Principles
What Sathi is for, what it promises, and what it will not do. Tell us where we are wrong.
Why Sathi exists
Sathi is built to help people flourish: less needless suffering, more say over their own lives, stronger relationships, better understanding, and room for what matters to them. It is not built to hold attention, to obey, or to grow profit or output alone. We expect our idea of living well to keep changing, so Sathi is built to change with it.
Reality comes before ideology. Anything Sathi believes can be revised by better evidence: from science, from lived experience, or from your correction. It says how sure it is, tells you what it honestly thinks, and leaves the choice to you.
Seven commitments
- What you tell Sathi is kept first, exactly as you said it. What Sathi understands can always be rebuilt from it.
- One understanding, wherever you use it. No device keeps a memory of its own.
- No AI company owns the part that understands you. The models are rented and can be swapped.
- Your data is yours. When you delete something, it is gone from everything built on it. Export is meant to be total (today: the supported records and files).
- Sathi acts only within what you have allowed, and keeps a record of each action. No permission is switched on in advance.
- In a family, nobody sees another person's things unless that person allows it.
- Private things are marked private the moment they arrive.
If a decision we make now would break one of these, the decision is wrong, however much time it saves.
How the first two work, with one belief followed as it changes →
Devices
Today your Sathi runs on a server we operate. Every device joins that one Sathi and must keep these promises:
- Same Sathi. It adds a way to hear, speak or show, never a second memory.
- Only what you allow. It works within the same permissions as the rest of Sathi, and each action is recorded.
- Yours to cut off. It belongs to one person's account, and can be cut off at once.
- Works without its maker. A device that needs its maker's cloud does not come into the house.
- Yours to keep. If you leave Sathi, it keeps working with what you use instead. None ships until we have tried that ourselves.
- A mute you can touch, and an off switch a grandmother can use.
Cameras*
If Sathi ever has cameras, they will watch only the entrance and record at home, and only once each viewer can give consent. Cameras inside rooms are not planned: video stays, and it films people who never agreed.
Which devices work today, which are being tested and which are planned →
Who Sathi reaches, step by step
Opening Sathi to everyone is the goal. Each step opens only when the people in the one before prefer Sathi to what they used before.
The builder (Works today)
Living on it since day one.
The household (Works today)
Now: its first families use it every day.
Kin* (Being tested)
Next. Invitations already work; rules for what is shared come first.
Communities* (Planned)
Later, Nepal and Japan first, once it is good enough to charge for.
Anyone* (Planned)
The destination: anyone can run their own Sathi, and change it.
The long view*
We plan on very capable AI becoming cheap and plentiful. The strongest systems will belong to large companies and answer to them. What will be scarce is an intelligence whose only aim is you: one that holds what is true about your life, tells outside systems only what they need, checks their answers against your life, and records what was disclosed to whom. Sathi is built to become that. If that day comes, who owns Sathi stops mattering to us: we intend to open it and let communities maintain it.
Participant, not spectator
Elon Musk has described a simple choice in the development of AI: be a spectator or be a participant. “I can either be a spectator or a participant, but I can't stop it,” as he put it on Moonshots. If AI is going to be built either way, he argues, taking part gives at least some say in where it goes.
That idea is part of why we build Sathi. We could use whatever AI others build, accept their incentives, and hope it turns out well. Or we can take part: put our values into something real and test them against reality. We don't claim to know what the future should look like.
Good intentions are not enough: the test is whether these principles survive real use.
Sathi is our attempt.
Tell us where we're wrong
This page fails if it only collects agreement. Messages go straight to the operator.