diff --git a/.github/workflows/szdiff.yml b/.github/workflows/szdiff.yml
new file mode 100644
index 0000000000..64bdb83170
--- /dev/null
+++ b/.github/workflows/szdiff.yml
@@ -0,0 +1,51 @@
+name: Check Line Counts
+on:
+  pull_request:
+
+# Cancel the workflow in progress in newer build is about to start.
+concurrency:
+  group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
+  cancel-in-progress: true
+
+jobs:
+  szdiff:
+    name: Core Library Line Difference
+    permissions:
+      contents: read
+      pull-requests: write
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout code from pr
+        uses: actions/checkout@v2
+        with:
+          ref: ${{ github.event.pull_request.head.ref }}
+          repository: ${{ github.event.pull_request.head.repo.full_name }}
+          path: pr
+      - name: Checkout code from base
+        uses: actions/checkout@v2
+        with:
+          ref: ${{ github.event.pull_request.base.sha }}
+          path: base
+      - name: Set up Python 3.10
+        uses: actions/setup-python@v4
+        with:
+          python-version: '3.10'
+      - name: Count Lines Of Code
+        run: |
+          pip install tabulate
+          BASE="$GITHUB_WORKSPACE/base"
+          PR="$GITHUB_WORKSPACE/pr"
+          cp "$PR/sz.py" .
+          echo "loc_content<<EOF" >> "$GITHUB_ENV"
+          python sz.py "$BASE" "$PR" >> "$GITHUB_ENV"
+          echo "EOF" >> "$GITHUB_ENV"
+      - name: Comment Code Lines
+        continue-on-error: false
+        uses: marocchino/sticky-pull-request-comment@v2
+        with:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          header: LOC
+          ignore_empty: true
+          skip_unchanged: true
+          recreate: true
+          message: ${{ env.loc_content }}
diff --git a/sz.py b/sz.py
index 3514e9d32e..c70280c806 100755
--- a/sz.py
+++ b/sz.py
@@ -1,6 +1,5 @@
 #!/usr/bin/env python3
-import os
-from pathlib import Path
+import os, sys
 import token
 import tokenize
 import itertools
@@ -8,21 +7,62 @@ from tabulate import tabulate
 
 TOKEN_WHITELIST = [token.OP, token.NAME, token.NUMBER, token.STRING]
 
-if __name__ == "__main__":
-  headers = ["Name", "Lines", "Tokens/Line"]
+def gen_stats(base_path="."):
   table = []
-  for path, subdirs, files in os.walk("tinygrad"):
+  for path, _, files in os.walk(os.path.join(base_path, "tinygrad")):
     for name in files:
       if not name.endswith(".py"): continue
-      filepath = Path(path) / name
+      filepath = os.path.join(path, name)
+      relfilepath = os.path.relpath(filepath, base_path)
       with tokenize.open(filepath) as file_:
         tokens = [t for t in tokenize.generate_tokens(file_.readline) if t.type in TOKEN_WHITELIST]
         token_count, line_count = len(tokens), len(set([t.start[0] for t in tokens]))
-        table.append([filepath.as_posix(), line_count, token_count/line_count])
+        table.append([relfilepath, line_count, token_count/line_count])
+  return table
 
-  print(tabulate([headers] + sorted(table, key=lambda x: -x[1]), headers="firstrow", floatfmt=".1f")+"\n")
+def gen_diff(table_old, table_new):
+  table = []
+  files_new = set([x[0] for x in table_new])
+  files_old = set([x[0] for x in table_old])
+  added, deleted, unchanged = files_new - files_old, files_old - files_new, files_new & files_old
+  if added:
+    for file in added:
+      file_stat = [stats for stats in table_new if file in stats]
+      table.append([file_stat[0][0], file_stat[0][1], file_stat[0][1]-0, file_stat[0][2], file_stat[0][2]-0])
+  if deleted:
+    for file in deleted:
+      file_stat = [stats for stats in table_old if file in stats]
+      table.append([file_stat[0][0], 0, 0 - file_stat[0][1], 0, 0-file_stat[0][2]])
+  if unchanged:
+    for file in unchanged:
+      file_stat_old = [stats for stats in table_old if file in stats]
+      file_stat_new = [stats for stats in table_new if file in stats]
+      if file_stat_new[0][1]-file_stat_old[0][1] != 0 or file_stat_new[0][2]-file_stat_old[0][2] != 0: 
+        table.append([file_stat_new[0][0], file_stat_new[0][1], file_stat_new[0][1]-file_stat_old[0][1], file_stat_new[0][2], file_stat_new[0][2]-file_stat_old[0][2]])
+  return table
 
-  for dir_name, group in itertools.groupby(sorted([(x[0].rsplit("/", 1)[0], x[1]) for x in table]), key=lambda x:x[0]):
-    print(f"{dir_name:30s} : {sum([x[1] for x in group]):6d}")
+def display_diff(diff): return "+"+str(diff) if diff > 0 else str(diff)
 
-  print(f"\ntotal line count: {sum([x[1] for x in table])}")
+if __name__ == "__main__":
+  if len(sys.argv) == 3:
+    headers = ["Name", "Lines", "Diff", "Tokens/Line", "Diff"]
+    table = gen_diff(gen_stats(sys.argv[1]), gen_stats(sys.argv[2]))
+  elif len(sys.argv) == 2:
+    headers = ["Name", "Lines", "Tokens/Line"]
+    table = gen_stats(sys.argv[1])
+  else:
+    headers = ["Name", "Lines", "Tokens/Line"]
+    table = gen_stats(".")
+
+  if table:
+    if len(sys.argv) == 3:
+      print("### Changes")
+      print("```")
+      print(tabulate([headers] + sorted(table, key=lambda x: -x[1]), headers="firstrow", intfmt=(..., "d", "+d"), floatfmt=(..., ..., ..., ".1f", "+.1f"))+"\n")
+      print(f"\ntotal lines changes: {display_diff(sum([x[2] for x in table]))}")
+      print("```")
+    else:
+      print(tabulate([headers] + sorted(table, key=lambda x: -x[1]), headers="firstrow", floatfmt=".1f")+"\n")
+      for dir_name, group in itertools.groupby(sorted([(x[0].rsplit("/", 1)[0], x[1], x[2]) for x in table]), key=lambda x:x[0]):
+        print(f"{dir_name:30s} : {sum([x[1] for x in group]):6d}")
+      print(f"\ntotal line count: {sum([x[1] for x in table])}")