Case study
Mutterly
Voice journaling, 100 percent on-device: speak in 20-second bursts and Mutterly turns scattered thoughts into daily reflections, running open-source models entirely on the iPhone so nothing ever leaves the device.
Journaling for people who will never keep a journal: open Mutterly, talk for twenty seconds, done. At the end of the day it synthesizes everything you muttered into one coherent reflection that helps you spot patterns and find clarity.
The engineering story is the privacy story. Mutterly runs its models completely on-device on an open-source foundation: Llama 3.2 by default, with Gemma, Nemotron, and other open models selectable in settings. Your voice and your thoughts never touch a server, and the whole thing works in Airplane mode. It will summarize literally anything you say to it, and no one but you will ever hear it.
