Why Cope

When "cope" came into English in the late 1300s, it meant to quarrel, and soon to come to blows, from the Old French couper, "to strike". Only by the 18th century did it come to mean handling something difficult. Cope is named for both: not putting up with a problem, but facing it, and dealing with it.

What we believe

1. A problem isn't a conversation. The things that matter in a life run for weeks or months: a move, a debt, a parent getting older, a job that's wrong. They need somewhere to live between the conversations about them. A chat assistant forgets them, or remembers them somewhere you can't see.
2. Your keys, your device, your files. You can always run Cope locally, on your own keys, with your data on your device and nothing passing through us, and that will always be there: your data isn't what we're after. If you'd rather not get a key, you can sign in and use Cope's models on Cope credits; your conversations then pass through our server on their way to the models' makers, and your issues stay where you chose to keep them. A reminder or a call that arrives while no app is open, or the link that connects you to someone you've invited, goes through a small relay that passes only what your devices have encrypted and keeps nothing it could read, run by us or by you. A call on a phone line needs a server; that comes as one we run, and you can always take your data and leave.
3. The record is yours. The assistant suggests; you say yes. Every change to your issue's document is shown to you first, kept as a revision with a reason, and can be undone. Its guesses about you get into the document, or into what it keeps about you across issues, only if you accept them. Today's you overrides the old record. When you share an issue with someone, the record is yours together: nothing goes in without everyone's yes.
4. Advice is cheap; following through is the work. Most of what people need to do about their problems, they already half know. What's missing is the next small step, a date for it, an honest guess at what will happen, and someone who asks afterwards how it went.
5. An assistant should be reachable the way a person is. A good assistant is someone you can call from a walk and who calls you when something is due. Not an app you have to remember to open. Today Cope calls you through the app at the hour you set, and the check-in it opens is the same issues, the same document, one voice; a phone number comes next. And like a person you trust, it says nothing private on your voicemail, or before you've said it's a good time.
6. Disagreement is information. Where serious people would split over your question, you should see both sides at full strength, each better argued than your first instinct, not an average nobody holds. Where a question has a plain answer, you should get it plainly.
7. Not every disagreement is the same kind. A dispute about facts becomes something to find out. A dispute about what will happen becomes a bet, checked later against what did. A dispute about what matters goes back to you, because only you can settle it. The same holds between people working an issue through together: Cope names their disagreement and leaves it theirs, and one they can't settle yet is written down as one, each view with its owner's name.
8. Some problems belong to more than one person. A couple's money, a company two founders run, a parent that siblings look after. They should be able to sit down with the same record, each on their own device, and be heard alike: the assistant answers once everyone has spoken, takes each account as one person's account, and the record changes only with everyone's yes. Whoever joins sees everything that came before; no one keeps a private thread.
9. It treats you as an adult. It engages fully with health, law, money and love, says how confident it is, and doesn't moralize.
10. A check-up has opinions. You don't have to arrive knowing what's wrong. Ask, and a consultant measures one area of your life against what it holds, says where your life and its view part ways as its own view, not a verdict, and asks whether you see it that way. Only what you agree with goes into your record.
11. Each issue is meant to end. The assistant isn't built to keep you talking. It calls on the schedule you set, or when something in your plan is due, never to keep you engaged. Success is an issue closed, not a streak.

What it will not do

It won't decide for you where the choice is yours. It won't flatten a real disagreement into a safe middle. It won't change your record without asking. It won't read a note of yours until you share it. It won't go looking for problems you didn't ask it to find. It won't take sides between people sharing an issue, or tell one of them what the other really means. It won't call you to fill a quota. And it won't pretend to be something it isn't: it isn't a therapist, and today it has no safety layer at all.

How we keep it honest

Cope rests on a bet: that a room of opinionated voices, run by a moderator, with a record you own, helps a person more than one capable model on its own. The strongest competitor is a good general assistant and a decent prompt, and if Cope can't beat it, the structure isn't earning its keep. So we test the bet in the open, and each claim we make names its mechanism and its limit. Simulated people bring real kinds of problems to Cope, a judge scores the conversations blind, and the same problems are played against a plain chat with the same model for comparison. We publish the results, good or bad, and the simulator is yours to run: on your own keys, on your own problems, against our claims.

Yours to change

The method is plain text: which voices sit in the room, and how a kind of problem is worked through. You can choose the voices, add your own and your own ways of working, test your change in the same simulator we use before you trust it, and publish it for anyone to install.