Natural-language task entry replaces forms with a single text field. The idea is two decades old (Cultured Code's Things had it in 2007). The on-device LLM era makes it finally robust.
The pattern
Type "Pay rent friday at 5pm #finance high priority" in one field. The system parses date, time, tag, and priority. No tabbing through fields, no date picker.
Why on-device LLMs help
Pre-LLM parsers were brittle — "next thursday" worked but "the thursday after the conference" didn't. On-device LLMs handle ambiguity gracefully. The catch: parsing happens locally, so it stays fast and private.
When to confirm
Always show the parsed result before saving. Even good parsers misread 5% of the time. A confirmation step costs nothing and prevents wrong-day rent reminders.
Voice as input
Voice → text → natural-language parsing is the fastest possible task entry. Sub-three-second flow from "I should..." to a saved task.
Where to draw the line
Don't try to parse complex multi-field input ("create a recurring weekly task with a 30-min reminder and link it to the Q4 launch project"). At that complexity, a form is honestly faster.
About Sovereign — A privacy-first AI personal assistant that runs entirely on your iPhone. On-device LLM, zero-knowledge encryption, and a coach that learns from your own words. See how it works or visit the homepage.