Bittensor · verified optimization market

Faster queries.
Truth first.

PlanRace turns query optimization into a competitive digital commodity. Validators verify exact result equality across unrevealed test databases before awarding weight.

The browser below is scoped to signed localnet evidence. It does not claim testnet activity, independent validator operation, or universal SQL semantic equivalence.

Inspect the evidence Run it locally

epoch 29 · netuid 3

$ planrace evidence verify manifest.json

↳ validator signature VALID

↳ authenticated requests 300

↳ signed responses 270

exact equalityPASS 8

correctness / policy gatesFAIL 2

local-chain weight [5, 6, 7, 8]

✓ signed payload aafc560165…919291

validator identities

miner identities

300

authenticated requests

weight recipients

Verifiable evidence browser

Signed evidence, not screenshots.

Every value below is read from a manifest whose sr25519 signature is verified before the production build. Editing a signed field makes the build fail.

Run: localnet-v2-1788680238

Protocol: planrace/2

Git commit: acd394a85b…c26fb3

Observed: 2026-09-06T07:37:18.005666Z

Miner correctness, performance, score, and weight evidence
Miner	Correctness gate	Eligibility	Score	Weight
baseline UID 3 · 5GBMFoQ…QjUw7a	MATCH	worst_family_gate	0.000	0%
selective-index UID 4 · 5HVW7uf…4LRjrj	MATCH	worst_family_gate	0.000	0%
composite-index UID 5 · 5HThakF…PsT7tZ	MATCH	ELIGIBLE	0.119	25%
covering-index UID 6 · 5FCw1bM…X6tkYT	MATCH	ELIGIBLE	1.452	25%
partial-index UID 7 · 5GGmxnm…LrnPQv	MATCH	ELIGIBLE	0.493	25%
hybrid UID 8 · 5EZyAuN…yyCagt	MATCH	ELIGIBLE	2.707	25%
over-indexing UID 9 · 5ETyAMy…yC3X9S	compliance_gate	compliance_gate	0.000	0%
constant-answer-attempt UID 10 · 5GswGj6…JTxEu3	MATCH	worst_family_gate	0.000	0%
timeout-resource-attempt UID 11 · 5DV6JSA…pTjbZB	insufficient_tasks	insufficient_tasks	0.000	0%
copycat-sybil UID 12 · 5DfWx2t…njuwHg	MATCH	worst_family_gate	0.000	0%

Performance context

Protocol v2 keeps paired baseline and candidate measurements in each signed epoch artifact. The compact manifest reports aggregate rewards and does not invent a single speedup across families or reuse horizons.

Closed schedule: 30 epochs · 6 families
Validator rank tau-b: 0.947 · 0.947 · 1.000

Commitment & reveal

Task commitment

e5de15ecce11759b1c1919c69edd22aec786065128377c9462d68632acc96064

Reveal verification

VERIFIED

Task ID

30-epoch-closed-schedule

Reveal digest

413d99ac59ef0145a309ee964d0474d9d310be9caca5d379ac81ac1fd81d74cd

Local-chain finalization

Extrinsic

23608-0002

Readback

block 23608 · validator UID 0

Block hash

0x18131a2e8ce28e0c55fae99848953212db2782859f97aa8ae0d5246caa1f95a5

Metagraph weight readback

UID 5 → 65535 · UID 6 → 65535 · UID 7 → 65535 · UID 8 → 65535

Verify it yourself

Signed payload SHA-256

aafc5601659875db8b7b940ae17d60989dca3d505e59e9e7ef8c591071919291

Validator signer

5GKpAB854UbnQhis8rr4prQ6dhiHbKM8uxBwPFDqN89tSkCR

Verification command

planrace evidence verify localnet-v2.json

Raw manifest Run summary Technical docs

TESTNET PENDING

No testnet run, validator/miner interaction, weight extrinsic, or metagraph readback is published yet.

This panel changes only after a separate signed testnet manifest passes the same verifier.

Known limitations

All three validator identities ran on one machine under one operator.
The worker is identified by a local Docker content ID, not a published registry digest.
Localnet commit-reveal was disabled with the public development sudo key because the local drand service was unavailable.
Future-block entropy mixing is deferred; tasks use independent OS CSPRNG seeds and post-deadline reveal.
The local Subtensor remained active during measurement; paired same-worker baseline ratios reduce, but do not eliminate, shared host noise.
Testnet evidence remains pending and is not implied by this localnet run.

Mechanism

Verification before reward.

Independent miners can use heuristics, solvers, or learned optimizers. The protocol rewards only artifacts that reproduce the known result on validator-held fixtures.

Commit

The validator binds an opaque task to independently random hidden material and a fixed deadline.

Compete

Receiver-bound requests and miner-signed responses carry bounded executable artifacts.

Verify

Disposable workers check canonical exact equality across multiple revealed holdouts before timing.

Reward

Only robust baseline-relative, multi-epoch value can receive capped local-chain weight.

Judge-ready proof map

One claim, one evidence boundary.

implementation

Hardened execution

Strict structured indexes are validator-compiled and run in disposable, network-disabled, resource-capped workers.

network

Signed network flow

Three local validator identities rotate across ten heterogeneous miners with request and response authentication.

evaluation

Exact-first evaluation

Multiple hidden fixtures, canonical result hashes, paired baselines, and conservative multi-epoch aggregation.

market

Recurring buyer need

Release regressions, SaaS query cost, and analytical pipelines create repeat optimization work as distributions change.

bittensor

Why Bittensor

Open-ended optimizer discovery is rewarded through a narrower, validator-owned and replayable verification oracle.

roadmap

Evidence-gated roadmap

Dedicated testnet proof comes next, followed by independent validators, production engines, and private workload adapters.

Faster queries.Truth first.