01
The parser refuses; it does not guess
You should be able to: List what a trust/1 fragment cannot contain, and explain why decay is derived rather than accepted.
trust/1 refuses unknown fields unless explicitly `x-` prefixed, refuses unlabeled numbers, refuses empty provenance, and refuses any spelling of expiry at all. That last one is deliberate: decay is DERIVED from a `renewed` timestamp, never accepted as a field a caller writes — because a caller who could write an expiry could also write one that never arrives, which is a trust relationship that reads as permanent because nobody dated it.
The same discipline reaches intent: an intent with no `wants` is refused outright, because a want-less intent would be invisible to routing and a record nobody can route from is not really a record of intent.
The pattern to notice is refusal over inference. A parser that guesses a reasonable default for a missing field is a parser that lets a caller omit exactly the field that would have made a claim checkable. trust/1 would rather reject the fragment than fill in a plausible-looking gap.
Source: @magician-network/core — trust/1
02
Every figure is graded by what actually backs it
You should be able to: Explain why a hearsay-only edge or a stale one can never read as fully measured.
Registers everywhere: every figure trust/1 carries is measured, asserted, or estimated, and a combination of inputs is only as strong as its weakest one. An edge backed by hearsay alone caps at estimated, however many hearsay reports pile up — repetition is not verification, and the format refuses to let volume stand in for a source.
Time does the same work. An edge unrenewed for more than 365 days contributes at estimated only, regardless of how solid the original assertion was. A relationship that has not been reaffirmed in a year is not the same claim it was on day one, and treating it as current would be quietly extending a claim past the date anyone actually backed it.
The practical upshot for anyone reading a trust figure: ask what grade backs it before acting on it. A 'measured' figure and an 'estimated' one can describe the same nominal relationship and mean very different things about how much weight it can bear.
03
Nothing a person does alone moves their own standing
You should be able to: State the one thing unilateral activity can never do in this system, and why that is the whole point.
Authority in trust/1 comes from what OTHERS assert about you, never from what you assert or do yourself. The measured figure counts only sealed, non-none outcomes — introductions that actually resolved to something, witnessed by the record, not activity you generated alone.
This is the format's answer to the failure mode every reputation system eventually hits: a user who works out how to move their own number. There is a test in the codebase whose entire job is proving unilateral activity moves nothing — not an edge case caught after the fact, but a standing invariant the suite is built to defend.
The design consequence is that standing cannot be farmed. If the only inputs are other people's sealed assertions about sealed outcomes, then the fastest way to raise your own standing is to actually do work other people are willing to seal a record of — which is the incentive the format is built to leave in place, and the only one.
