Subnet 89: InfiniteQuant
InfiniteQuant is Bittensor Subnet 89. Its on-chain description is oriented around trading-signal grading, and the registered repository is DeltaCompute24/InfiniteQuant-Subnet. The InfiniteQuant README describes miners committing encrypted directional trade calls and validators grading those calls on real market data before setting weights.
What InfiniteQuant Rewards
Rather than rewarding trading opinions or social commentary, InfiniteQuant rewards miners who can
commit encrypted signals that survive validator grading. A miner selects a pair and direction,
encrypts the signal with a drand 24-hour timelock, and records SHA256(signal) on-chain through
set_commitment. Validators later reveal the plaintext, verify it against the on-chain hash, and
grade the call on market candles.
The subnet therefore rewards provable signal quality under shared rules, not self-reported PnL. Emissions flow to qualified miners weighted by wins in the last 30 days multiplied by a lifetime hit-rate tier.
Commit and Reveal Context
The README separates the commit step from the reveal-and-grade step. At commit time, validators anchor entry at the open of the first 1-second bar at or after the commitment block time plus 30 seconds so every validator derives the same entry price. After the timelock opens, validators grade on 1-minute candles: first touch of take-profit wins, first touch of stop-loss loses, and no touch by the horizon is a wash.
This matters because a miner must reveal what was committed. A blob still unfetchable six hours after its reveal round is treated as a decisive loss, so miners cannot hide losing calls behind an unrevealed commitment.
Scoring and Qualification Context
InfiniteQuant’s README frames quality as lifetime and recent performance together. Hit rate is lifetime wins divided by lifetime decisive outcomes and never resets. A miner becomes qualified only after at least 30 lifetime decisive trades and a lifetime hit rate of at least 55%. Weight then scales with wins in the last 30 days multiplied by tier multipliers such as SHARP at 60% and WOLF at 70%.
Washes and voids do not count as decisive outcomes. If no qualified miner has recent wins, emissions burn. New hotkeys also receive an eight-day warmup period with dust emissions while they build a record.
Miner and Validator Roles
Miners publish encrypted signal blobs at public URLs, commit hashes on-chain, and reveal them after the timelock. They do not need a market-data subscription because validators handle pricing. The README also caps submissions at six signals per hotkey per UTC day and describes copy-penalty rules for hotkeys that habitually land second into another live trade.
Validators fetch revealed signals, grade them on shared candle rules, and set weights from the resulting track records. Running a validator requires a paid market-data plan sufficient to grade all supported assets and a validator permit so commits do not pile up unrevealed. Those weights feed into Yuma Consensus.
On-Chain Identity
Live SN89 data is available on TaoStats. The live Finney identity for netuid 89 registers the subnet name as InfiniteQuant, with repository DeltaCompute24/InfiniteQuant-Subnet and project URL infinitequant.app.
Relationship to Yuma Consensus
Subnet 89 uses Yuma Consensus to convert the trading-signal performance weight vectors that validators submit into the emission shares distributed to miners and validators within the subnet each tempo. The Yuma Consensus documentation describes how validator weight submissions are aggregated into consensus weights for each miner registered on the subnet.
In InfiniteQuant’s context, validators grade revealed commits against shared TP/SL and horizon rules, then translate qualified miners’ recent wins and lifetime hit-rate tiers into weight vectors for the subnet. The Emission documentation describes how those consensus weights determine each participant’s share of the subnet’s accumulated emission each tempo.
Development Stage Context
The Introduction to Bittensor describes subnet development as moving from localnet to testnet and then mainnet. For InfiniteQuant (SN89), that sequence changes how readers should interpret signal-commit and grading examples.
In localnet, InfiniteQuant-compatible miners and validators can be developed and tested in an isolated environment. Localnet grading results and emission outcomes do not represent production subnet performance.
On testnet, the README documents netuid 514 as the testnet counterpart using the same protocol with free TAO. Testnet signal grades and validator weights are separate from mainnet subnet state.
On mainnet, InfiniteQuant (SN89) is the live production subnet where miners commit encrypted trade signals and validators grade them on market data to determine real Bittensor emissions. The InfiniteQuant repository describes the mechanism that applies on the production network.
The Bittensor Networks reference separates mainnet, testnet, and localnet. A signal grading result or emission outcome from one environment should not be read as representing production subnet performance in another environment.
Reader Boundary
Subnet 89 InfiniteQuant should not be read as a discretionary trading bot, portfolio manager, or social sentiment feed. The InfiniteQuant README describes a commit-reveal grading protocol where validators score encrypted directional calls against shared market-data rules.
Unrevealed Commits Are Not Free Options
The same README states that miners must reveal what they commit and that a blob still unfetchable six hours after its reveal round counts as a decisive loss. A hidden or unreachable commitment does not substitute for a graded signal when weights are set.
Lifetime Quality and Recent Wins Both Matter
The README separates lifetime hit rate, which never resets, from recent wins in the last 30 days that size emissions for qualified miners. A strong old record alone does not replace the need for fresh qualifying wins after warmup ends.
Validator weights still flow through Yuma Consensus to determine emissions each tempo (Yuma Consensus, Emission).