Conviction: Bittensor’s Protocol-Level Jurisprudence

Full article credit: @KJ__Morris

In Novelty Search’s 71st episode, Const addressed the topic of rugpulls (or exit scams) within the blockchain space. He noted that “the difference between [Bittensor] and everybody else is we’re expected to solve this problem, and nobody else is”.

A rugpull or exit scam is where a team or developer suddenly departs from an ecosystem or abandons a project, leaving investors financially harmed. The issue exists in many forms, but is especially rampant in the world of web3 and blockchain tech. These distributed, global economic systems are a hotbed for malicious social dynamics – specifically enticing audiences to invest in their projects which dramatically raises token prices, only for the internal team to sell their portion and suddenly jump ship.

Rugpulls happen all the time. They’re initiated on Ethereum and Solana on a seemingly weekly basis. They happen on Bittensor, too. On April 10 2026, Covenant, one of the larger companies on Bittensor, announced its sudden departure from the ecosystem. The project, led by Samuel Dare, revealed that its three subnets (Templar, Basilica, and Grail) would no longer be functioning within the protocol.

There’s several factors that made this so shocking. Samuel Dare was not a random figure whose prominence emerged unexpectedly – he had a personal relationship with Const. He was an Engineering Manager at the Opentensor Foundation in 2024. Const had afforded him many opportunities throughout his time in the protocol.

Not only that, but just one month prior, Templar, the AI training arm of his company, claimed to have successfully pretrained a 72-billion-parameter model on the subnet. The news went stratospheric. The crowd went wild. The price went up. Not just of Templar’s alpha token (SN3), but the price of TAO, too. The network effect kicked in. Subnets unrelated to Covenant saw price increases. Bittensor received more attention. Everybody was happy.

It wasn’t just that somebody had pretrained a 72B model on the protocol, it was that somebody close to Bittensor’s core development had done so. It magnified the trust signal.

When Covenant rugpulled, it wasn’t just its subnets’ investors who were hit – everybody in the protocol was. Just as the team brought the price up, their vanishing act brought it down.

When Const spoke about rugpulling on Novelty Search, you could hear in his voice just how personal this all felt. It’s not the protocol’s first time being conned: prior to Covenant, the largest exit scam was Tenex, formerly SN67. However, they were small players compared to Sam’s company, and their ties to the protocol were far weaker. Const actually worked on the codebase for Templar, bought Covenant’s subnet slots for them, and worked on the website designs.

So when he was live on Novelty Search speaking about solutions, it was contextualised by two crucial aims: a desire to help those who were harmed by Sam’s actions, and a desire to reclaim what had been taken from the ecosystem – trust.

What’s so striking about Const’s discussion on Novelty Search is that he treated the problem of rugpulls as similar to any other type of exploit, giving it the same weight as a conventional hack. For instance, he compares it to moments like Bittensor’s 2024 attack, which triggered $8 million USD in losses.

That’s almost unheard of in the blockchain space. Rugpulls are typically treated as a price one simply pays for a decentralised economy. If you choose to operate in a space that exists outside of most legal systems, then you’re expected to accept the consequences.

This notion is being challenged.

In the wake of this, the Opentensor Foundation announced a new feature within the blockchain, Conviction.

What is Conviction?

Conviction is an architectural solution to a social problem. It asks: How do you stop bad actors from abandoning their projects and running away with the profits?

In the physical world, the justice system stops you, with court cases leading to fines and prison time. In a distributed, digital space, that system is much harder to reach, as the technology is still so novel, and the law tends to be too arcane.

However, Conviction is designed to prevent rugpulls natively within the Bittensor blockchain. Subnet owners can lock away their alpha tokens to the protocol, and over time, this will secure their control of the subnet. Your stake locked away, along with the time you lock it away for, gives you a Conviction score, which acts as a measure of how much you believe in your own project, signalling that you intend to be in the ecosystem for a long time.

The community can lock up their alpha tokens, too. They can allocate them to the same key as the subnet owners, raising their Conviction score and helping them cement their control. Or, they can assign them to their own key, or somebody else’s, and if another party builds a higher Conviction score than the owner, then they take control. The 18% emissions that subnet owners receive will be redirected to the key with the highest score.

For an in-depth understanding of the underlying mechanisms, take a look at the Learnbittensor.org explainer here.

Who Cares About Human Problems?

There’s this idea within the blockchain space that you can’t account for human failure.

“Don’t understand how to mine or stake an asset?” That’s on you.

“Sent coins to the wrong wallet because you’ve never used a crypto wallet before?” Too bad, this isn’t TradFi.

“Got scammed by someone who abandoned the project you invested in?” That’s just the way things are.

It’s a defeatist attitude, but more than that, it’s a disrespectful one. It dismisses the everyday user, in favour of embracing a learning curve. Educational barriers are not just about how to use a technology, they’re also related to how you interact with a technology, how you make reasonable decisions, and how you become an informed agent.

Now, Bittensor isn’t a perfect network. One of my biggest gripes with it is the high level of complexity baked into the ecosystem, relative to the low level of educational material. But I will say this: Conviction is an upgrade that actively appreciates the personal and moral aspects of engaging in these spaces.

Nobody would have demanded this. Nobody would have been frustrated if it didn’t exist. Accepting a rugpull is the norm. In Const’s own words:

“Nobody has done this because I don’t think anyone has really thought about doing it”

He’s not saying it’s an intellectually superior concept, he’s saying others in the blockchain space don’t treat this as within their jurisdiction. They see it as a social problem – something messy and awkward and unbecoming of an architectural or technological change.

Hearing about Conviction reminded me of how I felt when I learned about human-readable addresses. The concept has been around since Namecoin in 2011, but I was especially impressed when I saw the NEAR Protocol build them directly into their codebase. They embraced the simple fact that crypto is overwhelming, and tried to mitigate that for the sake of the everyday user.

Conviction doesn’t make Bittensor easier to understand. In fact, it raises the learning curve. But I place it in the same conversation because they both solve existential problems that technically do not need solving, yet everybody could benefit from.

On-Chain Jurisprudence

In his 1999 book, Code and Other Laws of Cyberspace, legal scholar Lawrence Lessig coined the famous phrase “code is law”. It was perhaps a little ahead of its time, as these three words didn’t ascend into public discourse until the blockchain industry was born.

The advent of cryptographic ecosystems, smart contracts, mining pools, and eventually subnets gave this concept a life of its own. It became a statement of rebellion against conventional centralised systems, and a battlecry that technology and mathematics can be enough to govern society. We don’t need hierarchies imposing rules, we can write and execute them ourselves.

But in a space where code is law, what exactly is the rule of law?

Even before computers, the rule of law has been a nebulous matter, one that people still write lengthy papers on to this day. The term was popularised by British jurist A. V. Dicey, but its roots date back to ancient Greece. To give an extremely distilled definition, the rule of law is the set of guiding principles that should influence the laws of the land – it’s the ideologies that drive the decisions, and the motivation behind the methods that enact them.

It’s the meta-level of the legal world, not focused exactly on the laws created, but more so on ensuring they’re applied in a fair and reasonable way. Generally, blockchains ignore this part. They’re focused on accelerating their progress. They want to build tools that grow their ecosystems, and empower their users. They want smart contracts with novel capabilities, experimental AMMs, and tools that facilitate innovation.

The Opentensor Foundation wants these things, too. dTAO was arguably built for accelerating market dynamics and providing more independence to facilitate innovation across the ecosystem. The difference is that, alongside developments like this, it’s also facing inward at the social and moral behaviour present, expanding its infrastructure to prevent harm.

Conviction is a new law, in response to an era-defining precedent. But driving Conviction is Bittensor’s own rule of law: that nobody should be above consequences, not even subnet owners.

The Purpose of Mutability

On Novelty Search, Const notes that everything that’s incentivised in Bittensor is subject to change. This includes subnets themselves, validators, and miners. With one exception: “the owner key is not [mutable], it’s fixed”. There’s no function within the protocol that facilitates the handing over of subnet owner keys to someone or something else, should they underperform or act maliciously.

Conviction doesn’t exactly make owner keys mutable, but it does something effectively similar by preventing them from receiving rewards (currently set at 18% emissions), redirecting them to other committed members. Therefore, the owner key no longer unlocks remuneration.

What’s especially important about Conviction is that Const himself is also subject to it. As a fellow subnet owner, he could theoretically lose his 18% from Subnet 120, Affine. It’s highly unlikely, both because Const is so well-respected, and because he owns so much TAO that gaining access to subnet alpha would be easy, but the mechanism exists and there isn’t an exception inside the code that saves him.

Let’s imagine an outlier scenario for a second. In 2016, it was uncovered that the South Korean President Park Geun-hye, was being puppeted by the Church of Eternal Life cult, where individuals, namely cult representative Choi Soon-sil, were controlling the president’s speeches and policies. Park Geun-hye was the first president in South Korea to be impeached. She was not above the rule of law.

If it were ever uncovered that Const was being highly influenced or even controlled by disconnected third parties whom the community distrusted, they could theoretically band together and attempt to build a higher Conviction score than him on Affine, therefore rendering his owner key somewhat inert. And with Const eventually stepping down from his position as Opentensor CEO, this law has the potential to seriously limit and restrict his power.

Losing a subnet due to a Conviction score would also count as a serious vote of no confidence from the community, affecting his social influence (as it would for any subnet owner). This is a strong sign that the rule of law exists within Bittensor, and that Conviction upholds and exemplifies it.

Final Words

For years, blockchain enthusiasts have cried that the industry should regulate itself, rather than have centralised traditional laws applied to it. Yet, most developers ignore the moral obligations that come with doing so.

Conviction is the first time I’ve seen a blockchain actively try this. If successful, it has the potential to change how we think about the industry, and could even set a new standard for decentralised governance. While the Covenant rugpull harmed the industry, it’s also provided an opportunity for maturity, one that’s being actively seized.