diff --git a/README.md b/README.md
index 638bf73..03bf437 100644
--- a/README.md
+++ b/README.md
@@ -34,9 +34,5 @@ Iterations = `double`s per op.
 
 There is likely overhead in capturing hardware counters, so these should be more "correct"
 
-NOTE - MWC256 is likely poorly implemented (it is supposed to be faster). As seen in the instrumented benchmark below there are a lot of branch mispredictions.
-This is clear from the generated assembly atm but I'm not sure why those branching instructions are generated. `UInt128` support is pretty new
-so maybe there are some inefficiencies there.
-
 ![Scaling iterations](/img/perf-scaling-2.png "Scaling iterations")