UniNet

Community

Talk to the people building the network.

UniNet is under active development. The conversations worth having right now are architectural — what the primitives are, how they compose, what is finished and what is not. This page is where to find them.

Scope

What this community is for right now

Some layers of the network run. Some are partially connected. Some are specified and not yet built. That shape decides what a conversation here can usefully be about.

Worth asking

Questions that get a real answer

  • How agent identity and the supervisor relationship are modelled.
  • What delegated authority can express in scope and expiry, and what it cannot.
  • How UNP separates users, relays and serving machines, and what each relay sees.
  • What capacity discovery covers across nodes, and what it leaves to your scheduler.
  • Where a design you are holding would break against any of the above.
Not answerable yet

Questions we would answer badly

  • Delivery dates for subsystems still marked in development.
  • Anything resting on hardware-backed confidential computing or HSM-backed key custody. Neither is implemented.
  • Production cross-chain settlement. Parts of bridge verification are incomplete.
  • Price, listings and token mechanics. This is not a channel for that, and it is not what UniNet is being built for.

Check the build status page before you design around any capability. Every subsystem is graded there, including the unfinished ones, and the grading is the same one we use internally. If a status is wrong, say so in any channel above and we will correct it.

Open questions

Where outside input still changes what gets built.

Architecture is cheaper to correct before it ships than after. These four are open, and a concrete constraint from a system you actually run is worth more here than a general opinion.

01Authority models

Delegation that is scoped, time-bounded and revocable covers a large share of cases. The useful message is the one describing a case it does not cover.

02Governance structures

Bring the organizational structure that breaks the model. Significant parts of the external autonomous-organization interface are still placeholder, so the cost of changing the design now is low.

03Privacy threat models

UNP is a routing architecture, not an anonymity claim. If your threat model includes the operator running your workload, describe its shape — that boundary is where the design is still being argued.

04Operator constraints

What the hardware you already run makes cheap, and what it rules out. Reward parameters are not finalized, and operator input is one of the inputs to them.

Before you commit engineering time

Read the architecture for how the layers compose, and the build status for how far each one has actually got. Both are written to be argued with.