diff --git a/extra/optimization/helpers.py b/extra/optimization/helpers.py
index 88f6f25b9d..8fd13000a4 100644
--- a/extra/optimization/helpers.py
+++ b/extra/optimization/helpers.py
@@ -28,6 +28,7 @@ from tinygrad.helpers import dedup
 def load_worlds(filter_reduce=True, filter_noimage=True, filter_novariable=True):
   fn = Path(__file__).parent.parent / "datasets/sops.gz"
   ast_strs = dedup(gzip.open(fn).read().decode('utf-8').strip().split("\n"))
+  assert len(ast_strs) > 5000, "dataset size too small"
   if filter_reduce: ast_strs = [x for x in ast_strs if "REDUCE_AXIS" in x]
   if filter_noimage: ast_strs = [x for x in ast_strs if "dtypes.image" not in x]
   if filter_novariable: ast_strs = [x for x in ast_strs if "Variable" not in x]