You're not missing anything. It's collapsing the state space at the value level rather than interpreter wide, but it's a similar process. Because it doesn't ca

You're not missing anything. It's collapsing the state space at the value level rather than interpreter wide, but it's a similar process. Because it doesn't capture the whole state it is vulnerable to false positives. But that in turn probably makes it more memory efficient. My solution would make Google disappointed in a whiteboard interview, imo.