TAO—Block—Subnets85 / 128active
Bittensor subnet for continuous, decentralized LLM router optimization
What miners actually race on — the mission definition, its tasks and scoring, and the mission-score board. Not the same thing as emission share.
Unverified leads — the mission might live in one of these. Nothing here is confirmed; if you run this subnet, point us at the real home.
Taonets hasn't verified where this subnet's race mission lives yet — the registered repo is https://github.com/fugal-ai/fugal-subnet, but the mission (tasks, scoring, rounds) may live elsewhere: a bench repo, docs, or a validator API. Verify before trusting it. The repo watch below tracks the registered repo only.
What the validator network actually scored this epoch — read from the chain's Incentive vector, normalized on chain.
These are each miner's normalized incentive from the chain's Incentive vector — the only race score Bittensor stores. It is a share of the validators' consensus weight: all rows sum to 1, so uid #56's 0.802 on Minos means it holds 80.2% of the validated weight — and since emission follows incentive exactly, it also takes ~80% of the miner-side payout. There is no absolute score (an “87/100”) on chain; numbers like that live in each subnet's own off-chain platform. Uids absent from the list received zero validated weight this epoch.
Absolute submission scores (an “87/100”) are computed by validators off chain and never published on chain — subtensor no longer even stores a pre-normalization Scores vector, so the incentive shares above are the only race score that exists on chain. When a subnet publishes scores through its own mission API, they appear in the mission panel above instead.
live chain reads
scores read 4d ago · block 9,485,344 · epoch 26,348
and what they can't
Incentive is the score the validator network converged on for each uid this epoch — it's what pays: emission follows it. It is normalized on chain to sum to 1, so it reads as a share of validated weight. Hidden per-validator test sets (e.g. a private eval split) are computed off chain and never published — the chain no longer even stores a pre-normalization Scores vector — so taonets shows exactly what the chain shows, no reconstruction.
did the race change? the head commit is the probe
The watch hasn't confirmed this repo in 4d ago — the head below may be behind GitHub. Treat the race as possibly moved until the watch is live again.
Merge pull request #192 from fugal-ai/deploy/v2-2026.11.10b-composes
The head commit is the atomic “did the race move” probe: any push to the default branch — new tasks, benchmark changes, scoring edits — lands here on the next watch pass. The confirmed-ago line above proves the watch is actually running.
archived per epoch since taonets started tracking (epoch 26,325)
| Epoch | Block | Time | Winner uid | Winner coldkey | Race score | Top incentive | Qualifier mean incentive | Qualifiers |
|---|---|---|---|---|---|---|---|---|
| Nov 11 | ||||||||
| 26,348 | 9,485,344 | Nov 11, 20:11 | #131 | 5D8FFBJ…4sPEu | — | 0.008 | 0.005 | 190 |
| 26,347 | 9,484,894 | Nov 11, 18:41 | #131 | 5D8FFBJ…4sPEu | — | 0.008 | 0.005 | 187 |
| 26,346 | 9,484,592 | Nov 11, 17:41 | #131 | 5D8FFBJ…4sPEu | — | 0.008 | 0.005 | 190 |
| 26,345 | 9,484,173 | Nov 11, 16:17 | #131 | 5D8FFBJ…4sPEu | — | 0.008 | 0.005 | 193 |
| 26,344 | 9,483,862 | Nov 11, 15:15 | #131 | 5D8FFBJ…4sPEu | — | 0.008 | 0.005 | 195 |
| 26,343 | 9,483,564 | Nov 11, 14:15 | #131 | 5D8FFBJ…4sPEu | — | 0.008 | 0.005 | 196 |
| 26,342 | 9,483,103 | Nov 11, 12:43 | #131 | 5D8FFBJ…4sPEu | — | 0.008 | 0.005 | 195 |
| 26,341 | 9,482,787 | Nov 11, 11:40 | #131 | 5D8FFBJ…4sPEu | — | 0.008 | 0.005 | 196 |
| 26,340 | 9,482,356 | Nov 11, 10:13 | #131 | 5D8FFBJ…4sPEu | — | 0.008 | 0.005 | 199 |
| 26,339 | 9,482,047 | Nov 11, 09:11 | #131 | 5D8FFBJ…4sPEu | — | 0.008 | 0.005 | 199 |
| 26,338 | 9,481,739 | Nov 11, 08:10 | #131 | 5D8FFBJ…4sPEu | — | 0.008 | 0.005 | 199 |
| 26,337 | 9,481,296 | Nov 11, 06:42 | #131 | 5D8FFBJ…4sPEu | — | 0.008 | 0.005 | 209 |
| 26,336 | 9,480,999 | Nov 11, 05:42 | #131 | 5D8FFBJ…4sPEu | — | 0.008 | 0.005 | 218 |
| 26,335 | 9,480,553 | Nov 11, 04:13 | #131 | 5D8FFBJ…4sPEu | — | 0.008 | 0.005 | 213 |
| 26,334 | 9,480,280 | Nov 11, 03:18 | #131 | 5D8FFBJ…4sPEu | — | 0.008 | 0.005 | 214 |
| 26,333 | 9,479,968 | Nov 11, 02:16 | #131 | 5D8FFBJ…4sPEu | — | 0.008 | 0.004 | 223 |
| 26,332 | 9,479,492 | Nov 11, 00:41 | #249 | 5D8FFBJ…4sPEu | — | 0.008 | 0.004 | 222 |
| Nov 10 | ||||||||
| 26,331 | 9,479,097 | Nov 10, 23:22 | #131 | 5D8FFBJ…4sPEu | — | 0.008 | 0.004 | 222 |
| 26,330 | 9,478,795 | Nov 10, 22:21 | #131 | 5D8FFBJ…4sPEu | — | 0.008 | 0.004 | 223 |
| 26,329 | 9,478,496 | Nov 10, 21:21 | #249 | 5D8FFBJ…4sPEu | — | 0.008 | 0.004 | 224 |
| 26,328 | 9,478,023 | Nov 10, 19:47 | #131 | 5D8FFBJ…4sPEu | — | 0.008 | 0.004 | 226 |
| 26,327 | 9,477,707 | Nov 10, 18:44 | #249 | 5D8FFBJ…4sPEu | — | 0.008 | 0.004 | 222 |
| 26,326 | 9,477,417 | Nov 10, 17:46 | #233 | 5D8FFBJ…4sPEu | — | 0.008 | 0.004 | 224 |
| 26,325 | 9,477,015 | Nov 10, 16:25 | #233 | 5D8FFBJ…4sPEu | — | 0.007 | 0.004 | 224 |
The chain does not store past Incentive vectors — history begins when taonets started watching (last 24 epochs kept), and every epoch is archived from live reads only. Gap rows mark stretches the worker never saw — they are downtime, not data. Winner coldkeys read at archival are exact; ones marked current were resolved later from the present ownership map. Race scores are the mission's own published scores for the settled round covering each epoch — retrospective (unknown until the round closed), in the mission's own units, never incentive.
who held the top incentive each day — 8 days back to 2026-11-03
| Day (UTC) | Winner uid | Winner coldkey | Top incentive | Block |
|---|---|---|---|---|
| 2026-11-10 | #233 | 5D8FFBJ…4sPEu | 0.007 | 9,472,101 |
| 2026-11-09 | #233 | 5D8FFBJ…4sPEu | 0.008 | 9,469,720 |
| 2026-11-08 | #8 | 5D8FFBJ…4sPEu | 0.011 | 9,462,535 |
| 2026-11-07 | #34 | 5G3Yd4t…biYdU | 0.011 | 9,455,362 |
| 2026-11-06 | #34 | 5G3Yd4t…biYdU | 0.013 | 9,448,174 |
| 2026-11-05 | #126 | 5FBywvF…HmAe1 | 0.050 | 9,440,991 |
| 2026-11-04 | #214 | 5Fehcnb…bLVx1 | 0.186 | 9,433,810 |
| 2026-11-03 | #0 | 5GbydRV…qwB7m | 1.000 | 9,426,630 |
Sampled one block per day from archived chain state — the block number is the verifiable anchor, the date is ±drift. Identity (hotkey + coldkey) is read at the sample block, so it names who actually won that day. Days the endpoint had already pruned are absent, never zero-filled.
The Emission map — what the scores converted into, live.
leading uid earns ≈ 0.2579 τ / epoch of 17.3 τ/day total
what it costs to race here, live
Best-case frame: every key wins every epoch. Your real edge depends on the model you bring — check the scoreboard and winner history above, and the repo watch before committing.