> Google tried to fix it by introducing an eager mode in Tensorflow but by then it was too late. And the fix was "new major version with a fundamentally differ

> Google tried to fix it by introducing an eager mode in Tensorflow but by then it was too late. And the fix was "new major version with a fundamentally different programming paradigm" But it turns out when your users are irritated with your product, and you tell them to change to a fundamentally different programming paradigm, the new programming paradigm they change to might not be yours.