What the token era did not answer

The five questions

A practice that built on public ledgers between 2017 and 2019 learned what that generation of protocols answered well: a token proved who controlled a balance, moved it without an intermediary, and left a public trail of the movement. Those were real answers to real questions, and they are the past tense of this field. What that era never answered — because nothing was asking software to act on an organisation’s behalf yet — is the set of questions an autonomous agent forces the moment it can do more than read.

There are five. Who is acting: an identity that is more than a key. On whose authority: a chain from a human, through an organisation, down to the agent and its session. Whom to trust: a way to decide whether a counterparty’s agent is worth dealing with. May value move: a policy that decides whether a payment goes, stops, or waits for a person. And what happened: a record another organisation can check without taking your word for it. The token era answered none of these as an open protocol, and the gap is where an agentic internet is being built.

An address is not an authorisation

The first confusion worth naming is that an address is not an authorisation. Control of a private key proves that whoever holds the key can sign; it says nothing about whether a person or an organisation authorised the holder to act, or how far that authority reaches. An agent that can sign can, by construction, act — and in an enterprise the question is never "can it sign" but "was it allowed to, and by whom, and up to what limit".

The identity layer’s answer is a chain that can only narrow. The open spec for it, delegation/1, refuses any child grant that carries a scope its parent lacks, a cap higher than its parent’s, or an expiry later than its parent’s — refused with a named reason, never clamped and never averaged. It refuses a grant that names no endpoint to enforce it, on the principle that authority nothing checks is theatre. And it refuses a root that is a machine: a chain must begin at a person or an organisation, so there is always a human at the top of the authority an agent exercises. That is the shape the token era’s bare address never had.

Discovery, trust and the questions a token never carried

The middle questions — what exists, who can act on an intent, whether to deal — were not even coherent in a world of anonymous addresses. An agent needs to find an organisation, learn what it offers, and decide whether its counterparty is real, all before any value moves. The graph layer answers "what exists" and refuses to carry a relationship edge with no evidence behind it: an assertion with no published basis is a claim, and the format does not carry claims. The discovery layer answers "who can act on this intent" and refuses to let a caller set an item’s visibility or its expiry — publication is a separate, human act, and staleness is derived, not declared.

The trust layer answers "should I deal with you", and its refusals are the sharpest of the stack: a declined introduction renders to the requester exactly as a path that never existed, so a refusal leaks nothing; only the owner of a relationship can consent to it being crossed; and standing comes from what others assert about you, never from how much you did. None of these could be expressed as a token. They are relationships and consent, and they need a protocol shaped for relationships and consent.

One layer per question

The remaining questions each get a layer. The gateway layer lets an agent reach an organisation that still lives on the ordinary web. Representation lets an organisation declare what it is, in a form a machine reads. Execution is how a piece of cross-organisation work actually runs; commerce is how value moves under a policy; audit is what happened, in a record a stranger can recompute. A single query walks the whole stack — identity, then graph, then discovery, then trust, then gateway, then representation, then execution, then commerce, then audit — and each layer hands the next a checkable fact rather than a promise.

What the token era got right, in the past tense

It is worth being precise about what carries forward. The token-sale era established that a record can be public and still be trustworthy, that value can move without a broker in the middle, and that an append-only log is a stronger primitive than a mutable database when several parties must agree. Every layer above inherits those lessons. What it discards is the assumption that an address, plus a balance, plus a public transaction, is enough to let software act for an organisation. It is not — and a practice that operated the earlier generation is exactly the one that should say so plainly rather than reselling it.

• The lab’s view (planned)FlashyLabs is preparing an engineering companion to this piece at https://flashylabs.com/insights/what-the-token-era-did-not-answer — planned, not yet published.
• The studio’s view (planned)The 4 Ventures thesis desk is preparing an investor-lens companion at https://4.ventures/thesis/what-the-token-era-did-not-answer — planned, not yet published.