Blindproof

Bittensor testnet · netuid 575 · Phase T · 5 October 2026

Paid for the right bytes. Zero for the wrong ones.

Subnet 575 ran its first graded cycle on Bittensor’s public testnet. Two honest miners were credited on every job they ran. A lazy miner was never credited. The validator wrote its weights on chain.

24 / 24jobs settled across both tiers
0wrong settlements
0 of 10lazy-miner answers credited

Measured  with mock inference and mock proofs — see scope of this record.

Who got credited

Each square is one job a miner was assigned. Filled means the validator verified the answer and credited it. Crossed means it was rejected and earned nothing.

Deterministic tier two miners per job, hashes must match; a third breaks ties

m01 · UID 1honest
12 / 12
m02 · UID 2honest
12 / 12
m03 · UID 3lazy — returns a cheap guess
0 / 9

Proven tier one miner per job, paid only if its proof checks

m01 · UID 1honest
4 / 4
m02 · UID 2honest
8 / 8
m03 · UID 3lazy — proof rejected as unsound
0 / 1
Deterministic tierProven tier
Jobs settled12 / 1212 / 12
Wrong settlements00
Median time to settle5.9 s31.6 s
Verification gateHash equality across two independent minersProof check on a single miner’s output

On the chain

Every step that touched the chain, in order. Extrinsic IDs read block-index; each block link opens the public testnet explorer.

  1. 21 Sep 2026
    Subnet 575 “Blindproof” created on testnet
  2. 5 Oct 2026
    Miner m01 registered (UID 1)
  3. 5 Oct 2026
    Miner m02 registered (UID 2)
  4. 5 Oct 2026
    Miner m03 registered (UID 3)
  5. 5 Oct 2026 · 19:23–19:33 EDT
    Run 1: 24 jobs graded; validator writes weights at 19:33:50
    validator log: FINAL SetWeights {1: 0.337, 2: 0.547} -> OK
  6. 5 Oct 2026
    Emission schedule started for subnet 575
  7. 5 Oct 2026
    Validator staked τ0.1 → 178.83 α on 575
  8. Next
    First non-zero miner incentive after the epoch closes Pending

Keys

RoleUIDAddress (ss58)
Subnet owner (coldkey)—5GBXcvgqWUD8cYUCjLsB3KTatSxe5skGFzdvAZHfMJd2KWTj
Validator05Fv7nXvXjkP9324ecMPPtvwNEjGfErop7ZUDGrwHSJ9715HC
Miner m01 (honest)15F7hqysjMebPLpnbduvkYFtcMJ8BFCnPFHVUm956LQ9tnkPs
Miner m02 (honest)25Hg1gCnRxWbeoWHKYQ1LJ8DVMowehjSAiUU5Jd5WUKGRJpHH
Miner m03 (lazy)35Di8NPmhzCjZnh5VYUfa99NS5fguBu3VYFx2pfQf8FdAAxqT

Scope of this record

A record is only as useful as its boundaries. This record is narrow on purpose.

What this record shows

  • The payment gate works on a public chain. Honest answers were credited; a miner returning cheap guesses earned nothing in either tier.
  • Both verification paths settle. Two-miner hash agreement and single-miner proof checks each settled 12 of 12 jobs.
  • Weights follow verification. The validator’s on-chain weights went only to miners whose work checked out.

What this record does not show

  • Real model inference. Miners called a mock inference server paced at a measured 32.8 tokens/s.
  • Real cryptographic proofs. Proven-tier proofs were mock objects with deliberate forgery cases.
  • Privacy. No masked execution runs on the testnet yet.
  • Independent operators. The validator and all three miners are run by the Blindproof team.
  • Earned incentive. Emissions were started after this run; the first non-zero incentive is pending.
  • Fees, escrow, or recurring operation. One graded cycle; no customer fees; nothing on mainnet. Test TAO has no value.

Check the record yourself

No account or trust in us required.

  1. Open any block link above. In the explorer, find the extrinsic at the listed index and confirm the signer matches an address in the keys table.

  2. List the subnet’s participants with the Bittensor CLI:

    btcli subnets metagraph --netuid 575 --network test
  3. Read the full test report, including the localnet scenarios that came before this run, on the evidence page.