> we optimize for code runtime but not for our own typing latency

Don't the layers of frameworks mean that the opposite is true.