Custody of evidence is the part that can improve
The trust deed does not need reinventing. Where it is written down, who can check it, and how quickly it can be proved genuine: those can.
What a public ledger actually gives you
Three things, and it is worth being exact about them because the category is full of overstatement.
- A timestamp that cannot be moved later. The single most common question in estate litigation is which version came last. A commitment recorded on a public chain answers that without relying on anyone's memory.
- Tamper evidence. Not tamper proof. If a record is altered, the alteration is visible. That is a different and more honest claim.
- Verification without permission. Anyone holding the reference can check it against the chain directly, rather than asking the party they may be in dispute with.
Privacy is not the trade. The design intent is that sensitive contents stay private and only a cryptographic commitment goes on chain. What is published is proof that a document existed and has not changed, not the document.
What it does not fix
Capacity
Whether the settlor understood what they signed is a question about a person, not a record. The Astor case turned on exactly this and a ledger would not have decided it.
Undue influence
A perfectly recorded document can still have been signed under pressure.
Legal validity
Whether terms are enforceable depends on the law where they must operate. A timestamp is evidence, not jurisdiction.
Lost keys
Non custodial means non custodial. If you lose your keys, nobody can recover them for you, including us.
Tax
Set by legislatures. Entirely indifferent to how the deed is stored.
Family disagreement
Better evidence narrows what can be argued about. It does not stop people arguing.
Where we intend to use AI, stated as intent
None of this exists yet. The platform is in development. Nothing in this section describes a capability you can use today, and we are not going to write it as though it does.
Two areas are under design. Both are assistive. Neither is intended to replace a lawyer, and neither is intended to make a decision that affects a beneficiary on its own.
- Drafting assistance. Turning a plain description of what someone wants into a structured first draft, for a professional to review. Intended to reduce the cost of the first hour, not to remove the professional.
- Reading and comparison. Summarising what a deed says in plain language for a beneficiary, and highlighting differences between versions. Intended to make an existing document legible, not to change it.
We are not designing an AI trustee, and we do not intend to. A trustee exercises discretion and bears a duty. Neither is something we would hand to a model.
Language models produce fluent text that is sometimes wrong. In a document that governs someone's inheritance for twenty years, fluent and wrong is the worst possible failure mode. Any AI assistance we ship will be reviewable, attributable, and subject to human sign off, or it will not ship.
Where the project actually stands
In development. No timeline is published, because we do not have one we are prepared to be held to.
What exists today is SWED, a TRC20 token on Tron, currently in the final phase of its presale. It confers no ownership, no dividend and no voting right, and no return is promised or implied.