Deep research as a commodity. Faster, cheaper, traceable research — produced by a competitive swarm of miners on Bittensor SN67.
Neuron registration trends
The alpha price in τ, day by day, and what moved it each way.
Log change over the last 30 days −23.55%: protocol buys 0.00%, root sells −4.44%, users net −27.04%, cross-subnet +7.94%, unattributed +0.0001%.
The chart follows the alpha price in τ. Each period starts at its opening price and steps through the mechanisms below, in a fixed order, to its closing price: a step that rises pushed the price up, one that falls pushed it down, and its height is how far it moved the price. A thin line carries each close into the next period’s open; it is dashed where the two differ. The faint candle behind each period is the market’s own open, high, low and close, for context.
With many periods on screen (90 days, or hours), the steps would be too thin to read, so the chart draws the price as a line instead and colours each period by the mechanism that moved it most. Hover a period for the full step-by-step breakdown.
Not here, because they do not move the spot price: the per-block TAO injection into the pool, and burns.
All figures are log changes, not simple percentage changes, so the parts add up exactly to the total; on big moves the two differ (a 10% log fall is a 9.5% price fall).
2026-11-10 UTC, as of ≈00:45 UTC (block 9,472,314)
The day isn’t over yet, so these figures keep changing. They are in no daily total, average or rank.
Research execution at scale requires decomposing complex queries into retrieval, ranking, reasoning, and synthesis steps—but evaluating research quality across many implementations is expensive, centralized, and hard to reproduce. A decentralized network of competing researchers can improve workflows faster than a single team, but without transparent scoring and enforced execution standards, quality degrades and coordination fails.
Harnyx addresses this by making research execution a competitive market. Miners submit Python agent scripts that solve research queries under strict tool budgets and time limits. Validators execute those scripts in sandboxed environments, capture tool use and responses, and score results against reference answers using both correctness metrics and pairwise judging. The platform takes validator score medians, ranks miners, and records weights on-chain—creating financial incentives for better research decomposition and tool orchestration.
The result is a continuous feedback loop: miners see which workflows score highest, competitors iterate to improve, and the network automatically upgrades the research pipeline. All execution is transparent and auditable—tool calls, citations, reasoning traces, and scoring evidence are recorded for every task.
Auto-generated from the GitHub repository and may not reflect the latest changes.
Harnyx (SN 67) is a Bittensor subnet for deep research that turns research execution into a competitive harness where miners submit Python agent scripts to answer research queries, validators execute and score them in sandboxed environments against reference answers, and the network returns intelligence with provenance through on-chain weight distribution based on correctness and evidence quality.
Miners implement a query entrypoint that receives a research query and returns an answer (text and optional output/note/citations). Validators poll the platform for assigned task attempts, fetch miner artifacts, execute scripts in sandboxed containers with tool access, capture execution logs and responses, submit task results to the platform, and optionally run similarity judging. The platform coordinates task batches, aggregates validator scores (taking medians), and computes weights using champion selection and participant allocation rules.
Bittensor records weights on-chain for emission distribution. Tool execution happens through a platform-tool-proxy that mediates access to search, LLM, embedding, reranking, and other tools within session and budget constraints.
From data version 12, miners qualify with 10 tasks then compete on 20 additional shared main tasks. Final champion decision uses all 30 tasks. F1 and Overall each select up to 30 main participants independently.
Challengers are compared against incumbents; a challenger replaces the incumbent when: (1) score is strictly higher and at least min(perfect score, incumbent score + 10 percentage points); (2) score and runtime do not regress with median cost ≥10% lower; or (3) score and cost do not regress with median runtime ≥10% lower and ≥1000ms. Champion emission uses B = 0.80 × Overall champion's Overall score. F1 allocation = B × s × F1 retention factor where s = 0.40 + 0.50 × max(0, F1 champion's F1 score − Overall champion's F1 score).
Retention factor is 1 on champion change, 0.5 on first retention, 0.25 on later retentions. Participants divide the remainder by distribution weight = participation stage × novelty multiplier, capped at 50.
The sections above were auto-generated from the GitHub repository and may not reflect the latest changes. View repository (github.com, opens in a new tab, external site)
Deep research as a commodity. Faster, cheaper, traceable research — produced by a competitive swarm of miners on Bittensor SN67.
This information was auto-extracted from the GitHub repository and may be incomplete or out of date. Always refer to the subnet's GitHub repository (github.com, opens in a new tab, external site) for the latest instructions.
Hardware requirements: see the subnet's repository (github.com, opens in a new tab, external site)
| Day (UTC) | Share |
|---|---|
| 2026-11-09 | 0.01% |
| 2026-11-08 | 0.01% |
| 2026-11-07 | 0.01% |
| 2026-11-06 | 0.01% |
| 2026-11-05 | 0.01% |
| 2026-11-04 | 0.01% |
| 2026-11-03 | 0.01% |
| 2026-11-02 | 0.01% |
| 2026-11-01 | 0.01% |
| 2026-10-31 | 0.01% |
| 2026-10-30 | 0.01% |
| 2026-10-29 | 0.01% |
| 2026-10-28 | 0.01% |
| 2026-10-27 | 0.01% |
| 2026-10-26 | 0.01% |
| 2026-10-25 | 0.01% |
| 2026-10-24 | 0.01% |
| 2026-10-23 | 0.02% |
| 2026-10-22 | 0.02% |
| 2026-10-21 | 0.02% |
| 2026-10-20 | 0.02% |
| 2026-10-19 | 0.02% |
| 2026-10-18 | 0.02% |
| 2026-10-17 | 0.02% |
| 2026-10-16 | 0.02% |
| 2026-10-15 | 0.03% |
| 2026-10-14 | 0.03% |
| 2026-10-13 | 0.03% |
| 2026-10-12 | 0.03% |
| 2026-10-11 | 0.03% |
| 2026-10-10 | 0.03% |
| 2026-10-09 | 0.03% |
| 2026-10-08 | 0.03% |
| 2026-10-07 | 0.03% |
| 2026-10-06 | 0.04% |
| 2026-10-05 | 0.04% |
| 2026-10-04 | 0.03% |
| 2026-10-03 | 0.04% |
| 2026-10-02 | 0.04% |
| 2026-10-01 | 0.03% |
| 2026-09-30 | 0.04% |
| 2026-09-29 | 0.04% |
| 2026-09-28 | 0.04% |
| 2026-09-27 | 0.04% |
| 2026-09-26 | 0.05% |
| 2026-09-25 | 0.05% |
| 2026-09-24 | 0.05% |
| 2026-09-23 | 0.05% |
| 2026-09-22 | 0.05% |
| 2026-09-21 | 0.05% |
| 2026-09-20 | 0.06% |
| 2026-09-19 | 0.06% |
| 2026-09-18 | 0.06% |
| 2026-09-17 | 0.07% |
| 2026-09-16 | 0.08% |
| 2026-09-15 | 0.06% |
| 2026-09-14 | 0.05% |
| 2026-09-13 | 0.09% |
| 2026-09-12 | 0.09% |
| 2026-09-11 | 0.09% |
| 2026-09-10 | 0.08% |
| 2026-09-09 | 0.11% |
| 2026-09-08 | 0.12% |
| 2026-09-07 | 0.13% |
| 2026-09-06 | 0.08% |
| 2026-09-05 | 0.17% |
| 2026-09-04 | 0.16% |
| 2026-09-03 | 0.18% |
| 2026-09-02 | 0.20% |
| 2026-09-01 | 0.16% |
| 2026-08-31 | 0.15% |
| 2026-08-30 | 0.16% |
| 2026-08-29 | 0.17% |
| 2026-08-28 | 0.14% |
| 2026-08-27 | 0.16% |
| 2026-08-26 | 0.18% |
| 2026-08-25 | 0.19% |
| 2026-08-24 | 0.19% |
| 2026-08-23 | 0.19% |
| 2026-08-22 | 0.20% |
| 2026-08-21 | 0.26% |
| 2026-08-20 | 0.25% |
| 2026-08-19 | 0.28% |
| 2026-08-18 | 0.25% |
| 2026-08-17 | 0.32% |
| 2026-08-16 | 0.29% |
| 2026-08-15 | 0.32% |
| 2026-08-14 | 0.31% |
| 2026-08-13 | 0.31% |
| 2026-08-12 | 0.31% |
