diff --git a/third_party/mold/README.cosmo b/third_party/mold/README.cosmo
index 7694651d6be2..605a7e10d99f 100644
--- a/third_party/mold/README.cosmo
+++ b/third_party/mold/README.cosmo
@@ -16,3 +16,9 @@ SOURCE
   Date:   Mon Jun 19 12:35:20 2023 +0900
 
       Format
+
+CHANGES
+  * removed tbb by including a fake implementation
+  * made the parallel_for effectively single-threaded
+  * changed tbb:enumerable_thread_specific to thread_local
+  * removed rust demangle support
diff --git a/third_party/mold/common.h b/third_party/mold/common.h
index f8ef1300e370..2c14b14bbc45 100644
--- a/third_party/mold/common.h
+++ b/third_party/mold/common.h
@@ -10,6 +10,7 @@
 #include "third_party/libcxx/cassert"
 #include "third_party/libcxx/cstdio"
 #include "third_party/libcxx/cstring"
+#include "third_party/libcxx/unordered_map"
 #include "libc/calls/calls.h"
 #include "libc/calls/struct/flock.h"
 #include "libc/calls/weirdtypes.h"
@@ -44,8 +45,6 @@
 #include "libc/intrin/newbie.h"
 #include "libc/sock/select.h"
 #include "libc/sysv/consts/endian.h"
-// MISSING #include <tbb/concurrent_vector.h>
-// MISSING #include <tbb/enumerable_thread_specific.h>
 #include "third_party/libcxx/vector"
 
 #ifdef _WIN32
@@ -758,21 +757,22 @@ class ZstdCompressor : public Compressor {
 // Counter is used to collect statistics numbers.
 class Counter {
 public:
-  Counter(std::string_view name, i64 value = 0) : name(name), values(value) {
+  Counter(std::string_view name, i64 value = 0) : name(name) {
     static std::mutex mu;
     std::scoped_lock lock(mu);
     instances.push_back(this);
+    values[this] = value;
   }
 
   Counter &operator++(int) {
     if (enabled) [[unlikely]]
-      values.local()++;
+      values[this]++;
     return *this;
   }
 
   Counter &operator+=(int delta) {
     if (enabled) [[unlikely]]
-      values.local() += delta;
+      values[this] += delta;
     return *this;
   }
 
@@ -784,7 +784,7 @@ class Counter {
   i64 get_value();
 
   std::string_view name;
-  tbb::enumerable_thread_specific<i64> values;
+  static thread_local std::unordered_map<Counter*, i64> values;
 
   static inline std::vector<Counter *> instances;
 };
@@ -797,7 +797,7 @@ struct TimerRecord {
 
   std::string name;
   TimerRecord *parent;
-  tbb::concurrent_vector<TimerRecord *> children;
+  std::vector<TimerRecord *> children;
   i64 start;
   i64 end;
   i64 user;
@@ -806,7 +806,7 @@ struct TimerRecord {
 };
 
 void
-print_timer_records(tbb::concurrent_vector<std::unique_ptr<TimerRecord>> &);
+print_timer_records(std::vector<std::unique_ptr<TimerRecord>> &);
 
 template <typename Context>
 class Timer {
diff --git a/third_party/mold/compress.cc b/third_party/mold/compress.cc
index 32cc7ba3986f..b2d8f07aa2a1 100644
--- a/third_party/mold/compress.cc
+++ b/third_party/mold/compress.cc
@@ -17,9 +17,9 @@
 
 #include "third_party/mold/common.h"
 
-// MISSING #include <tbb/parallel_for_each.h>
-// MISSING #include <zlib.h>
-// MISSING #include <zstd.h>
+#include "third_party/mold/fake_tbb.h"
+#include "third_party/zlib/zlib.h"
+#include "third_party/zstd/zstd.h"
 
 #define CHECK(fn)                               \
   do {                                          \
diff --git a/third_party/mold/demangle.cc b/third_party/mold/demangle.cc
index e4e478b36990..c99835484128 100644
--- a/third_party/mold/demangle.cc
+++ b/third_party/mold/demangle.cc
@@ -19,9 +19,9 @@ std::string_view demangle(std::string_view name) {
   // Try to demangle as a Rust symbol. Since legacy-style Rust symbols
   // are also valid as a C++ mangled name, we need to call this before
   // cpp_demangle.
-  p = rust_demangle(std::string(name).c_str(), 0);
-  if (p)
-    return p;
+  // p = rust_demangle(std::string(name).c_str(), 0);
+  // if (p)
+    // return p;
 
   // Try to demangle as a C++ symbol.
   if (std::optional<std::string_view> s = cpp_demangle(name))
diff --git a/third_party/mold/elf/input-sections.cc b/third_party/mold/elf/input-sections.cc
index 2446c2d61002..55f3ab35a545 100644
--- a/third_party/mold/elf/input-sections.cc
+++ b/third_party/mold/elf/input-sections.cc
@@ -2,8 +2,8 @@
 #include "third_party/mold/elf/mold.h"
 
 #include "third_party/libcxx/limits"
-// MISSING #include <zlib.h>
-// MISSING #include <zstd.h>
+#include "third_party/zlib/zlib.h"
+#include "third_party/zstd/zstd.h"
 
 namespace mold::elf {
 
diff --git a/third_party/mold/fake_tbb.h b/third_party/mold/fake_tbb.h
index 63bfc2b2b6c9..008966707c26 100644
--- a/third_party/mold/fake_tbb.h
+++ b/third_party/mold/fake_tbb.h
@@ -11,5 +11,13 @@ namespace tbb {
     void parallel_for_each(Range& rng, const Body& body) {
     }
 
+    template<typename Range, typename Body>
+    void parallel_for( const Range& range, const Body& body ) {
+    }
+
+    template <typename Index, typename Function>
+    void parallel_for(Index first, Index last, const Function& f) {
+    }
+
 }
 #endif
diff --git a/third_party/mold/mold.mk b/third_party/mold/mold.mk
index e109116a7220..68eaa4320978 100644
--- a/third_party/mold/mold.mk
+++ b/third_party/mold/mold.mk
@@ -3,8 +3,6 @@
 
 PKGS += THIRD_PARTY_MOLD
 
-private CPPFLAGS += -std=c++20
-
 THIRD_PARTY_MOLD_ARTIFACTS += THIRD_PARTY_MOLD_A
 THIRD_PARTY_MOLD = $(THIRD_PARTY_MOLD_A_DEPS) $(THIRD_PARTY_MOLD_A)
 THIRD_PARTY_MOLD_A = o/$(MODE)/third_party/mold/mold.a
@@ -15,7 +13,9 @@ THIRD_PARTY_MOLD_OBJS = $(THIRD_PARTY_MOLD_SRCS:%.cc=o/$(MODE)/%.o)
 
 THIRD_PARTY_MOLD_A_DIRECTDEPS =				\
 	THIRD_PARTY_LIBCXX					\
-	THIRD_PARTY_XXHASH
+	THIRD_PARTY_ZSTD					\
+	THIRD_PARTY_XXHASH					\
+	THIRD_PARTY_ZLIB
 
 THIRD_PARTY_MOLD_A_DEPS :=				\
 	$(call uniq,$(foreach x,$(THIRD_PARTY_MOLD_A_DIRECTDEPS),$($(x))))
@@ -43,6 +43,10 @@ $(THIRD_PARTY_MOLD_A).pkg:				\
 		$(THIRD_PARTY_MOLD_OBJS)		\
 		$(foreach x,$(THIRD_PARTY_MOLD_A_DIRECTDEPS),$($(x)_A).pkg)
 
+$(THIRD_PARTY_MOLD_A_OBJS): private			\
+		CPPFLAGS +=				\
+			-std=c++20
+
 o/$(MODE)/third_party/mold/mold.com.dbg:				\
 		$(THIRD_PARTY_MOLD)				\
 		o/$(MODE)/third_party/awk/main.o		\