Glossary

The process of permanently destroying TAO tokens. Used for subnet registration (burn registration ) and other protocol mechanisms. Burned TAO is removed from circulation.

The process of assigning your stake to a validator 's hotkey . Delegators earn a portion of the validator 's emissions minus the take rate.

New TAO tokens distributed each block to validators and miners based on their performance and stake . The emission rate decreases over time following a halving schedule.

A mechanism on test networks to distribute free TAO tokens for development and testing purposes. Not available on mainnet.

The amount of TAO required to be locked when registering a new subnet . The cost adjusts dynamically based on network demand and decreases over time if no new subnets are registered.

The smallest unit of TAO , similar to satoshis in Bitcoin or wei in Ethereum. 1 TAO = 10^9 RAO. Named after Rao Bittensor.

An alternative to burning where TAO or Alpha tokens are returned to the emission pool rather than destroyed. Recycled tokens eventually flow back to validators and miners.

TAO tokens locked to support validators and participate in consensus. Staking earns emissions proportional to the amount staked and validator performance.

The percentage of emissions that a validator keeps before distributing to delegators. Set by the validator , typically 0-18%.

The native cryptocurrency of the Bittensor network. 1 TAO = 1,000,000,000 RAO (10^9). Used for staking, incentives, and transaction fees.

The total amount of TAO tokens in existence. Increases with emissions and decreases with burns. Maximum supply is 21 million TAO .

Subnet -specific stake tokens. When you stake TAO on a subnet , you receive alpha tokens representing your share. Alpha appreciates based on subnet performance.

The process of shutting down a subnet and returning locked TAO to the owner. Can be scheduled by the subnet owner or forced by root network governance.

Dynamic TAO — the mechanism where each subnet has its own Alpha token with a liquidity pool enabling TAO ↔ Alpha swaps. Alpha price reflects market demand: more TAO staked into a subnet increases its Alpha value. Subnet emission shares are determined by TAO Flow (staking activity) rather than static root network weights .

A complete cycle of validation, weight-setting, and emission distribution in a subnet . Occurs every tempo blocks.

A mode where Alpha tokens can be freely traded and transferred between accounts. When disabled, Alpha is locked to the staking relationship.

The consensus and incentive rules used by a subnet . Different mechanisms can implement different validation logic, weight-setting rules, and emission distribution formulas.

A unique identifier (u16) for each subnet in the network. Netuid 0 is the root network . New subnets are assigned sequential netuids.

Subnet 0, which coordinates emissions distribution across all subnets. Root validators set weights on subnets to determine their share of total emissions .

An independent network within Bittensor focused on a specific task or domain (e.g., text generation, image processing). Each subnet has its own validators, miners, and incentive mechanism .

A mechanism allowing subnets to be leased rather than permanently owned. Lease holders pay periodic fees and receive subnet emissions . Leases can be terminated or transferred.

The account that registered a subnet and controls its hyperparameters . Subnet owners can adjust settings like tempo , weights limits, and registration requirements. Receives a cut of subnet emissions .

The percentage of subnet emissions that goes to the subnet owner . Set by the network, typically a small percentage that incentivizes subnet creation and maintenance.

An EMA-smoothed measure of net TAO moving into or out of a subnet 's liquidity pool. Subnets with higher TAO Flow (more staking activity) receive a larger share of block emissions . Replaces static root network weights for emission allocation.

The number of blocks between weight-setting epochs in a subnet . During each tempo, validators evaluate miners and update their weights .

A short identifier (like a ticker) for a subnet 's Alpha token. Subnet owners can set custom symbols to brand their subnet (e.g., "SN1" for subnet 1).

A Substrate account that holds TAO and controls stake . Should be kept offline in cold storage . Used to manage hotkeys and withdraw funds.

A Substrate account used for operational tasks like setting weights and signing transactions. Should be kept on validators/miners. Can be regenerated if compromised.

On-chain metadata about a coldkey or subnet , including name, description, URL, and social links. Set via set_identity and set_subnet_identity calls. Helps users identify validators and subnets.

A multi-signature account requiring M-of-N signers to approve transactions. Used for shared custody, DAOs, and security. Transactions are proposed, approved, then executed.

An account authorized to submit transactions on behalf of another account. Proxies can be limited to specific call types (staking, governance, etc.). Useful for operational security.

The address format used by Substrate chains including Bittensor. Bittensor uses prefix 42, resulting in addresses starting with '5'. Encodes a 32-byte public key.

A unique identifier for a neuron (validator or miner ) within a specific subnet . UIDs are assigned when registering and can be recycled when neurons are deregistered.

The maximum number of blocks a neuron can go without activity before being considered inactive. Inactive neurons may be pruned or lose their permit .

The stake relationship between validators and miners. When a validator gives weight to a miner , they form a bond. Dividends flow through these bonds.

A two-phase weight submission mechanism . Validators first commit a hash of their weights , then reveal the actual weights later. Prevents front-running and weight copying attacks.

A measure of how much a validator 's weight assignments agree with other validators. High consensus means the validator rates miners similarly to peers. Used in emission calculations.

The removal of a neuron from a subnet , freeing its UID for reuse. Occurs when a neuron is pruned due to low performance or when the subnet needs to make room for new registrations.

A validator 's share of subnet emissions , calculated from their bonds with high-performing miners. Validators earn dividends proportional to their stake in successful miners.

A grace period after registration during which a neuron cannot be deregistered due to low performance. Allows new participants time to set up and prove their value.

A miner 's share of subnet emissions , calculated from their normalized rank in Yuma Consensus . Higher-ranked miners receive more incentive.

A consensus parameter (κ) that controls the sharpness of the softmax function used in Yuma Consensus . Higher kappa means more winner-take -all distribution of incentives.

A node that performs useful work (computation, inference, data processing) and is evaluated by validators. Miners earn emissions based on their performance scores.

A general term for any participant (validator or miner ) registered in a subnet . Each neuron has a UID and associated hotkey .

Authorization for a validator to set weights in a subnet . Only validators with sufficient stake and meeting subnet requirements receive permits. Checked before weight submissions are accepted.

The process of removing the lowest-performing neurons from a subnet to make room for new registrations. Neurons with the lowest pruning scores are removed first. Protected during immunity period .

A miner 's position in the subnet ranking, derived from stake -weighted validator scores. Rank determines the miner 's share of incentive emissions .

The process of joining a subnet as a validator or miner . Requires burning TAO or solving a proof-of-work puzzle. Assigns a UID to the new neuron .

A consensus parameter (ρ) that controls the temperature scaling in Yuma Consensus calculations. Works with kappa to tune the incentive distribution curve.

Protection against attacks where an adversary creates many fake identities to gain disproportionate influence. Yuma Consensus achieves this via stake -weighted median: creating many low-stake validators has no effect on consensus since only majority stake can shift the median.

A measure of consensus among validators about a miner 's performance. High trust means validators agree on the miner 's score; used to filter outlier ratings.

A node that evaluates miners, sets weights , and participates in consensus. Validators stake TAO and earn emissions based on their contribution to the network.

A score indicating how reliable a validator 's weight-setting is, based on agreement with the consensus. Also called vtrust. Validators with low trust have reduced influence on miner rankings.

Scores set by validators to rank miners based on performance. Weights determine how emissions are distributed among miners in a subnet .

Bittensor's consensus mechanism that aggregates validator weights to determine miner rankings and emission distribution. Uses a stake -weighted median to be resistant to Sybil attacks and collusion.

The latest version of Yuma Consensus with improved resistance to manipulation and more efficient bond mechanics. Enabled per-subnet via the Yuma3On storage flag.

A Uniswap V3–style AMM design where liquidity providers deposit assets within a specific price range (tick range) rather than across the entire price curve. This provides deeper liquidity around the current price with less capital. Used by the Swap pallet for TAO ↔ Alpha conversions.

A record of assets deposited into a concentrated liquidity pool within a specific tick range. Each position tracks the owner, tick_low, tick_high, liquidity amount, and accrued fees. Identified by a PositionId.

The square root of the current price stored as a Q64.64 fixed-point number in AlphaSqrtPrice. Used for efficient swap math: the core invariant L = Δy / Δ(√P) means swap amounts can be computed with a single multiplication.

A price discretization point in concentrated liquidity . Each tick is an integer index where price = 1.0001^tick, representing a 0.01% price increment. Liquidity providers choose a tick range (tick_low, tick_high) to define the price interval where their capital is active.

Structured data published on-chain by an account via the Commitments pallet . Used by miners and validators to publish metadata like serving endpoints, model versions, and configuration. Requires a deposit proportional to storage used.

A commitment where the data is hidden until a specified block number. Before the reveal block , only a hash is visible on-chain. After the reveal block , the actual data is published. Enables commit-reveal patterns for arbitrary data.

A 32-byte identifier for an account, typically displayed as an SS58 -encoded address. Used for hotkeys, coldkeys, and any on-chain identity .

A waiting period during sensitive operations like coldkey swaps. Allows time for the original owner to contest fraudulent swaps. Duration is configurable by governance.

Authority Round consensus for block production in Subtensor. Block authors take turns producing blocks in a round-robin fashion. Works alongside GRANDPA for finality .

A feature where emissions are automatically staked to a designated hotkey rather than being sent as free balance. Reduces transaction overhead for compounding.

The network endpoint that miners expose for validators to connect to and request work. Includes IP address, port, and protocol information. Set via serve_axon call .

A batch of extrinsics processed together. Bittensor produces blocks approximately every 12 seconds. Each block has a number and hash.

A function that can be invoked on the blockchain, typically requiring a signature. Examples: transfer, add_stake, set_weights.

A delegation mechanism where a hotkey can assign a portion of its stake weight to child hotkeys. Children receive a share of the parent's emissions based on the assigned proportion.

The process of migrating all stake and ownership from one coldkey to another. Requires a waiting period (arbitration ) for security. Used for key rotation or recovery.

The computational difficulty of proof-of-work registration puzzles. Automatically adjusts to maintain target registration rates. Higher difficulty requires more computation.

A distributed randomness beacon (https://drand.love) run by a consortium of independent organizations. Produces verifiable, unpredictable random values at regular intervals using threshold BLS signatures. Subtensor imports drand pulses via the Drand pallet for on-chain verifiable randomness.

A notification emitted when something happens on-chain. Events are indexed and can be queried to track state changes. Examples: StakeAdded, NetworkAdded.

Ethereum Virtual Machine compatibility layer in Subtensor. Allows Ethereum-style smart contracts and enables association of EVM addresses with Substrate accounts.

A transaction submitted to the blockchain. Includes calls (signed transactions from users) and inherents (unsigned data from block authors).

The guarantee that a block cannot be reverted. In Subtensor, GRANDPA provides deterministic finality, meaning once finalized, a block is permanent.

GHOST-based Recursive ANcestor Deriving Prefix Agreement. The finality gadget used by Subtensor to finalize blocks. Validators vote on chains rather than individual blocks for efficiency.

The process of replacing a hotkey while maintaining stake relationships. Useful when a hotkey is compromised or needs rotation. May be subnet -specific.

Configurable settings that control subnet behavior, including tempo , weights limits, registration costs, immunity period , and consensus parameters. Set by subnet owners or root network governance.

Unsigned data included in every block by the block author. Includes timestamp and other consensus-required information. Unlike signed transactions, inherents don't require a signature.

A machine-readable description of the runtime interface. Lists all pallets, calls, events, storage , constants, and errors. Queried via RPC at runtime .

Miner /Maximal Extractable Value — profit that block producers can extract by reordering, inserting, or censoring transactions. Common attacks include front-running (buying before a large trade) and sandwich attacks (buying before and selling after). On Bittensor, MEV risks arise in staking swaps and Alpha pool operations.

An encrypted mempool system that protects transactions from front-running. Users encrypt their transaction using ML-KEM-768 (post-quantum key encapsulation) and XChaCha20-Poly1305 (authenticated encryption) with a key announced by the next block author. The block author decrypts and executes the transaction, but cannot have seen its contents before committing to their key.

A module in Substrate that encapsulates related functionality. SubtensorModule is the main pallet containing Bittensor-specific logic.

A metrics endpoint that neurons expose for monitoring and observability. Named after the Prometheus monitoring system. Set via serve_prometheus call .

A registration method where neurons solve computational puzzles to join a subnet . Alternative to burn registration . Difficulty adjusts based on registration rate.

Restrictions on how frequently certain operations can be performed. Prevents spam and abuse. Different limits apply to staking, weight-setting, and other operations.

Remote Procedure Call interface for querying the blockchain. Used to read storage , submit transactions, and subscribe to events via WebSocket or HTTP.

The state transition function of the blockchain, compiled to WebAssembly. Defines all pallets, calls, events, and storage . Can be upgraded without hard forks.

An emergency state that restricts certain operations on the network. Can be entered by governance to protect the network during attacks or critical bugs.

Simple Concatenated Aggregate Little-Endian encoding. The binary format used by Substrate for serializing data on-chain. More compact than JSON.

On-chain state that persists between blocks. Can be queried using RPC . Examples: TotalStake, Neurons, SubnetworkN.

The blockchain framework used to build Subtensor, also used by Polkadot. Provides pallets, runtime compilation, and RPC interfaces.

Transport Layer Security encryption for axon connections. Enabled via serve_axon_tls to secure communication between validators and miners.

WebAssembly, the format Subtensor's runtime compiles to. Allows the blockchain logic to run in a sandboxed virtual machine and be upgraded without hard forks.