docs/governance.mdpinned to impactium@637886d

Governance

Impactium is meant to be governed by the people who build it, and to arrive at its principles through recorded deliberation rather than hardcoded decree. Two mechanisms carry most of the weight: the Universal Constitution and the ActionCatalog.

The Universal Constitution

The network's principles live in a Universal Constitution — a set of articles that mature through a knowledge-refinement ladder rather than appearing fully formed:

MUSE → IDEA → THEORY → LAW → HARDENED

An article is proposed (entering as a MUSE), then advanced up the ladder, each advance carrying a recorded rationale. Members vote; advancing to LAW requires a majority of all Entities, and ratifying to HARDENED (effectively immutable) requires a supermajority. Nothing is erased — the full deliberation history of every article, ratified or not, is retained on-chain.

This matters beyond process: the deliberation record the Constitution produces is part of what the founding Genesis Capsule embeds. Impactium's principles arrive at its launch with provenance — as a documented convention, not as text someone typed in.

Pre-genesis, voting is one Entity, one vote, within a closed vetted membership. How votes are weighted at full scale is a question the network is designed to answer for itself.

The ActionCatalog

Not everything is a constitutional question. The everyday economic parameter — how much a given type of action is worth in principle (the QUANT magnitude fed into minting) — is set by the governed ActionCatalog, one row per action type. A row carries the action's base value, the WhyModifiers that context can apply to it, and the impact domains that kind of action may touch.

What a row carries

An ActionType is the what: the kind of act being recorded. Its base value sits on the 0SCALE, from -3 to +3, and every action starts at 1 — an action type nobody has tuned yet is worth one, not zero. A type the catalog has never heard of resolves to a row named Generic, so a new kind of action is never blocked for want of an entry, only un-tuned.

A WhyModifier is the why: context that changes what an instance of that act was worth. Driving is not the same when you are delivering medicine as when you are driving for pleasure, and both are still driving.

A WhyModifier can never cross zero. It moves the value within the sign its action type already carries. If context changes the kind of act rather than its degree — self-defence is not a murder that scored differently — that is a different ActionType, and the catalog gets another row.

The reason is the same one that seals the impact domains: a modifier that flipped the sign would be netting. It would fold "a life was taken" and "lives were saved" into a single number, and whichever way that number landed, one of those two facts would vanish from the record. Impactium refuses that everywhere else, and refuses it here. Both facts keep their own home — the impact is a Capsule, and the cost that delivered it is a linked 0TRUECOST record.

Rows are cheap. A netted number is not.

A claim names one Why. Not several — the sign rule is checked per modifier when a row is proposed, and that check does not compose: two modifiers can each keep the base's sign and still cross zero together. Allowing several would move the guarantee off the catalog and onto the moment of minting. A second reason is a second ActionType.

A row may also declare which impact domains that kind of action generates, and a record of it may not claim a domain outside them. A row that declares none restricts nothing — the catalog fills in gradually, and an un-tuned action type is never unmintable for want of an entry.

The ActionCatalog is where two of Impactium's most distinctive governance ideas live:

  • Everyone votes. Any member with a wallet can vote on the weights. Governance of the network's values is not reserved for validators or founders.
  • Votes are weighted by proximity. A voter's influence on a given weight is scaled by their proximity to the subject matter — how close they are, in lineage and in participation, to the kind of work being weighted. Those who do the work have more say over how it is valued.

The leading way to compute that proximity is graph-derived: measured from the Hive-Knowledge Graph — lineage distance and participation edges to the subject. This turns the graph from a read surface into a genuine input to governance, which is why its proximity queries must be as deterministic, versioned, and provable as anything else the chain commits to.

Weight balancing is the single largest area of ongoing work in the network's evolution — a standing workstream, not a one-time setting. And because every minted record references the matrix version in force at the time, revising the weights never rewrites history: past impact keeps the value it was assessed at.

Validators

Consensus is run by a governed, vetted validator set — power is assigned by governance, not bought through stake. This fits a network of known, accountable Entities and keeps the early network coordinatable, while leaving room to derive validator standing from net-impact reputation as it grows.

The through-line

Governance in Impactium is deliberate and recorded: principles mature through a visible ladder, economic values are set by proximity-weighted vote, and every change is versioned so it never rewrites the past. The network is meant to be shaped by its members, in the open, with a permanent record of how it got where it is.