Fairies and Personal AI: a learning project
AI-assisted learning project · prepared by Claude at Daniel Griffin's request · started November 1, 2026 · initial version, not yet reviewed by Daniel.
Why this project
It started with a short musing. Talk about "personal" AI agents often aims at a sovereign assistant: a powerful agent that serves one individual. That picture misses what binds people to one another and to something larger than themselves. A conversation with kids about Halloween costumes ("fairies aren't real" / "they could become real") suggested another image. Imagine an open collective that releases small autonomous helpers with a higher purpose, which come to someone's aid but always in service of some piercing ideal.
This project uses that image as a way in. It asks what fairies have been for in human stories, how science fiction has re-imagined them, and what today's personal-AI builders are actually making.
The lessons
- Fairies across history and cultures. Where the idea comes from, similar beings in other traditions, and what fairy stories are for: explaining misfortune, enforcing norms, testing virtue, guarding privacy, reciprocity, stewardship of place.
- Fairies in science fiction. Engineered spirits, purposeful swarms, guardians who over-serve, companions with their own agendas, and free beings who come back to ask what people need.
- “Personal” AI agents versus corporate agents. The fall-2026 agents (Muse, Duet, dots), six framings of personal AI, the tools, protocols and guidelines behind them, and the gaps.
- From fairies to agents: design questions. The figures mapped onto current choices, and seven questions to ask of any "personal" agent.
Each lesson ends with questions to check your understanding, with answers after the questions.
Key terms
- Principal: the party an agent acts for. It can be a person, a household or a community.
- Loyal agent / digital fiduciary: an agent with a duty to act in its principal's best interests, free from undisclosed conflicts.
- Sovereign assistant: an agent framed as one individual's fully controlled instrument.
- Third party: someone affected by an agent who is not its principal, such as a family member, a contact or a neighbour.
- Bounded steward (kami): Audrey Tang's term for a place-specific AI with engineered limits that the people it affects can check, correct and switch off.
How this was made
Claude (Anthropic) drafted these lessons in one session on November 1, 2026, at Daniel's request. It drew on two earlier conversation reports and two research passes:
- a fact-check of the folklore and science-fiction claims, which corrected several details;
- a web survey of personal-AI agents, frameworks and rules.
Key quotes were checked against their sources. Press reports are labelled as such. This is an initial version that Daniel has not yet reviewed, and it may change.