Indexed BittensorBlockchain Data

Bittensor is a Substrate chain, running since 2021, that produces a block every twelve seconds and spends most of them paying out rather than settling transfers. TAO is the native token, divisible into a billion rao. The chain’s unit of organization is the subnet: an independent incentive market identified by an integer netuid, where miners produce a commodity such as inference or storage, validators score what the miners produced, and a subnet’s creator defines what counts as good work. Netuid zero is the root network. At each epoch boundary the runtime combines the weight vectors validators published, weighted by their stake, into ranks and incentive scores, and emissions follow those scores. Every subnet also has a token of its own, called alpha, traded against TAO in that subnet’s pool. Accounts use two keys: a coldkey holds funds, a hotkey operates on a subnet.

SonarX indexes what the chain records: blocks and extrinsics, subnet registrations, the weights validators set, epoch-boundary emission events, staking and unstaking against hotkeys, subnet token pools and their movements, transfers, and derived account balances. Everything is normalized into the model SonarX uses across every network it covers, so subnet accounting sits beside data from chains built for entirely different purposes.

Full historical coverage is available today as a data-share or a file delivery. Real-time delivery is enabled on request.

Transfers are the least interesting thing here. The records that carry meaning are registrations, the weights validators publish and the emissions the runtime derives from them, which means income exists without any signed payment behind it: nobody sends a miner its earnings, the protocol credits them at an epoch boundary. Identity is split, so a participant is a coldkey and its hotkeys rather than one address, and a position read from either half alone is incomplete. Denomination is the third problem, because stake in a subnet is held in that subnet’s own alpha token, so two positions of equal size in different subnets are not the same quantity of anything.

Stake and reward accounting is the first workload, and it is harder here than it sounds: income accrues per epoch, per hotkey and in a subnet’s own token, so a return series has to be built before it can be reported. Staking analytics covers reward accounting across the chains SonarX indexes. Subnet analysis is the second, tracking registrations, emission share and validator concentration per netuid over time. Participant attribution is the third, resolving coldkey and hotkey pairs into operators. Teams following networks that pay machines rather than traders usually hold Fetch.ai and IoTeX beside it.

Tell us the delivery target, how far back you need and how fresh it has to be, and the share is scoped against that. Bittensor history is available now as a share or a file delivery, and streaming is provisioned on request. Say which subnets matter, because scoping by netuid cuts a dense history down sharply. The published coverage matrix for every network SonarX indexes lives in the data documentation, and the rest of the catalogue is at all indexed chains.