Update README.md

README.md

@@ -14,7 +14,7 @@ tags:
 **The first reinforcement learning agent to finish the whole game in one continuous run.** A network by krzim
 ([GitHub](https://github.com/krzim), [Hugging Face](https://huggingface.co/zimmerk4)), trained in a JAX simulator of
 the game, climbs the real Steam game from bottom to top with no save/load: 17 of 20 wins in krzim's test. Below, one
-of its wins (2 min 34 s); the circle in the corner is its mouse input.
+of its wins (1 min 35 s); the circle in the corner is its mouse input.
 
 <video controls src="https://huggingface.co/miiwater/goi-rl/resolve/main/videos/krzim-network-wins.mp4"></video>