The hard questions this design invites, answered plainly. If yours is not here, the essays go deeper — start with One question, many rooms.
No. The software rehearses the argument and provokes; people deliberate and decide. The output is a contestability map — where a deliberation converges, where it splits and along which lines, what is under-explored, where the evidence is thin. It is never a verdict, a recommendation, or an answer. When a rehearsal reaches agreement too fast, the tool flags it as under-explored rather than settled. The machine’s whole job is to make the human decision better informed and the record of it verifiable.
In a rehearsal, yes — they are composite personas, and we say so. A rehearsal exists to surface the fault lines of an argument before real people spend their time on it: where it is strong, where it is thin, what a real room would fight about. It is a wind-tunnel, not the flight. When real members deliberate, the composites are gone; only what real people actually decided is ever described as where they stand.
The one thing the map may never do is manufacture agreement by blending dissent away. A referendum’s majority erases its minority by design — 51% is a result and 49% is nothing. A consolidated assembly keeps the minority legible: a lone position is surfaced and attributed to the room that held it, a split inside a room is carried up verbatim. Dissent is treated as a finding, not as noise to be smoothed out.
You should not have to, and you are not asked to. Every deliberation’s record is sealed on sovereign New Zealand and EU infrastructure — no US-owned or China-owned cloud in the stack — write-once, its signature checkable offline by anyone, without trusting us or any platform. The consolidated map is sealed the same way and traceable down to each room’s record, with a cryptographic binding that the sealed record is the very report that entered the consensus. Do not take the platform’s word for it. Check the record.
A synthetic voice is barred from standing in for tangata whenua, iwi, hapū or Māori as an identity — a refusal built into the code, not a settings option. That bar is a best-effort control, labelled provisional in the code itself, explicitly subject to confirmation by tangata whenua we have not yet had. Where a question engages tikanga or Māori data, it routes to real people deliberating in the ways they choose.
No. The federated tier — the consolidation that draws many sealed assemblies into one dissent-preserving map — is built and has been validated end to end against our own live governance infrastructure: real synthesis, real seals, real offline verification. It is not running in production for any organisation. It can be demonstrated; it is not in anyone’s hands today. Anyone who tells you otherwise is selling.
A single open-weight model of modest size, run on sovereign infrastructure — no US or China cloud in the inference path. We say “single model” deliberately: one model is not a panel of independent minds, and we do not pretend it is. The humility is the point. The tool’s value is in structuring the argument and sealing the record, not in the brilliance of any one model.
A referendum has breadth and no reasoning — a whole-country yes or no, counted. An assembly has reasoning and no breadth — a small room given the evidence and time. A federated assembly runs the same question in many rooms at once and draws the sealed results together: the breadth of a vote and the reasoning of an assembly, with every room’s dissent kept legible.
Any organisation that has to decide together and is too spread out to sit in one hall — a party’s branches, a co-operative’s members, a union’s worksites, a professional body’s regions, a board over its subsidiaries. Democracy is the hardest setting, which is why it leads: what survives a nation’s argument will survive an AGM. The country is just the biggest room.