Probably not, that's certainly a fair point. But in that case why go into it at all? Just say it's magic "faster" juice (and don't post it to HN). One thing I'

Probably not, that's certainly a fair point. But in that case why go into it at all? Just say it's magic "faster" juice (and don't post it to HN). One thing I'm curious about as a developer is how this is different than the type inference that SpiderMonkey already did. The JIT certainly didn't represent numbers as strings in memory or something equally silly prior to this new type inference. Is it better inference that optimizes away boxing or can optimize across function calls or something? (Just musing "out loud", I don't expect answers to these questions here on HN.)