Playing around with Racket. I think what the Racket people fundamentally got right is its focus. The strength of lisps comes mostly from being able to combine a lot of different programming styles in a single semi-coherent framework (which is great for learning) and, related, being able to create specialized languages easily (which is great for CS language experimentation).
Ultimately I don't think any lisp will ever become a truly popular language because the same quality of flexibility that makes it great for teaching and computer science will in a larger group inevitably result in a mess of person-specific mini-languages. Lisps are kind of the anti-Go in that way.