Sappho
Laozi
Confucius
Sun Tzu
Socrates
Plato
Aristotle
Cicero
Cleopatra
Seneca
Epictetus
Marcus Aurelius
Leonardo da Vinci
Niccolò Machiavelli
Nicolaus Copernicus
Elizabeth I
Galileo Galilei
Johannes Kepler
Isaac Newton
Benjamin Franklin
Adam Smith
Abigail Adams
Mary Wollstonecraft
Sojourner Truth
John Stuart Mill
Abraham Lincoln
Charles Darwin
Ada Lovelace
Henry David Thoreau
Frederick Douglass
Susan B. Anthony
Florence Nightingale
Andrew Carnegie
John D. Rockefeller
Nikola Tesla
Booker T. Washington
Jane Addams
Ida B. Wells
Marie Curie
Winston Churchill
For experts, coaches & the people who learn from them
Good counsel has always been rationed by the hours in one person’s day.
Simulations of leaders in politics, business and science, each answering from that person’s actual record — and room beside them for a mind built on yours.
Free tier of 20 messages a month · $129 a year or $14.99 a month
- 1,200+
- minds on the record
- 0
- quotes made up, of 1,323 checked
- 50 × 1,000
- sources and pages, in your own
People already take their questions to AI. What nobody has fixed is what the AI is standing on when it answers.
A 47-country study found 66% of people relying on AI output without checking its accuracy, while only 46% are willing to trust AI at all. University of Melbourne & KPMG, 2025 (48,000 respondents, 47 countries, fielded Nov 2024–Jan 2025)
Three things happenbefore a word of advice reaches you.
- 01
Somebody assembles the record
A person, not a crawler, chooses what stands behind each mind: speeches, letters, published work, interviews. Every document keeps its maker, its date, its kind, and where it is held.
- 02
The pages get read before you ask
Scans that are only pictures are typed up a page at a time, and the scan stays beside the text so you can compare. A talk becomes a transcript; a transcript becomes searchable.
- 03
The mind goes looking
It lists its archive, searches it, opens a document, reads the page before. Then it answers from what it found, with the passage attached.
“The man who dies thus rich dies disgraced.”
The AI never types a quote. It marks the passage, our system copies the words straight out of the document, and anything that is not an exact match is thrown away. Advice set out as a quotation cannot have been invented.
Ask a mind something its record does not cover and the answer comes back short, with no quote under it. There was nothing to copy.
The honest numbers
If an AI will speak near your reputation,ask for the failure rate first.
So we publish ours. 1,250 questions put to one simulation, answered from 17 documents split into 11,117 searchable pieces, every answer marked against checks written in advance. It measures whether an answer stays tied to its record — the thing that matters when the record is yours, or when you are taking its counsel.
- 97%
- more of our benchmark's rubric checks passed than the same model with no documents attached
- 0
- quotes made up, across 1,323 checked against the documents
- 0
- highlights on the wrong passage
- 0
- quotes from outside the mind's own record
Humy internal benchmark, August 2026: 58.2% of the rubric checks passed with the record attached, 29.6% without, for the one configuration scored. Across the 1,323 quotes in that run, one citation opened nothing, and we report it rather than round it away.
The sentences around a quote are still written by the AI. Attaching the documents cut outright invention from 22.6% of answers to 12.2%, which still leaves roughly one answer in eight carrying an invented date or reference in that surrounding text. The quote itself is always real, and the document sits one click away. Read the full benchmark.
People have always taken counsel from the dead.This makes the archive answer back.
Reading the letters of statesmen and the notebooks of scientists for guidance is an old practice. The difference here is that you can put your own situation to the mind, and it will search its record for what bears on it.
- PoliticsStatecraft, from the people who practised itAbraham Lincoln on holding a country together; Winston Churchill on holding a line. Each answers from their own speeches and letters.
- BusinessEnterprise, argued by its buildersAndrew Carnegie and John D. Rockefeller, questioned on capital, competition and what wealth is for, out of what they actually wrote.
- ScienceDiscovery, explained by the discoverersIsaac Newton and Marie Curie, answering from papers and letters rather than from a textbook's summary of them.
- LettersCounsel that has kept for centuriesMarcus Aurelius on conduct, Benjamin Franklin on habit and money. The originals of most advice you have already heard secondhand.
You have said more than anyone can absorb in a call with you.
Books, talks, a newsletter, a course, years of answers to the same forty questions. A simulation built from up to 50 of your own sources — 1,000 pages each — answers in your words, quoting you verbatim, in more than 50 languages, at three in the morning. For a course you teach, a practice you run, or the people your mission is meant to reach. Your sources stay private to your account unless you share them.
- 01
You cannot charge for access
There is no paywall, storefront or revenue share for your audience. You can share your simulation and put it in front of anyone; selling seats to it is not something Humy does today.
- 02
It does not remember a visitor
Every conversation starts clean. A returning client is a stranger each session. Memory across sessions is coming, and it is not here today.
- 03
The prose around a quote can still slip
In our benchmark, roughly one answer in eight carried an invented date or reference in the AI-written text around the quote. The quote itself is always real, with the document one click away.
- 04
It is not you
It answers from what you gave it and nothing else. A question outside the record comes back short, and it will not improvise your judgment on a situation you never wrote about.
- 05
Voice is capped, tables stay in text
Live voice calls run up to 10 minutes with natural turn-taking. Images and tables live in text chat.
Scoped against your material
For a larger record — books to transcribe, years of talks, an organisation behind the mission. We scope the build against what you have and you test it before anyone else sees it.
- A 30-minute walkthrough, with your material as the example
- A simulation built from writing and talks you already have
- How we ran our benchmark, in writing
- Conversation logs, searchable and exportable
$129 a year, or $14.99 a month.
One price covers the whole library and everything you build from your own record. Someone assembles each archive by hand and the pages are read months before you ask anything, which is the part you never see and most of what you are paying for.
Free
20 messages a month
No credit card. Enough for the one test worth running before you pay: ask a mind about the field you know best, then look at where the quotes came from.
Plus
$129 per year
or $14.99 per month, billed monthly
- The whole library of 1,200+ minds, in text or live voice
- Up to 50 simulations of your own, built from documents you upload
- Every quote copied out of its document, with that document one click away
- More than 50 languages, spoken and written
- A pool of 12,480 messages a year, or 1,040 a month if you pay monthly
Humy is paid for by 92 schools, 6 universities and 13 businesses, and by 143 people who bought it for themselves. Larger deployments are scoped on a call rather than off a price list.
What experts ask us first.
Source-grounded AI simulations of people, events and places. Each one is built on a curated set of real documents — speeches, letters, published work, interviews — and searches that record before it answers. When it quotes, our system copies the words straight out of the document; the AI never types a quote itself. The ready-made library holds more than 1,200 minds, from statesmen to scientists to builders of industry, and you can build your own from documents you upload.
The library is historical and public figures, each grounded in their own record. A living expert can build a simulation of themselves from their own material — books, talks, newsletters, course notes — and share it with the people they teach, coach or advise. What we do not offer is building a simulation of a living person from someone else's account: the record it stands on should be yours to give.
Because the AI cannot write one. It marks which passage of which document it means, and our system copies those words out of the source, discarding anything that is not an exact match. In our August 2026 internal benchmark, all 1,323 quotes matched their documents word for word, none highlighted the wrong passage, and none came from outside the figure's own sources. One citation in that run opened nothing, and we report it rather than round it away.
Sometimes, in the prose around the quote. The quotation itself is copied from the document and cannot be invented, but the sentences framing it are still written by the AI. In the same benchmark, attaching the documents cut outright invention from 22.6% of answers to 12.2% — roughly one answer in eight still carried an invented date or reference in that surrounding text. That is why every claim keeps its source one click away.
As a study and counsel tool, yes: put your situation to a mind and it will search its record for what bears on it, quoting the source so you can read the original in full. It is not the person, and it is not professional advice — legal, medical or financial questions belong with someone licensed to answer them. Read it the way you would read the collected letters of someone you admire: evidence first, guidance second.
It does not remember a visitor between sessions — every conversation starts clean, and memory across sessions is coming but not here today. It has no paywall or revenue share, so you can share your simulation but not sell seats to it. It answers only from its record: a question outside the documents comes back short rather than improvised. And voice calls cap at 10 minutes, with images and tables staying in text chat.
Upload the record: up to 50 sources per subject and 1,000 pages per source — books, PDFs, scans, transcripts, audio. Pages without machine-readable text are transcribed page by page. You set the instructions and reading level, test it privately with your own hardest questions, and share it when you are satisfied. Your sources stay private to your account unless you choose to share them.
When the record is large or the stakes are institutional: books that need transcribing, years of recorded talks, an organisation that wants review before anything is public, or a deployment across a course or a practice. Book a call and we will scope the build against your material; you can put your own questions to the result before anyone decides anything.
$129 per year, or $14.99 per month, with a free tier of 20 messages a month to try it first. One subscription covers the whole library and everything you build from your own record. Team-scoped builds are priced against the material once we have seen it. You can cancel whenever you like and keep access until the end of the paid period.
No. Nothing you upload trains a model, ours or a provider's, and conversations are never used for training either. Your documents are stored as source records that a simulation searches at question time, held in the US or Sweden on Microsoft Azure. Withdraw a document and the answers that leaned on it stop coming; there is nothing left to quote.
Something we have not answered? Write to us at hello@humy.ai.