec
@@ -60,32 +62,15 @@ import Data.Primitive.ByteArray.Offset (MutableByteArrayOffset(..))
 import Data.ByteString.Short.Internal (ShortByteString(SBS))
 import Data.Text.Short (ShortText)
 import Data.Char (ord)
+import Data.ByteArray.Builder.Unsafe (Builder(Builder))
+import Data.ByteArray.Builder.Unsafe (stringUtf8,cstring)
 
 import qualified GHC.Exts as Exts
 import qualified Data.Text.Short as TS
 import qualified Data.Primitive as PM
 import qualified Data.Vector as V
-import qualified Data.ByteArray.Builder.Small.Bounded as Bounded
-
--- | An unmaterialized sequence of bytes that may be pasted
--- into a mutable byte array.
-newtype Builder = Builder
-  -- This functions takes an offset and a number of remaining bytes
-  -- and returns the new offset.
-  (forall s. MutableByteArray# s -> Int# -> Int# -> State# s -> (# State# s, Int# #))
-
-instance Semigroup Builder where
-  {-# inline (<>) #-}
-  Builder f <> Builder g = Builder $ \arr off0 len0 s0 -> case f arr off0 len0 s0 of
-    (# s1, r #) -> case r /=# (-1#) of
-      1# -> g arr r (len0 +# (off0 -# r)) s1
-      _ -> (# s1, (-1#) #)
-
-instance Monoid Builder where
-  mempty = Builder $ \_ off0 _ s0 -> (# s0, off0 #)
-
-instance IsString Builder where
-  fromString = shortTextUtf8 . TS.fromString
+import qualified Data.ByteArray.Builder.Bounded as Bounded
+import qualified Data.ByteArray.Builder.Bounded.Unsafe as UnsafeBounded
 
 -- | Run a builder. An accurate size hint is important for good performance.
 -- The size hint should be slightly larger than the actual size.
@@ -101,7 +86,7 @@ run hint b = runByteArrayST $ do
           Just len -> do
             shrinkMutableByteArray arr len
             unsafeFreezeByteArray arr
-  go hint
+  go (max hint 1)
 
 -- | Variant of 'pasteArrayST' that runs in 'IO'.
 pasteArrayIO ::
@@ -186,9 +171,10 @@ construct f = Builder
         Nothing -> (# s1, (-1#) #)
         Just (I# n) -> (# s1, n #)
 
+-- | Convert a bounded builder to an unbounded one.
 fromBounded :: forall n. KnownNat n => Bounded.Builder n -> Builder
 {-# inline fromBounded #-}
-fromBounded (Bounded.Builder f) = Builder $ \arr off len s0 ->
+fromBounded (UnsafeBounded.Builder f) = Builder $ \arr off len s0 ->
   case fromIntegral (natVal' (proxy# :: Proxy# n)) of
     I# req -> case len >=# req of
       1# -> f arr off s0
@@ -226,7 +212,7 @@ slicedUtf8TextJson !src# !soff0# !slen0# = construct $ \(MutableBytes dst doff0
                     then PM.writeByteArray dst doff (c2w c) *> go (soff + 1) (slen - 1) (doff + 1)
                     else do
                       write2 dst doff '\\' 'u'
-                      doff' <- Bounded.pasteST
+                      doff' <- UnsafeBounded.pasteST
                         (Bounded.word16PaddedUpperHex (fromIntegral (c2w c)))
                         dst (doff + 2)
                       go (soff + 1) (slen - 1) doff'
diff --git a/src/Data/ByteArray/Builder/Small/Bounded.hs b/src/Data/ByteArray/Builder/Bounded.hs
similarity index 82%
rename from src/Data/ByteArray/Builder/Small/Bounded.hs
rename to src/Data/ByteArray/Builder/Bounded.hs
index 892b36d..5db8a12 100644
--- a/src/Data/ByteArray/Builder/Small/Bounded.hs
+++ b/src/Data/ByteArray/Builder/Bounded.hs
@@ -1,4 +1,3 @@
-{-# language GADTs #-}
 {-# language KindSignatures #-}
 {-# language ScopedTypeVariables #-}
 {-# language BangPatterns #-}
@@ -12,23 +11,18 @@
 {-# language TypeApplications #-}
 
 -- | The functions in this module are explict in the amount of bytes they require.
-module Data.ByteArray.Builder.Small.Bounded
+module Data.ByteArray.Builder.Bounded
   ( -- * Builder
-    Builder(..)
-  , construct
+    Builder
     -- * Execute
   , run
-  , pasteST
   , pasteGrowST
-  , pasteIO
     -- * Combine
   , empty
   , append
     -- * Bounds Manipulation
-  , (<=)
-  , lessThanEqual
-  , isLessThanEqual
-  , raise
+  , weaken
+  , substitute
     -- * Encode Integral Types
     -- ** Human-Readable
   , word64Dec
@@ -56,46 +50,33 @@ import Data.Bits
 import Data.Char (ord)
 import Data.Primitive
 import GHC.Exts
-import GHC.ST
-import GHC.Word
-import GHC.Int
-import Data.Kind
-import GHC.TypeLits (KnownNat,Nat,type (+),natVal')
-import qualified GHC.TypeLits as GHC
+import GHC.ST (ST(ST))
+import GHC.Word (Word8(W8#),Word16(W16#),Word32(W32#),Word64(W64#))
+import GHC.Int (Int64(I64#))
+import GHC.TypeLits (KnownNat,type (+),natVal')
 import Data.Primitive.ByteArray.Offset (MutableByteArrayOffset(..))
-import qualified Control.Category as Cat
+import Arithmetic.Types (type (<=), type (:=:))
+import Data.ByteArray.Builder.Bounded.Unsafe (Builder(..))
 
+import qualified Data.ByteArray.Builder.Bounded.Unsafe as Unsafe
 import qualified Data.Primitive as PM
 
--- | A builder parameterized by the maximum number of bytes it uses
--- when executed.
-newtype Builder :: Nat -> Type where
-   Builder ::
-        (forall s. MutableByteArray# s -> Int# -> State# s -> (# State# s, Int# #))
-     -> Builder n
-
+-- Used internally.
 knownNat :: KnownNat n => Proxy# n -> Int
+{-# inline knownNat #-}
 knownNat p = fromIntegral (natVal' p)
 
--- | Execute the builder. This function is safe.
+-- | Execute the bounded builder.
 run :: forall n. KnownNat n
   => Builder n -- ^ Builder
   -> ByteArray
 {-# inline run #-}
 run b = runST $ do
   arr <- newByteArray (knownNat (proxy# :: Proxy# n))
-  len <- pasteST b arr 0
+  len <- Unsafe.pasteST b arr 0
   shrinkMutableByteArray arr len
   unsafeFreezeByteArray arr
 
--- | This function does not enforce the known upper bound on the
--- size. It is up to the user to do this.
-pasteST :: Builder n -> MutableByteArray s -> Int -> ST s Int
-{-# inline pasteST #-}
-pasteST (Builder f) (MutableByteArray arr) (I# off) =
-  ST $ \s0 -> case f arr off s0 of
-    (# s1, r #) -> (# s1, (I# r) #)
-
 -- | Paste the builder into the byte array starting at offset zero.
 -- This reallocates the byte array if it cannot accomodate the builder,
 -- growing it by the minimum amount necessary.
@@ -112,67 +93,40 @@ pasteGrowST b !(MutableByteArrayOffset{array=arr0,offset=off0}) = do
   let sz1 = off0 + req
   if sz1 <= sz0
     then do
-      off1 <- pasteST b arr0 off0
+      off1 <- Unsafe.pasteST b arr0 off0
       pure (MutableByteArrayOffset arr0 off1)
     else do
       arr1 <- PM.resizeMutableByteArray arr0 sz1
-      off1 <- pasteST b arr1 off0
+      off1 <- Unsafe.pasteST b arr1 off0
       pure (MutableByteArrayOffset arr1 off1)
 
--- | This function does not enforce the known upper bound on the
--- size. It is up to the user to do this.
-pasteIO :: Builder n -> MutableByteArray RealWorld -> Int -> IO Int
-{-# inline pasteIO #-}
-pasteIO b m off = stToIO (pasteST b m off)
-
--- | Constructor for 'Builder' that works on a function with lifted
--- arguments instead of unlifted ones. This is just as unsafe as the
--- actual constructor.
-construct :: (forall s. MutableByteArray s -> Int -> ST s Int) -> Builder n
-{-# inline construct #-}
-construct f = Builder
-  $ \arr off s0 ->
-    case unST (f (MutableByteArray arr) (I# off)) s0 of
-      (# s1, (I# n) #) -> (# s1, n #)
-
-infixr 9 `append`
-
 -- | The monoidal unit of `append`
 empty :: Builder 0
 empty = Builder $ \_ off0 s0 -> (# s0, off0 #)
 
+infixr 9 `append`
+
 -- | Concatenate two builders.
-append :: Builder n -> Builder m -> Builder (n + m)
+append :: Builder m -> Builder n -> Builder (m + n)
 append (Builder f) (Builder g) =
   Builder $ \arr off0 s0 -> case f arr off0 s0 of
     (# s1, r #) -> g arr r s1
 
--- | A proof that n is less than or equal to m
-newtype (n :: Nat) <= (m :: Nat) = LessThanEqual Int -- m - n
+-- | Weaken the bound on the maximum number of bytes required. For example,
+-- to use two builders with unequal bounds in a disjunctive setting:
+--
+-- > import qualified Arithmetic.Lte as Lte
+-- >
+-- > buildNumber :: Either Double Word64 -> Builder 32
+-- > buildNumber = \case
+-- >   Left d  -> doubleDec d
+-- >   Right w -> weaken (Lte.constant @19 @32) (word64Dec w)
+weaken :: forall m n. (m <= n) -> Builder m -> Builder n
+weaken !_ (Builder f) = Builder f
 
-instance Cat.Category (<=) where
-  id = LessThanEqual 0
-  -- b <= c (c - b) -> a <= b (b - a) -> a <= c (c - a)
-  LessThanEqual cb . LessThanEqual ba = LessThanEqual (cb + ba)
-
--- | Dynamically check than 'n' is less than or equal to 'm'
-isLessThanEqual :: (KnownNat n, KnownNat m) => Proxy# n -> Proxy# m -> Maybe (n <= m)
-isLessThanEqual n m = if 0 <= diff then Just (LessThanEqual diff) else Nothing
-  where diff = knownNat m - knownNat n
-
--- | Statically check than 'n' is less than or equal to 'm'. 'n' and 'm' must be known at compile time.
-lessThanEqual :: forall n m. (KnownNat n, KnownNat m, n GHC.<= m) => n <= m
-lessThanEqual = LessThanEqual (knownNat (proxy# :: Proxy# m) - knownNat (proxy# :: Proxy# n))
-
--- | Weaken the bound on the maximum number of bytes required.
--- >>> :{
--- buildNumber :: Either Double Word64 -> Builder 32
--- buildNumber = \case
---   Left d  -> doubleDec d
---   Right w -> raise lessThanEqual (word64Dec w)
--- :}
-raise :: forall m n. n <= m -> Builder n -> Builder m
-raise !_ (Builder f) = Builder f
+-- | Replace the upper bound on size with an equal number.
+substitute :: forall m n. (m :=: n) -> Builder m -> Builder n
+substitute !_ (Builder f) = Builder f
 
 -- | Encode a double-floating-point number, using decimal notation or
 -- scientific notation depending on the magnitude. This has undefined
@@ -207,7 +161,7 @@ int64Dec (I64# w) = int64Dec# w
 -- the word. This is only used internally.
 wordCommonDec# :: Word# -> Builder n
 {-# noinline wordCommonDec# #-}
-wordCommonDec# w# = construct $ \arr off0 -> if w /= 0
+wordCommonDec# w# = Unsafe.construct $ \arr off0 -> if w /= 0
   then internalWordLoop arr off0 (W# w#)
   else do
     writeByteArray arr off0 (c2w '0')
@@ -230,7 +184,7 @@ internalWordLoop arr off0 x0 = go off0 x0 where
 -- | Requires up to 19 bytes.
 int64Dec# :: Int# -> Builder 20
 {-# noinline int64Dec# #-}
-int64Dec# w# = construct $ \arr off0 -> case compare w 0 of
+int64Dec# w# = Unsafe.construct $ \arr off0 -> case compare w 0 of
   GT -> internalWordLoop arr off0 (fromIntegral w)
   EQ -> do
     writeByteArray arr off0 (c2w '0')
@@ -286,7 +240,7 @@ word8PaddedUpperHex (W8# w) = word8PaddedUpperHex# w
 -- might not be. Benchmark this.
 word64PaddedUpperHex# :: Word# -> Builder 16
 {-# noinline word64PaddedUpperHex# #-}
-word64PaddedUpperHex# w# = construct $ \arr off -> do
+word64PaddedUpperHex# w# = Unsafe.construct $ \arr off -> do
   writeByteArray arr off (toHexUpper (unsafeShiftR w 60))
   writeByteArray arr (off + 1) (toHexUpper (unsafeShiftR w 56))
   writeByteArray arr (off + 2) (toHexUpper (unsafeShiftR w 52))
@@ -309,7 +263,7 @@ word64PaddedUpperHex# w# = construct $ \arr off -> do
 
 word32PaddedUpperHex# :: Word# -> Builder 8
 {-# noinline word32PaddedUpperHex# #-}
-word32PaddedUpperHex# w# = construct $ \arr off -> do
+word32PaddedUpperHex# w# = Unsafe.construct $ \arr off -> do
   writeByteArray arr off (toHexUpper (unsafeShiftR w 28))
   writeByteArray arr (off + 1) (toHexUpper (unsafeShiftR w 24))
   writeByteArray arr (off + 2) (toHexUpper (unsafeShiftR w 20))
@@ -326,7 +280,7 @@ word32PaddedUpperHex# w# = construct $ \arr off -> do
 -- GHC make the decision. Open an issue on github if this is
 -- a problem.
 word16PaddedUpperHex# :: Word# -> Builder 4
-word16PaddedUpperHex# w# = construct $ \arr off -> do
+word16PaddedUpperHex# w# = Unsafe.construct $ \arr off -> do
   writeByteArray arr off (toHexUpper (unsafeShiftR w 12))
   writeByteArray arr (off + 1) (toHexUpper (unsafeShiftR w 8))
   writeByteArray arr (off + 2) (toHexUpper (unsafeShiftR w 4))
@@ -338,7 +292,7 @@ word16PaddedUpperHex# w# = construct $ \arr off -> do
 -- Definitely want this to inline. It's maybe a dozen instructions total.
 word8PaddedUpperHex# :: Word# -> Builder 2
 {-# inline word8PaddedUpperHex #-}
-word8PaddedUpperHex# w# = construct $ \arr off -> do
+word8PaddedUpperHex# w# = Unsafe.construct $ \arr off -> do
   writeByteArray arr off (toHexUpper (unsafeShiftR w 4))
   writeByteArray arr (off + 1) (toHexUpper (unsafeShiftR w 0))
   pure (off + 2)
@@ -350,28 +304,28 @@ word8PaddedUpperHex# w# = construct $ \arr off -> do
 ascii :: Char -> Builder 1
 ascii c = word8 (fromIntegral @Int @Word8 (ord c))
 
--- | Encode an UTF8 char. This only uses as much space as is required.
+-- | Encode a character as UTF-8. This only uses as much space as is required.
 char :: Char -> Builder 4
 char c
-  | codepoint < 0x80 = construct $ \arr off -> do
+  | codepoint < 0x80 = Unsafe.construct $ \arr off -> do
       writeByteArray arr off (unsafeWordToWord8 codepoint)
       pure (off + 1)
-  | codepoint < 0x800 = construct $ \arr off -> do
+  | codepoint < 0x800 = Unsafe.construct $ \arr off -> do
       writeByteArray arr off       (unsafeWordToWord8 (byteTwoOne codepoint))
       writeByteArray arr (off + 1) (unsafeWordToWord8 (byteTwoTwo codepoint))
       return (off + 2)
-  | codepoint >= 0xD800 && codepoint < 0xE000 = construct $ \arr off -> do
+  | codepoint >= 0xD800 && codepoint < 0xE000 = Unsafe.construct $ \arr off -> do
       -- Codepoint U+FFFD
       writeByteArray arr off       (0xEF :: Word8)
       writeByteArray arr (off + 1) (0xBF :: Word8)
       writeByteArray arr (off + 2) (0xBD :: Word8)
       return (off + 3)
-  | codepoint < 0x10000 = construct $ \arr off -> do
+  | codepoint < 0x10000 = Unsafe.construct $ \arr off -> do
       writeByteArray arr off       (unsafeWordToWord8 (byteThreeOne codepoint))
       writeByteArray arr (off + 1) (unsafeWordToWord8 (byteThreeTwo codepoint))
       writeByteArray arr (off + 2) (unsafeWordToWord8 (byteThreeThree codepoint))
       return (off + 3)
-  | otherwise = construct $ \arr off -> do
+  | otherwise = Unsafe.construct $ \arr off -> do
       writeByteArray arr off       (unsafeWordToWord8 (byteFourOne codepoint))
       writeByteArray arr (off + 1) (unsafeWordToWord8 (byteFourTwo codepoint))
       writeByteArray arr (off + 2) (unsafeWordToWord8 (byteFourThree codepoint))
@@ -418,7 +372,7 @@ char c
 -- | Requires exactly 8 bytes. Dump the octets of a 64-bit
 -- word in a big-endian fashion.
 word64BE :: Word64 -> Builder 8
-word64BE w = construct $ \arr off -> do
+word64BE w = Unsafe.construct $ \arr off -> do
   writeByteArray arr (off    ) (fromIntegral @Word64 @Word8 (unsafeShiftR w 56))
   writeByteArray arr (off + 1) (fromIntegral @Word64 @Word8 (unsafeShiftR w 48))
   writeByteArray arr (off + 2) (fromIntegral @Word64 @Word8 (unsafeShiftR w 40))
@@ -432,7 +386,7 @@ word64BE w = construct $ \arr off -> do
 -- | Requires exactly 4 bytes. Dump the octets of a 32-bit
 -- word in a big-endian fashion.
 word32BE :: Word32 -> Builder 4
-word32BE w = construct $ \arr off -> do
+word32BE w = Unsafe.construct $ \arr off -> do
   writeByteArray arr (off    ) (fromIntegral @Word32 @Word8 (unsafeShiftR w 24))
   writeByteArray arr (off + 1) (fromIntegral @Word32 @Word8 (unsafeShiftR w 16))
   writeByteArray arr (off + 2) (fromIntegral @Word32 @Word8 (unsafeShiftR w 8))
@@ -442,13 +396,13 @@ word32BE w = construct $ \arr off -> do
 -- | Requires exactly 2 bytes. Dump the octets of a 16-bit
 -- word in a big-endian fashion.
 word16BE :: Word16 -> Builder 2
-word16BE w = construct $ \arr off -> do
+word16BE w = Unsafe.construct $ \arr off -> do
   writeByteArray arr (off    ) (fromIntegral @Word16 @Word8 (unsafeShiftR w 8))
   writeByteArray arr (off + 1) (fromIntegral @Word16 @Word8 w)
   pure (off + 2)
 
 word8 :: Word8 -> Builder 1
-word8 w = construct $ \arr off -> do
+word8 w = Unsafe.construct $ \arr off -> do
   writeByteArray arr off w
   pure (off + 1)
 
@@ -469,9 +423,6 @@ reverseBytes arr begin end = go begin end where
 c2w :: Char -> Word8
 c2w = fromIntegral . ord
 
-unST :: ST s a -> State# s -> (# State# s, a #)
-unST (ST f) = f
-
 shrinkMutableByteArray :: MutableByteArray s -> Int -> ST s ()
 shrinkMutableByteArray (MutableByteArray arr) (I# sz) =
   primitive_ (shrinkMutableByteArray# arr sz)
diff --git a/src/Data/ByteArray/Builder/Bounded/Unsafe.hs b/src/Data/ByteArray/Builder/Bounded/Unsafe.hs
new file mode 100644
index 0000000..9bad04f
--- /dev/null
+++ b/src/Data/ByteArray/Builder/Bounded/Unsafe.hs
@@ -0,0 +1,59 @@
+{-# language DataKinds #-}
+{-# language GADTSyntax #-}
+{-# language KindSignatures #-}
+{-# language MagicHash #-}
+{-# language RankNTypes #-}
+{-# language ScopedTypeVariables #-}
+{-# language UnboxedTuples #-}
+
+module Data.ByteArray.Builder.Bounded.Unsafe
+  ( -- * Types
+    Builder(..)
+    -- * Construct
+  , construct
+    -- * Run
+  , pasteST
+  , pasteIO
+  ) where
+
+import GHC.TypeLits (Nat)
+import Data.Kind (Type)
+import GHC.IO (stToIO)
+import GHC.ST (ST(ST))
+import GHC.Exts (Int(I#),RealWorld,Int#,State#,MutableByteArray#)
+import Data.Primitive (MutableByteArray(..))
+
+-- | A builder parameterized by the maximum number of bytes it uses
+-- when executed.
+newtype Builder :: Nat -> Type where
+   Builder ::
+        (forall s. MutableByteArray# s -> Int# -> State# s -> (# State# s, Int# #))
+     -> Builder n
+
+-- | Constructor for 'Builder' that works on a function with lifted
+-- arguments instead of unlifted ones. This is just as unsafe as the
+-- actual constructor.
+construct :: (forall s. MutableByteArray s -> Int -> ST s Int) -> Builder n
+{-# inline construct #-}
+construct f = Builder
+  $ \arr off s0 ->
+    case unST (f (MutableByteArray arr) (I# off)) s0 of
+      (# s1, (I# n) #) -> (# s1, n #)
+
+-- | This function does not enforce the known upper bound on the
+-- size. It is up to the user to do this.
+pasteST :: Builder n -> MutableByteArray s -> Int -> ST s Int
+{-# inline pasteST #-}
+pasteST (Builder f) (MutableByteArray arr) (I# off) =
+  ST $ \s0 -> case f arr off s0 of
+    (# s1, r #) -> (# s1, (I# r) #)
+
+-- | This function does not enforce the known upper bound on the
+-- size. It is up to the user to do this.
+pasteIO :: Builder n -> MutableByteArray RealWorld -> Int -> IO Int
+{-# inline pasteIO #-}
+pasteIO b m off = stToIO (pasteST b m off)
+
+unST :: ST s a -> State# s -> (# State# s, a #)
+unST (ST f) = f
+
diff --git a/src/Data/ByteArray/Builder/Unsafe.hs b/src/Data/ByteArray/Builder/Unsafe.hs
new file mode 100644
index 0000000..18c2eba
--- /dev/null
+++ b/src/Data/ByteArray/Builder/Unsafe.hs
@@ -0,0 +1,94 @@
+{-# language BangPatterns #-}
+{-# language DuplicateRecordFields #-}
+{-# language LambdaCase #-}
+{-# language MagicHash #-}
+{-# language RankNTypes #-}
+{-# language ScopedTypeVariables #-}
+{-# language UnboxedTuples #-}
+
+module Data.ByteArray.Builder.Unsafe
+  ( -- * Types
+    Builder(..)
+    -- * Safe Functions
+    -- | These functions are actually completely safe, but they are defined
+    -- here because they are used by typeclass instances. Import them from
+    -- @Data.ByteArray.Builder@ instead.
+  , stringUtf8
+  , cstring
+  ) where
+
+import Data.Primitive (MutableByteArray(MutableByteArray))
+import Foreign.C.String (CString)
+import GHC.Exts ((-#),(+#),(/=#),(>#))
+import GHC.Exts (Addr#,Int(I#),Ptr(Ptr))
+import GHC.Exts (IsString,Int#,State#,MutableByteArray#)
+import GHC.ST (ST(ST))
+import GHC.Base (unpackCString#,unpackCStringUtf8#)
+
+import qualified GHC.Exts as Exts
+import qualified Data.ByteArray.Builder.Bounded as Bounded
+import qualified Data.ByteArray.Builder.Bounded.Unsafe as UnsafeBounded
+
+-- | An unmaterialized sequence of bytes that may be pasted
+-- into a mutable byte array.
+newtype Builder = Builder
+  -- This functions takes an offset and a number of remaining bytes
+  -- and returns the new offset.
+  (forall s. MutableByteArray# s -> Int# -> Int# -> State# s -> (# State# s, Int# #))
+
+instance IsString Builder where
+  {-# inline fromString #-}
+  fromString = stringUtf8
+
+instance Semigroup Builder where
+  {-# inline (<>) #-}
+  Builder f <> Builder g = Builder $ \arr off0 len0 s0 -> case f arr off0 len0 s0 of
+    (# s1, r #) -> case r /=# (-1#) of
+      1# -> g arr r (len0 +# (off0 -# r)) s1
+      _ -> (# s1, (-1#) #)
+
+instance Monoid Builder where
+  {-# inline mempty #-}
+  mempty = Builder $ \_ off0 _ s0 -> (# s0, off0 #)
+
+-- | Create a builder from a cons-list of 'Char'. These
+-- are be UTF-8 encoded.
+stringUtf8 :: String -> Builder
+{-# inline stringUtf8 #-}
+stringUtf8 cs = Builder (\arr off0 len0 s0 -> goString cs arr off0 len0 s0)
+
+-- | Create a builder from a @NUL@-terminated 'CString'. This ignores any
+-- textual encoding, copying bytes until @NUL@ is reached.
+cstring :: CString -> Builder
+{-# inline cstring #-}
+cstring (Ptr cs) = Builder (\arr off0 len0 s0 -> goCString cs arr off0 len0 s0)
+
+goString :: String -> MutableByteArray# s -> Int# -> Int# -> State# s -> (# State# s, Int# #)
+{-# noinline goString #-}
+goString [] _ off0 _ s0 = (# s0, off0 #)
+goString (c : cs) buf off0 len0 s0 = case len0 ># 3# of
+  1# -> case unST (UnsafeBounded.pasteST (Bounded.char c) (MutableByteArray buf) (I# off0)) s0 of
+    (# s1, I# off1 #) -> goString cs buf off1 (len0 -# (off1 -# off0)) s1
+  _ -> (# s0, (-1#) #)
+
+-- We have to have a rule for both unpackCString# and unpackCStringUtf8#
+-- since GHC uses a different function based on whether or not non-ASCII
+-- codepoints are used in the string.
+{-# RULES
+"Builder stringUtf8/cstring" forall s a b c d.
+  goString (unpackCString# s) a b c d = goCString s a b c d
+"Builder stringUtf8/cstring-utf8" forall s a b c d.
+  goString (unpackCStringUtf8# s) a b c d = goCString s a b c d
+#-}
+
+goCString :: Addr# -> MutableByteArray# s -> Int# -> Int# -> State# s -> (# State# s, Int# #)
+goCString addr buf off0 len0 s0 = case Exts.indexWord8OffAddr# addr 0# of
+  0## -> (# s0, off0 #)
+  w -> case len0 of
+    0# -> (# s0, (-1#) #)
+    _ -> case Exts.writeWord8Array# buf off0 w s0 of
+      s1 -> goCString (Exts.plusAddr# addr 1# ) buf (off0 +# 1# ) (len0 -# 1# ) s1
+
+unST :: ST s a -> State# s -> (# State# s, a #)
+unST (ST f) = f
+
diff --git a/test/Main.hs b/test/Main.hs
index 649711e..37b2cb0 100644
--- a/test/Main.hs
+++ b/test/Main.hs
@@ -5,7 +5,7 @@
 
 import Control.Monad.ST (runST)
 import Data.Bytes.Types (MutableBytes(..))
-import Data.ByteArray.Builder.Small
+import Data.ByteArray.Builder
 import Data.Word
 import Data.Char (ord)
 import Data.Primitive (ByteArray)
@@ -14,15 +14,19 @@ import Test.Tasty (defaultMain,testGroup,TestTree)
 import Test.QuickCheck ((===))
 import Text.Printf (printf)
 import Test.Tasty.HUnit ((@=?))
+
+import qualified Data.ByteString as ByteString
 import qualified Data.ByteString.Builder as BB
-import qualified Data.Primitive as PM
-import qualified Data.List as L
-import qualified Data.Vector as V
-import qualified Test.Tasty.QuickCheck as TQC
-import qualified Test.QuickCheck as QC
-import qualified GHC.Exts as Exts
 import qualified Data.ByteString.Lazy.Char8 as LB
+import qualified Data.List as L
+import qualified Data.Primitive as PM
+import qualified Data.Text as T
+import qualified Data.Text.Encoding as TE
+import qualified Data.Vector as V
+import qualified GHC.Exts as Exts
+import qualified Test.QuickCheck as QC
 import qualified Test.Tasty.HUnit as THU
+import qualified Test.Tasty.QuickCheck as TQC
 
 import qualified HexWord64
 
@@ -54,6 +58,9 @@ tests = testGroup "Tests"
         (runArray word64Dec (V.fromList xs))
         ===
         pack (foldMap show xs)
+    , THU.testCase "stringUtf8" $
+        packUtf8 "¿Cómo estás? I am doing well." @=?
+          run 1 (stringUtf8 "¿Cómo estás? I am doing well.")
     , THU.testCase "doubleDec-A" $
         pack (show (2 :: Int)) @=? run 1 (doubleDec 2.0)
     , THU.testCase "doubleDec-B" $
@@ -88,8 +95,8 @@ tests = testGroup "Tests"
   , testGroup "alternate"
     [ TQC.testProperty "HexWord64" $ \x y ->
         run 1
-          (  fromUnsafe (HexWord64.word64PaddedUpperHex x)
-          <> fromUnsafe (HexWord64.word64PaddedUpperHex y)
+          (  fromBounded (HexWord64.word64PaddedUpperHex x)
+          <> fromBounded (HexWord64.word64PaddedUpperHex y)
           )
         ===
         pack (showWord64PaddedUpperHex x <> showWord64PaddedUpperHex y)
@@ -99,6 +106,9 @@ tests = testGroup "Tests"
 pack :: String -> ByteArray
 pack = Exts.fromList . map (fromIntegral @Int @Word8 . ord)
 
+packUtf8 :: String -> ByteArray
+packUtf8 = Exts.fromList . ByteString.unpack . TE.encodeUtf8 . T.pack
+
 -- This is used to test pasteArrayST
 runArray ::
      (a -> Builder) -- ^ Builder