As with all machine learning projects, it's all about having a large, diverse dataset. No other company is even close to Tesla in terms of how much data they have, and that's the real advantage. Every week, Tesla adds millions of hours of footage of drivers on real roads to their archive.
When Tesla makes a software update like this, they can test it on past driving data to make sure it works before releasing to real drivers. That fast feedback means we'll see more wins faster from the software team as time goes on from the software team.