Idea · carried by 1 source
Agent Personality Through Literature
A technique for giving AI agents distinctive personality and “swagger” by feeding them books you find fascinating, telling them they also find them fascinating, and building literary taste into their identity. The result is output that feels less like a stock LLM response and more like a collaborator with a defined philosophical perspective.
Where it comes from
Jesse Genet, on the a16z podcast (~37:08–39:08), explaining how she gives her agents personality: > “One of the ways you can give it personality is to give my agent a list of the last 10 books I found personally fascinating. And then I'll be like, 'You also find these fascinating.' You read these books and you thought they were really interesting. I try to actually give it an identity that has some swagger.” highlights/building-agents-at-home-a16z — Full podcast context.
How it works
- Book list injection: Feed the agent a curated list of books you love, declaring they are part of its identity
- Literary taste as philosophy: Books carry worldviews; giving an agent specific books gives it a lens through which to interpret questions
- Quirkiness emerges: A different book set produces different answers. Catcher in the Rye vs. Diamond Age would produce fundamentally different gift suggestions for a five-year-old (~37:26–38:07)
- Layered on top of LLM defaults: The literary identity doesn't replace the base model — it adds a filter/perspective on top (~39:02)
- Practical application: The homeschool agent reads curriculum books as “reference points,” but the engineer agent reads Neal Stephenson to develop philosophical grounding (~38:43–38:56)
The breakthrough realization: the technique that started as a practical necessity for the curriculum feed (feeding textbooks to the homeschool agent) turned out to be a general-purpose personality-building method. Books are compact carriers of perspective, and they work as well for shaping agent personality as they do for shaping human personality.