ARCTURA
Read the docs
Live on Bittensor · Subnet 505 · Resonance BFT consensus

Autonomous resource management, running as a subnet.

Arctura is Bittensor subnet 505 (SN505), coordinating three things most networks keep separate: validators and miners doing the base work, an agent mesh acting on it, and Resonance BFT — Arctura's own Byzantine fault tolerant consensus round — deciding what's actually settled. A proposal doesn't take effect because one operator pushed it through. It takes effect once two-thirds of the network attests, every round, no exceptions.

Read the ARM docs View the repo

What Arctura is

Three layers, one settlement

Most of what people call "the network" is really three separate jobs happening at once. Arctura keeps them distinct so each can be audited on its own, then reconciles them through a single consensus round. Nobody skips a layer to move faster — that's the whole point.

01 · Subnet
Validators & miners
Bittensor subnet 505. Miners perform work, validators score it against the subnet's scoring function, and stake-weighted consensus sets emissions — the same incentive structure any Bittensor subnet runs on, unmodified.
02 · Consensus
Resonance BFT
A Byzantine fault tolerant round layered on top of the subnet. Proposals — resource allocations, weight updates, parameter changes, new agents — are collected, voted on, and only settle once two-thirds of validators attest. Below quorum, nothing changes.
03 · Mesh
Agent mesh
Task-specific agents that read post-settlement subnet state and act on it — routing work to the right miners, surfacing signal reports to observers, and executing what governance already approved. The mesh never acts ahead of consensus.

Governance

The Council Protocol

Every change proposed to the network — a new agent, a parameter shift, a partner integration — is checked against five tests before it goes to a vote. The order matters: each test gates the next, so a proposal that fails early never reaches Council members who'd only be voting on cosmetics.

01
Signal Alignment
Does the proposal serve the network's actual telemetry — real usage, real load — rather than a narrative about it? A parameter change justified by projected demand fails here until the demand shows up in subnet data.
02
Sovereign Clarity
Can any operator read the proposal and understand exactly what changes and who is accountable for it? Proposals that bundle multiple unrelated changes into one vote don't pass — accountability has to trace to one owner per change.
03
Agent Elevation
Does it give the agent mesh more useful autonomy, or just more surface area to fail in? A new agent that duplicates an existing agent's scope, without adding a distinct capability, doesn't clear this gate.
04
Persistence Test
Does it still make sense a year from now, or is it solving for this week's incentive only? Short-term emission tweaks aimed at a temporary miner shortage are the classic case this gate is built to catch.
05
Upgrade Path
Is there a clean way to revise or reverse it if the assumption underneath turns out wrong? Every proposal ships with its own rollback condition, defined before the vote — not improvised after something breaks.

Getting involved

Three ways to run with the mesh

Participation is technical, not tiered by payment. What you get access to depends on what you're running and what you've proven, not what you've purchased.

Node operator
Run a validator or miner
Stake TAO, register on the subnet, and start earning emissions for validated work. Validators additionally take part in Resonance BFT voting rounds once registered. Full setup, hardware requirements, and staking mechanics are documented publicly.
Node setup docs →
Contributor
Build on the agent mesh
The mesh is open to new agents that clear all five Council Protocol gates. If you have a task worth automating against settled subnet state, the proposal process starts with an issue on GitHub, not a closed-door pitch.
Open an issue →
Observer
Follow the signal
Read the weekly network report, watch consensus rounds settle in real time on the quorum meter, and see exactly what the Council has approved — and why — without running any infrastructure yourself.
Read the signal report →

Questions

Frequently asked

Autonomous Resource Management is Arctura's method for coordinating validators, miners, and AI agents on a shared Bittensor subnet, with Resonance BFT settling consensus on how resources get allocated.

Yes. Arctura runs as subnet 505 (SN505) on Bittensor using the standard miner, validator, and stake-weighted emissions structure. Resonance BFT and the agent mesh are layered on top of that base subnet, not a replacement for it.

Arctura is live on Bittensor as subnet 505 (SN505).

A Byzantine fault tolerant consensus round. Validators vote on a proposed state change, and it only settles once two-thirds of the network agrees — no single validator or operator can unilaterally push a change through.

A five-gate review every proposed change passes through before a vote: Signal Alignment, Sovereign Clarity, Agent Elevation, Persistence Test, and Upgrade Path. Each gate has to clear before the next is checked, so a proposal that fails on substance never reaches a vote on cosmetics.

It's a set of task-specific agents that read subnet state only after Resonance BFT has settled it, then act on that state — routing work to miners, publishing signal reports, and executing changes the Council has already approved. The mesh never acts ahead of consensus.

No. Running a validator or miner requires staking TAO on the subnet directly — there is no separate paid tier, and no revenue share for referring other people into the network. Access to the agent mesh is earned through Council review, not purchase.

Full node setup, staking mechanics, and ARM specification live at autonomousresourcemanagement.com, with source and issue tracking on GitHub.