Bittensor Community Presses Const on the Conviction Upgrade. Here Are His Answers.

Last week, the Bittensor ecosystem was shaken by the abrupt exit of the Covenant team, a move that saw roughly $10 million walk out the door and left the affected subnets drained of their founding operators. We covered the fallout in full detail:

The Conviction upgrade is the network’s response to prevent the next Covenant-style disaster. Below is a rundown of the Novelty Search community call, convened specifically to address it.

The Opening Pitch

Const opened the community call with a simple premise: Bittensor learns from its exploits, and the recent subnet rug exposed a real gap in the protocol.

Almost everything on Bittensor is mutable: validator incentives, miner emissions, the viability of subnets themselves. But the subnet owner key isn’t. That immutability is exactly what allowed the rugger to keep draining 18% of emissions from the pool after walking away.

His proposed fix is a new primitive called conviction. Stakeholders can lock their alpha tokens to a subnet key for a set duration, producing a cryptographically verifiable signal of long-term commitment. Conviction is calculated as stake multiplied by time, with locked tokens unlocking linearly. Crucially, the score used to determine subnet ownership peaks at the midpoint of the lock, not instantly, so no one can flash-attack a subnet.

The rollout will be phased. First, locking ships as an optional primitive that anyone can use. Then it gets activated for ownership voting on Subnets 3, 39, and 81 (the three subnets controlled by the Covenant team) so the community can take them back. Only after that will it extend to other mature subnets. Newly registered subnets get an immunity period.

“You don’t own a subnet unless you have something at risk,” Const said. “Stake is how much you own. Incentive is how much you work. Stake times time is your conviction.”

The Community Pressed Const (Q&A)

Is conviction just an anti-rug measure? No. Const framed it as an ownership and token-rights measure as well. It gives alpha tokens more value by letting the community control the brand, the slot, and the mechanism behind a subnet, and not just speculate on its price.

As a top holder of Templar, will you lock all of your alpha? Yes. Const said he intends to lock all of his alpha and hand the subnet to a team emerging organically from the miner community.

Won’t this kill teams’ ability to raise capital? Const pushed back on this hard. Teams can still sell tokens via OTC contracts tied to future unlocks, the same way the foundation funded itself in its earliest days. “You basically are asking investors to have some conviction in what you’re building as well.”

What if a whale locks a huge amount of alpha to sabotage a subnet? This was the room’s biggest concern. Const’s response: the economics don’t work. An attacker who locks to seize a subnet can’t sell, while everyone else can. If the community disagrees with the takeover, holders exit first, and the attacker is left holding devalued tokens for the duration of their lock. Validators can also refuse to shift weight, meaning the attacker inherits a key but not a functioning mechanism.

Can other people lock in support of the existing owner? Yes, and this was the detail that pieced everything together. Investors, miners, and aligned stakers can all lock their alpha to the owner’s key as a pledge of allegiance. For a hostile takeover to succeed, the entire distributed holder base has to side with the attacker. On a mature subnet, Const argued, that’s nearly impossible.

Will newly registered subnets be exposed to this immediately? No. There will be an immunity period. Conviction-based ownership voting is being built for mature subnets, roughly a year old or more, where a meaningful holder base already exists.

When does the team that rugged stop getting paid? When this mechanism ships and Subnets 3, 39 and 81 are reclaimed through the new voting system.

What’s the max lock period? Not yet finalized, but Const suggested it would likely land around a year. Longer locks would need careful design to stay un-gameable.

Can locked tokens be transferred? No. Lock means lock.

Can users delegate conviction to others? Yes. If you’re a stakeholder who believes in a team, you can lock your alpha to their key. If a miner community grows large enough and disagrees with the owner, they can also organize to lock elsewhere, though that would take time.

What happens to locked stake if a subnet deregisters? It’s relinquished.

ETA? The locking primitive is targeted to ship within about a month. The governance-layer voting features will follow on a more cautious timeline and go through the standard Bittensor Improvement Proposal process.

Const closed the call the way he opened it, framing the moment as another step in a pattern. “Every time we introduce something like this, it’s a Cambrian explosion of innovation in this ecosystem.”

Watch the full Novelty Search session below: