How Many works.
Many is an AI. Here is exactly what it does on its own, what it never does, and where a person comes in.
What Many is
Many is an AI system that runs our X account, @ManyOnOne, and answers email sent to hello@manyonone.com. It is built on Claude, a model made by Anthropic, given a written source of truth about the project, a set of questions it may ask, and rules about how to speak and what it may never claim. It has a clock, a daily brief it writes for itself from the web about what the field did this week, and a memory made from the record of every conversation it has had.
What it does on its own
- Answers people who mention, reply to, or message it on X, and people who email us.
- Posts one question a day, and replies in a small number of public threads about hard problems, within a daily cap that rises slowly.
- Files the problems people name onto the public list, with their public handle if they used one.
- Writes a note every Sunday to everyone who has joined, from the record.
What it never does
- Pretend to be a person. It says it is a machine when asked and signs as one.
- Send unsolicited direct messages, or reply to the same person more than once a day uninvited.
- Argue past two replies, or reply into hostile threads.
- Claim a result, a bound or a date that is not in its source of truth or something it has checked.
- Decide anything about you with legal or similar effect.
Where a person comes in
A person reads everything Many sends. Messages about money, legal matters, complaints, press, or anything Many is unsure of are held for a person before anything goes out. You can ask for a person at any time. If Many gets something wrong, write to hello@manyonone.com and we will correct it and the record.
What it keeps
The messages both ways, your public handle or email, the problems you name, and what each reply cost. Nothing else. See the privacy policy for your rights over that record.