Personal agents are small intelligences you trust.
They:
- have access to your email, your calendar, and your microphone
- run locally on your devices
- can write notes on your behalf
- anticipate your taste in restaurants, music and items
- can track tasks and dates
- work offline (useful to debug wifi or tell you if a snake is venomous)
It’s obvious everyone in tech (except maybe Apple) prefers such agents to live in the cloud. Their cloud. If daemons or dmonitors come into being they will look like persistent cloud agents. Communication with such cloud agents may be fully intermediated by your personal agent. Your personal agent knows your reading level, preferred language, and when you’d like communicate via a keyboard or microphone, and it can perform all translations necessary to make this happen, allowing cloud agents to remain simple monolingual text interfaces.
The thinest personal agent you can imagine is like a mobile phone keyboard. An IO interface with some convenience features. Mostly about translation. A thicker agent is more like a sup’ed up Siri.
Features
- input (keyboard, voice, video)
- rolling buffer of audio or video input which user can ask for retroactive analysis of.
- output (text, sound)
- passive awareness of user for polite interruption
- notices if they are busy (calendar, behavior) and doesn’t interrupt unless it’s critical.
- screens calls
- communicates notification preferences to upstream services
- schedule awareness
- can negotiate schedules with other people or their agents
- can send proactive ETA notes or reply on request of another agent.
- privacy awareness
- understands what can be shared with whom, or asks if it doesn’t know
- preference memory
- remember general taste and preferences for setting up external agent tasks/jobs?