Bittensor decentralisation under scrutiny

Covenant AI spent months becoming one of the best-known builders on Bittensor, then walked away on April 9 after calling the network’s governance “decentralization theatre.” Bittensor’s token, TAO, fell sharply after the exit became public, with crypto outlets reporting drops in the mid-teens to about 18% over the next day. (theblock.co) (cointelegraph.com) Bittensor is a blockchain network that breaks artificial intelligence work into separate “subnets,” which are like small marketplaces for one job at a time. Its own documentation says miners do the work, validators grade the work, and TAO rewards are split based on those grades. (docs.learnbittensor.org) Those validators have real power because their scores decide who gets paid. Bittensor says only the top 64 nodes by emissions are eligible to validate by default, and each subnet governor can raise or lower that validator cap. (docs.learnbittensor.org) That design means Bittensor is not one giant open field where everyone participates equally. It is a stack of gated competitions, and the rules on who can validate, how many validators exist, and how rewards flow all shape who stays inside the network. (docs.learnbittensor.org 1) (docs.learnbittensor.org 2) The governance layer above those subnets is also more structured than the word “decentralized” usually suggests. Bittensor’s governance docs say the network is still in a transition from foundation control to community ownership, and that proposals are created by a three-person “Triumvirate” made up of Opentensor Foundation employees before the Senate approves them. (docs.learnbittensor.org) The Senate is not a town hall of all token holders either. Bittensor says Senate seats go to top delegate hotkeys, and a proposal only executes after Senate approval and after a Triumvirate member closes it onchain. (docs.learnbittensor.org) That was the opening Covenant AI attacked. In Sam Dare’s exit statement, as reported by The Block and Cointelegraph, he said Jacob Steeves still had effective control over the Triumvirate and could push changes “without process and without consensus.” (theblock.co) (cointelegraph.com) Dare tied that accusation to specific actions, not just philosophy. Reports on April 9 and April 10 said Covenant alleged subnet emissions were suspended, moderation powers in community channels were restricted, subnet infrastructure was deprecated, and token sales were used as pressure during the dispute. (theblock.co) (cointelegraph.com) Steeves denied those claims. Cointelegraph reported that he said he could not suspend subnet emissions, did not hold privileges beyond normal TAO holders, and had only sold a small portion of his holdings in subnets that were “not running” and near “100% burn code.” (cointelegraph.com) The fight landed hard because Covenant was not a random participant. Cointelegraph reported that Jensen Huang mentioned Covenant’s decentralized training run on the All-In Podcast on March 19, calling it a remarkable technical achievement, which had helped make Covenant one of the network’s highest-profile success stories. (cointelegraph.com) So the argument now is less about whether Bittensor has governance on paper and more about where power sits when a serious dispute starts. If a network says no single person can steer it, builders will test that claim the first time emissions, validator access, moderation rights, or treasury-linked economics become leverage. (docs.learnbittensor.org) (theblock.co) That is why this exit traveled beyond one crypto token’s price chart. Bittensor’s own documents describe a system still moving away from foundation control, and Covenant’s departure turned that transition from an abstract roadmap into a live stress test. (docs.learnbittensor.org) (cointelegraph.com)