Reward efficient, reproducible model outputs by scoring quality per compute, verifying dataset provenance, auditing compute claims, and maintaining rolling reputations for stable trusted AI
MeritNet is a Bittensor subnet that rewards efficient, reproducible intelligence rather than raw compute or unverifiable accuracy. Miners are scored on quality per unit of compute, with validator-audited compute claims, dataset provenance verification, and a rolling on-chain reputation system.
Current incentive schemes favor brute-force scaling and short-term benchmark gaming. MeritNet corrects this by penalizing excessive compute and rewarding sustained, cost-effective performance, improving decentralization and signal quality.
Miners submit outputs plus signed compute telemetry, dataset fingerprints, and environment hashes.
Validators re-run sampled inferences, verify compute claims, and attest to provenance.
Scoring combines quality, cost, latency, provenance, and miner reputation:
efficiency = (quality × provenance × latency) / (compute^β)
score = normalized_efficiency × (1 + γ × reputation)
Epoch rewards are distributed using an exponent α to avoid winner-take-all dynamics.
Randomized validator sampling
Stake-backed slashing for false attestations
Reputation decay to prevent permanent dominance
Task-class normalization for fair cross-task comparison
MeritNet strengthens Bittensor’s core mission by turning TAO emissions into a credible market signal for efficient, trustworthy AI, discouraging wasteful centralization while preserving permissionless competition.
Project: Quality and Efficiency Subnet (QE-Subnet)
Status: Pre-Seed / Seed Fundraising
Date: February 4, 2026
The Quality and Efficiency Subnet (QE-Subnet) is a Bittensor subnet designed to reward efficient, reproducible, and verifiable model performance relative to compute cost. By combining cost-aware scoring, validator-audited compute claims, dataset provenance verification, and an on-chain reputation system, QE-Subnet produces a credible market signal for high-value neural compute.
We are seeking pre-seed and seed funding to complete core protocol development, bootstrap validator infrastructure, run pilot deployments, and progress toward mainnet launch.
Total Raise: $500K–$1.5M USD (or equivalent in crypto / TAO)
40% — Core Team Expansion: Protocol engineering, DevOps, validator operations
30% — Validator Infrastructure: Hardware, deployment tooling, validator grants
20% — Pilot Programs: Cloud, research, and enterprise integrations
10% — Operations: Legal, compliance, marketing, and general administration
Large and Growing Market: >$30B annual ML inference spend; decentralized compute exceeding $5B and expanding.
Clear Differentiation: First subnet to combine efficiency-based rewards, dataset provenance, and rolling on-chain reputation.
Strong Token Alignment: QE-Subnet increases the informational value of TAO emissions by rewarding sustainable intelligence rather than raw compute.
Execution Path: Identified pilot partners including GPU infrastructure providers and research organizations.
Technical Credibility: Team experience across Web3 infrastructure, decentralized systems, and ML operations.
Pre-Seed (Current): $100K–$300K
Objective: MVP completion, testnet launch, initial validator set.
Seed (Q2 2026): $500K–$1M
Objective: Validator network scaling, pilot deployments, reputation system hardening.
Series A (Q4 2026): $3M–$5M
Objective: Mainnet launch, enterprise onboarding, ecosystem expansion.
Q1 2026: Testnet live with initial validator infrastructure and pilot partners.
Q2 2026: Pilot inference tasks running; reputation system in beta.
Q3 2026: Mainnet proposal submitted; governance activation.
Q4 2026: Public mainnet launch with 50+ validators and initial enterprise adoption.