Bittensor Co-Founder Jacob Steeves Responds to Covenant AI

Not sure what happened with Covenant AI and Bittensor? Read our full breakdown here to get up to speed.

The Covenant AI departure from the Bittensor ecosystem sent shockwaves through the community. On April 11, Jacob Steeves, known as Const, the co-founder of Bittensor, published a detailed response on X in the form of a long-form article. The Jacob Steeves (Const) response addressed the situation directly, combining raw personal reflection with a forward-looking technical proposal designed to prevent similar events from ever happening again.

Here is a breakdown of everything Const said and what it means for Bittensor moving forward.

Const Addresses the Fallout

Const opened his response by describing the events as deeply personal. He referred to Samuel Dare, the founder of Covenant AI, as someone he once considered a brother. According to Const, Dare acted with the intention to cause maximum pain to both him personally and to the Bittensor protocol he has dedicated his life to building.

The emotional weight of the statement was clear. Const acknowledged that the departure hurt everyone who held tokens tied to Dare’s subnets and everyone who believed in the shared vision behind them. He issued a direct and sincere apology to those who lost money, lost faith, or had the financial rug pulled from underneath them.

At the same time, Const did not dismiss the work that Covenant AI accomplished. He credited Samuel Dare with proving something extraordinary. Through intense effort, the Covenant team demonstrated that open-source, permissionless networks can genuinely compete with centralized, billion-dollar, fiat-funded AI labs. Const called this innovation profound and stressed that it should not be forgotten, regardless of how things ended.

The Case for Cryptographic Accountability

A significant portion of the Jacob Steeves Const response focused on the broader structural problem that the departure exposed. Const argued that in an industry built on freedom, relying on traditional legal contracts and governmental accountability is not enough. These mechanisms are slow, manual, and ultimately insufficient for the fast-paced world of AI and crypto.

He went further by painting a picture of the near future. Const predicted that it will no longer be humans running subnets or companies. Instead, faceless agents will build digital systems at lightning speed, rendering legacy understandings of value and law obsolete. The real protection, he argued, must be cryptographic rather than legal.

This framing set the stage for the most concrete proposal in the entire response.

Locked Stake: A New Protocol-Level Feature

Const introduced a concept called Locked Stake, a proposed mechanism that would add a commitment dimension to subnet ownership on Bittensor. The idea is straightforward but powerful. Subnet owners would be able to lock their tokens for a defined period of time, making it impossible to move them before that period expires. This combination of time and stake would serve as a transparent, on-chain signal of long-term conviction.

For investors, Locked Stake would provide predictability and concrete assurance against sudden exits like the one Covenant AI executed. For subnet teams, it would open a new competitive dimension. Teams willing to lock their tokens for longer periods would stand out as more committed and trustworthy.

Const acknowledged that this is one of the hardest unsolved problems in Web3. How do you measure a team’s commitment in a permissionless, open system beyond just incentives and ownership? Locked Stake offers a cryptographic answer to that question.

In a notable detail, Const revealed that the Locked Stake concept was ironically one of the last things Samuel Dare worked on before leaving the Opentensor Foundation. He considered his real mistake not implementing the feature sooner.

SN3, SN39, and SN81 Will Continue

Const moved to reassure the community about the future of the affected subnets. He confirmed that the code behind Subnets 3, 39, and 81 remains fully open source. The vision for what was being built there, he emphasized, was never owned by a single person.

Community members and potentially former Covenant AI team members are already in the process of organizing to continue the work. Const expressed confidence that miners and holders would rally around these subnets and revive them. Functionally, the subnets should not change.

He announced plans to discuss both the Locked Stake proposal and the subnet recovery strategy during the open Thursday call on the Bittensor Discord, inviting everyone to join, ask questions, and provide feedback.

Bittensor Stands as the Most Decentralized AI Protocol

Const closed by reaffirming Bittensor’s place in the broader AI landscape. He called it the most decentralized AI protocol in existence today, while acknowledging that it is not perfect and still has a long way to go. The core message was clear. Bittensor remains the only project building state-of-the-art artificial intelligence with a genuine path toward being truly co-owned by the people who participate in it.

His final words captured the defiant, forward-looking tone of the entire response. Const declared that the next time the network trains a large language model, it will be a one-trillion-parameter model.

The Bittensor community now waits for the technical details behind Locked Stake and the organizational plans for the affected subnets. Both topics are expected to take shape during next week’s open call.

Frequently Asked Questions