We built this to change that — and to be very careful about how. The whole product is an argument that an honest silence is worth more than a convincing invention.
A character never invents a memory. Where there is no memory, there is no answer.
Not a value we aspire to — a rule the product enforces before a single word is generated. Everything else here is downstream of it.
Rule one of the product spec, and the thing we test for before every release.
Why we exist
Photographs survive. Voices and reasoning usually do not — how someone thought, why they made the choice they made, what they would have told you. That part disappears quietly, and nobody notices it is going until it has gone.
What makes this different
This is not a fantasy character. Everything a Legacy Echo character says traces back to something a real person recorded, and it shows you the source. Where there is no memory, there is no answer — and it says so rather than filling the gap.
How we think about consent
A character of someone still here requires that person’s own account and their own consent. Nobody can build one of you on your behalf. And no character ever claims to be the person — it is labelled as AI built from real memories, every time it speaks.
Each of these would make the product more impressive in a demo and worse in the only situation that matters.
It would answer everything, and you would never know which answers were real. The refusal is the feature.
Spoken answers in someone’s own voice are planned, and gated on that person’s explicit permission. Not on a relative’s.
Who inherits administration of a legacy is genuinely unsettled. We would rather say that than publish a promise.
You do not have to decide today whether this is for you or for them. One recording is enough to find out.
The mechanics are on How it works, and the handling of the files is on Security.