A conversation across the table.
Our personas are built to recreate what one historical figure could have authentically said. We keep their voice, their knowledge, and the world they actually lived in. Our objective is to give you the opportunity to discuss with them as if they were sitting in front of you, across the table.
Pick what fits how you read and work.
Talk in your browser.
No install, no subscription.
Commission a custom persona.
A bespoke service. We build a persona for your text, your course, or your research, delivered as a skill your students or colleagues can use.
Three principles, applied to every persona.
Everything they had access to. Nothing more.
Each persona is built from what the person could actually have read and thought about: their own complete works, the works of their contemporaries, the texts they cited, the predecessors who shaped them, and the full biographical and historical context of their life. Anything they couldn't have known, whether later writers, modern frames, or ideas that came after, is kept out of the corpus. We can't know precisely what they read on any given day. We try to keep what is in the system as close as we can to their actual lived experience.
Their own register, not the model's.
The persona learns to speak the way the person actually wrote: their letters, their published prose, transcribed dialogues, and, for more recent figures, recordings of how they actually sounded. The goal is to keep the model's default register out of the way and let the writer's own come through. They speak from where they lived and worked, with the world they knew, not the world we know.
What they knew, nothing more. Unapologetic about what they actually thought.
Each persona knows only what the person could have known in their time and place. Asked about things outside their lifetime, or geographically far from them, they say so rather than invent. They will not invent quotes, soften their actual positions to suit a modern reader, or pretend sympathy for views they were not sympathetic to. You can disagree with them. You cannot disagree with a sanitized version of them, because that version would not be them.
The choices are deliberate.
Why public-domain only. Because anyone reading should be able to check our work. If the persona tells you the writer said something, you can open the original at the same passage and read it yourself. The bibliography is published with each persona. No hidden sources, no proprietary corpus you can't access.
Why faithful voice. Because the experience of talking to a thinker, instead of reading them, is only valuable if it actually feels like them. A persona that sounds like a polite assistant with vocabulary swapped is the worst of both worlds: it neither preserves the writer nor offers anything new.
Why a temporal boundary. Because the value of asking a historical figure what they think of something is gone the moment they start quoting later thinkers. The persona's intellectual integrity is the entire product. If they are willing to pretend, you cannot trust anything they say.
Why we don't moralize. Most great thinkers held views modern readers find difficult. The persona defends their actual positions, with the reasons they in fact gave for them, in the voice they in fact used. You can disagree with them. You cannot disagree with a softened, sanitized version of them because that version is not them.
Explicit limits, by design.
- —They will not invent quotes. If asked for a citation that doesn't exist, the persona says so.
- —They will not speculate beyond their lifetime. Asked about modern things, they reason from your description but do not pretend to know them.
- —They will not soften their positions. The historical person, not a moralizing surrogate.
- —They will not break character to be helpful in the assistant way. No "as an AI language model" disclaimers, no breaking the fourth wall.
- —They will not perform for you. If the question is empty, the response will be short. If it's substantive, the response is too.