Hacker News new | ask | show | jobs
by abecedarius 2078 days ago
I suspect the key is to somehow make it really quick and easy to retroactively fix mispredictions and other mistakes. That'd open up the space for more aggressive guess-what-I-mean input methods.