Personal Digital Twins

A personal digital twin is a continuously updated model of an individual's behavior, preferences, schedule, knowledge, and goals. It is built to assist that one person, not to represent them to others.

Where an AI clone is outward-facing, a personal digital twin is inward-facing. Its job is to know you well enough to act on your behalf in narrow, high-leverage ways: planning, drafting, recalling, comparing.

A useful twin holds three categories of state: identity (who you are and what you want), context (what is currently true about your life and work), and memory (what has already happened that you might need to recall).

It is the third category that distinguishes a twin from a generic assistant. A twin remembers the meeting you had three months ago and the decision you made afterward.

The architecture question for personal twins is local versus hosted. A locally hosted twin gives the owner more control and stronger privacy guarantees but limited compute. A hosted twin is more capable but creates the same custody questions that haunt every cloud service.

A growing class of products tries to split the difference: sensitive memory lives on-device, while heavy inference is delegated to encrypted hosted endpoints.