Three apps, one approach. Each listens on your own machine, keeps a fast lane that judges what just changed and a slower one that reasons it through, and never lets a late answer overwrite a newer one. Live speech is messy — people talk over each other, misspeak, correct themselves — so the hard part is deciding what actually happened, and when to act. realtim.ing is where that machinery is becoming one shared engine, so each new app on top of it gets smaller.