The grandmother and the seed phrase

Every few months someone asks why this is not on a blockchain. The question is fair. Identity, ownership, a record no one can forge: it sounds like the thing blockchains were invented for. The answer is a woman in a village in Mymensingh, sixty-eight years old, who has never owned an email address, and who is the patient this has to work for.

Put her on a blockchain and you have handed her a seed phrase. Twelve or twenty-four words that are the only way back into her own health record. Lose them and the record is gone, permanently, with no one to call. There is no help desk for a private key. There is no reset. A teenager in San Francisco who loses a seed phrase loses some money. She would lose the only proof that her heart condition was ever diagnosed.

That reason would be enough on its own. There are more. A blockchain is expensive to write to, and slow, and its cost moves with a token price set by speculation on the other side of the world. A prescription should not cost a fluctuating gas fee to issue. A village clinic with intermittent electricity should not depend on a global consensus network being reachable to record that a woman came in with high blood pressure.

And the thing blockchains are genuinely good at, distributing trust across thousands of strangers who do not know each other, is a problem Bangladesh's health system does not have. The trustworthy parties already exist. The medical council. The diagnostic chains. The pre-departure office. They are known, named, and authoritative. The job is not to invent trust among strangers. It is to let the trust that already exists travel.

So the identifier is did:web. It resolves the way the rest of the web resolves. An identifier like did:web:khamhealth.com:doctor:bmdc-A-54219 is a document sitting at a known address over HTTPS, holding a public key. To check a doctor's signature, you fetch his key and verify. No chain, no token, no global network that has to be reachable. The same protocol that loads a news site loads a clinical identity.

The signature underneath is Ed25519. It is not exotic. It is the signature in SSH, the thing engineers use to log into servers every day. It is in Signal. It is in modern TLS, which means it is in the lock icon on every bank login in the country. It is fast enough to run on a cheap phone and old enough to be trusted. Choosing it was the opposite of clever, and clever is the enemy here.

This honesty has a cost, and the cost has to be said out loud. did:web is not decentralized. The identifiers resolve over a domain one operator maintains. If that operator vanishes, the identifiers stop resolving. A blockchain would not have that single point of failure. This is the real trade, made with eyes open: a system one operator must keep alive that a grandmother can actually use, over a trustless system she cannot. The mitigation is boring and institutional, not a token. The credentials are W3C standard and the wallet exports, so the records survive the operator even when the resolver does not. A succession obligation is written into the structure, so the registry can pass to another keeper. Portability is the answer to single-operator risk.

The grandmother never sees any of this. She sees a card on a phone, or a number an assistant holds for her. She holds no seed phrase, because a system that asks a sixty-eight-year-old to safeguard twenty-four words has already failed her. The cryptography is real and the cryptography is hidden, and both of those are the point. The most important property of national health infrastructure is that the person it serves never has to understand it to be protected by it.