This is not attempt number one
From the outside, this project looks recent: a prototype in early 2026, a production system by summer, a family using it daily. The truth is slower and, we think, more convincing.
The idea — one trusted system that holds a family's life, in its own languages, on its own terms — dates to 2019. It has been started more than once since then, under more than one name, and it stalled every time in the same two places: nights-and-weekends time, and the absence of a team. Those stalls are not a secret and they are not the story; what matters is what persisted through them.
Dates, for the record
In late September 2022, years before any of the current code existed, the names were secured — sathi.ai on September 29, and altruistic.ai two days later. That pairing is the oldest naming of the architecture itself: the product and the open framework it would answer to, named in the same week, as one commitment.
Decentralized identity was part of the thinking from early on: if identity and trust interactions ever become wallet-native, a sovereign node should be addressable there too — trust mechanisms should not depend on any registrar's goodwill, ours included. sathi.eth was registered in September 2023, while that year's attempt was underway, and is held today: an option for that future, never a pivot. Wrapped names even allow a member's subname to be made irrevocable by the operator — the dismissible-operator principle extended into the naming layer. (All of these dates are public record, for anyone who cares to look.)
Through 2023 there was another full attempt — branding, product imagery, working sessions — that once again outran the hours available to it. We keep an image from that iteration in the repository the way you keep a photograph of a house you didn't finish building: not as failure, as foundation.
Why this attempt is different
Not discipline. The missing ingredient was never conviction — it was a team, and in 2026 the team finally arrived in an unexpected form: models became capable enough that one person working with an AI could do what previously required several people who never materialized. There is something almost too neat about that, because it is the project's own thesis folded back on itself: intelligence became abundant, exactly on schedule — which is precisely why the scarce things this system protects (trust, continuity, sovereignty) needed to be built now. The tool that unblocked the build is the same force that makes the build urgent.
So the current system is best read not as a first draft but as the surviving draft: the decisions that made it through every stalled iteration — the family first, the archive as the asset, deletion as a right, exit as a door that must actually open — are the ones that were re-chosen from scratch, multiple times, years apart. An idea that gets rebuilt after every collapse is the closest thing a builder has to evidence that it is not a whim.