Bittensor’s Conviction Upgrade Changes Everything

Added: 2026-05-15

[00:00:00]A major change is coming to Bit Tensor.

[00:00:03]In fact, in recent times, we haven't seen a change as big as the one we're about to see. And so, in this video, I would like to explain to you exactly what this change is, what this means for Bit Tensor and how you can go ahead and take advantage of it. So, if that's something you want to watch, I think personally, if I was invested into a cryptocurrency and there was a major update, I would want to watch exactly what that change is going to be, then go ahead and watch the video. Let's get into it. So, this is Bit Tensor's conviction upgrading. Yes, my friends, it really does change everything. It's not just an update. It is Bit Tensor's attempt to make subnet ownership measurable, visible, and harder to abuse. So, there is a famous saying in the crypto space. It is do not trust, verify. And what that means is instead of just trusting someone for what they say, how about you just see it on the blockchain? How about if you want to lock up your money, give me the transaction hash. Let me see it. And that's exactly what they're trying to do here with a brand new upgrade for Bit Tensor called Conviction. So, we're going from trust me to prove it. This is very, very big because Bitensor is testing whether an open AI market can police itself. There was a controversy from a series of subnets. They all went under one umbrella. They called themselves the Covenant. where the owner of the Covenant decided to dump a heck of a lot of Tal about $10 million worth of Tal on his stakeholders. He did that in light of a massive technical accomplishment which was a very big deal for TAL called Covenant Covenant 72B. This was a 72 billion parameter LLM model. Now, all you really need to know is that it was such a big deal that the Anthropic CEO and the Nvidia CEO both acknowledged the accomplishment. That's how big of a deal it was because it was foreseen to be impossible to coordinate as much compute. Those are the little chips on your computer to the point where you'd be able to build as big of a model as they did in a decentralized manner. That's getting data from anywhere in the world. Because right now, the way that these big AI companies are doing it is they're building these mega data centers and they're doing it all there in this very centralized location. The problem is they can only kind of do so many chips before they run out of space or they run out of chips to actually put in. There is a massive supply issue in these chips. There's just not enough compute right now. on. While some of that is going to be alleviated by the end of this year, there are just so many problems right now uh with getting compute. In fact, if you wanted some very direct proof in retail, I recently picked up a Mac Mini and a Mac Studio to build my own AI agents. I'm going to be making videos on Agentic Trading. That's something I'm very interested in. Uh, but the point is, if you try to get a Mac Mini or a Mac Studio right now, either you're going to have to wait a couple months or you're going to have to get very, very, very lucky that a refurbished one comes up. There are very big supply shortages when it comes to these Mac minis and these Mac Studios.

[00:03:13]And that is retail's way of getting access to compute at a semiafordable price. I mean, it's still like $1,000 base price practically to get yourself a Mac Mini right now. very expensive but nonetheless there are very big supply issues when it comes to compute and the great thing that Covenant did was they made 72B. Now it was built on Subnet 3 which at the time was called Templar and when the founder exited you know a lot of Bit Tensor holders were like oh no I mean that was the cool thing that we had andropic co talked about us because of Covenant 72B so everybody started freaking out. There were a lot of holders that misunderstood what actually happened and sold. Now the context you need now to understand the situation in its whole is that despite the fact that the subnet owner left and dumped the actual technology and the people that actually built it didn't go anywhere. So So in essence, while yes, the guy dumped $10 million and he's unethical and you know he deserves what's coming for him.

[00:04:17]Um, what I have to say is it had no impact at all in Bitensor's ability to build a bigger model than Covenant was.

[00:04:27]And so what we're trying to solve is in essence we we have the tech layer secured. If the subnet owner leaves, it's not going to magically kill the tech. But what if we could make it more difficult for these subnet owners to dump $10 million on our heads? Why do we have to go ahead and just trust them? I get it. If you don't trust the founder, don't invest into the cryptocurrency.

[00:04:49]But what if I rebuttal? I gave you a rebuttal here. And my rebuttal was, well, why do I need to trust him in the first place? If we had a mechanism in place called conviction where I don't need to trust this guy and we can have it transparent on the blockchain, when is unlocked, we're going to happen.

[00:05:04]Winnie was able to dump on us. And that is exactly what Covenant has produced.

[00:05:12]He has they have produced a world in which we've created a new technology called conviction and I will explain it to you my friends. We're I'm going to explain it to you because there's a there's a new model coming up and it's very exciting. And make sure by the way my friends you're you're drinking your coffee black. Make sure you're drinking your coffee black.

[00:05:28]I like my coffee black because it's just so much healthier. You get the raw coffee beans. But anyway, make sure that um you understand that Covenant 72B while it was cool and it was a very big technical accomplishment um just because a subnet owner decides to leave doesn't really impact anything at all. It really doesn't. And it's I think it's been a month, right? It's been about a month since this controversy happened. We're already starting to build a bigger model. We're already building a bigger one. It started at a 1 billion parameter model and then they scaled it to 8 billion and I I think somewhere along the line they went to 24 billion. Now they're going to 80. So remember Covenant was 72, Tutonic is 80. It's marginally larger. They're just proving that we could build a larger model and we could do it better than you ever did Sam. And so just so you can understand what a subnet is in layman's terms, very simply put, a bit tensor subnet is an AI market. So there are 128 of these. All of them are doing very cool things in their own respect. They have three things, three little pillars here that make up a subnet. One is the miners. The miners are all doing a cool task. So they provide useful AI work. Whatever that task is, there's 128 of them. All of them are tasks. There are validators which measure the quality of that task.

[00:06:48]Let's make sure you're actually doing that well. Let's make sure you're doing the best. And the people that do that task the best get emissions. they get rewarded for their performance. You can see this loop, this very hyper capitalist loop where only the best survive. And so naturally, you're going to be getting the best results. And so the magic of Bit Tensor isn't one model.

[00:07:08]It's an open market for useful machine intelligence, which is really cool because open markets are great, but they still need accountability. If anyone can launch, attract emissions, and build hype, the network needs a way to separate long-term builders from those shortterm extractors.

[00:07:28]And so before the attention game could outrun alignment, if you could gain heat, let's say like you become the most exciting subnet of the day and you start pumping 50% of a single day, well, that's great and all, but the only problem is as a subnet owner, you can kind of just press the big old sell button and get out. Now, that's a big risk because owner liquidity matters.

[00:07:49]They get about 18% of all rewards and so we need to have a visible commitment layer and that's a very very big deal.

[00:07:58]So, this is a couple of the details here on Covenant 72B. It was a 1.1 trillion token model. It was listed on hugging face. We had that model scale and it's not that Bitensor failed, but that Bit Tensor needs better market accountability. This is something that it's not only a Bit Tensor problem. This is every single crypto chain. All crypto chains have this problem. How do we trust the founders? And so the cool thing is these are screenshots directly from the mouth themselves. This is Con, founder of Bit Tensor. Check this out yourself. Please uh do not trust me. I am just a YouTuber. I'm just a content creator. Yes, I might be an investor and I might have been full-time for 5 years so far. I'm probably going to be here for many more years. Um but remember, I am not as technical as Consta. So, please do the verification yourself. Go to tutonic.ai, check it out yourself.

[00:08:47]Um, const is building out the biggest right now decentralized training model ever. That's what Tutonic ADB is going to be. You can check the website yourself. The point is not everything is fixed. The point is that Bitensor can keep producing new experiments after a messy chapter. So, what is conviction?

[00:09:02]What is this big new change coming to Bit Tensor? Conviction turns support into a visible commitment signal. So, buying alpha shows interest. Locking alpha shows conviction. So if you buy alpha, which is the subnet tokens, that's cool. It's like, I'm interested in you. But if you lock that alpha, it shows you've got some conviction. And the longer you lock it for, the more conviction accumulates.

[00:09:26]And so that's the general idea of what conviction is. Users can lock Subnet Alpha to what's called a hotkey to show long-term support. you know, typically speaking, the person with the most uh the locked subnet alpha for the longest you because they have to build that conviction um ends up getting control over a subnet. And so you can imagine if you're doing a great job and you know your name is John, you're doing a great job. Shout out to this fictional character I've created. Some of you are like, "Who's Timmy? I don't know who Timmy is." Timmy is a It might be an AI agent I've created. Uh but um let's say you're John, right? And you own a subnet. You might be a little scared.

[00:10:09]You might be shaking a little bit because you haven't been doing a good job lately. You have been doing a good job and people have decided to go with uh, you know, maybe a little Timmy. Go with little Timmy cuz he's a he's an 18-year-old super genius and he's he's got a better solution than you do for your subnet. And also he's willing to hold for longer than you are. And so backers of that subnet, maybe the biggest holders, the whales, they can all decide to shift their stake over to little Timmy instead of you and you lose control. Oh no, that's bad. But also, you weren't doing a very good job in the first place. So yeah, it was probably time to go to the next guy. And so that's the general idea. Or maybe you're a bad actor and you want to you want to go ahead and dump everything right now.

[00:10:51]You want to dump everything. You want to make sure that this subnet burns to the ground as Sam has shown is, you know, an actual thing that subnets may be thinking or at least submit owners may be thinking. Uh well, you're not going to be able to do that. Well, you might be able to dump everything eventually, but uh you know, you're going to lose control a heck of a lot faster. And also, like if you do actually maliciously take over the sub, everybody's just going to dump before you get your tokens. Oh no. Oh no.

[00:11:20]That's not good. That's not good. So, the owner emissions are the key. So, let's say you get owner emissions.

[00:11:27]Remember, they get 18% of all of the emissions on a bit sensor subnet. Well, 100% of those emissions, all of their emissions are automatically locked into what's called conviction. Very cool. It doesn't remove bad actors. you know those guys could still be very bad actors but it makes extraction slower more visible and easier for the market to price. So instead of someone dumping on your head overnight which personally is something that institutions and I myself being myself and I and I imagine you do not want now they have a system in which it linearly goes over time.

[00:12:06]It's called vesting. If you if you've ever um invested in angel rounds or KO rounds or you know seed rounds, you're going to understand what I'm talking about. In the VC world, it's very streamlined. They're just bringing that over to Bit Tensor. And so I asked um I I was trying to do the research on this and and I and I figured you'd have this question. I'm like, does that mean that they can just grab all the emissions and just say like, okay, I want to sell all of them now. How does this work? So this is kind of how it works. Every single day, the subnet owner earns ongoing rewards. So they earn their 18% share.

[00:12:44]Now 100% of those automatically get unlocked to the owner's hotkey. Every single times they get a reward, they get conviction that has about a halflife of 62 days. Now there's no automatic release after a year. It's going to stay unlocked until they personally press the button to unlock it. If they press the unlock button, then there is vesting involved, right? There is a timeline involved before they can sell it. It's it's over time instead of all at once.

[00:13:21]And so if they unlock it, their conviction drops immediately. So let's say they want to fully exit out of the ecosystem. One, they're not going to be able to immediately do that. But two, if they lose all of their conviction, they're going to lose control of the subnet. And so we don't have a guy that's going to be dumping over the course of a year and also has control of the subnet. He'll lose control and then he'll be able to dump or do whatever he wants over a more predictable time frame. Owner emissions become a rolling public commitment stream, not an instant exit wallet. So this is kind of how we can see it. So from zero days, you can see that they don't get anything. Let's say they have 10,000 locked alpha that they want to go ahead and and go ahead and and release and and get score for.

[00:14:02]So, they're only going to get full score on that after about a year. As you can see there, the halflife at 62 days, it goes to about half of that stake. And so, commitment is earned over time. It's not instantly claimed, right? You don't get the full weight of all of your stake until about a year's time, which incentivizes people to lock their alpha, especially if they're subnet owner over time to maintain control of the subnet.

[00:14:29]Unlocks become public before liquidity becomes fully available. So, when alpha unlocks, conviction drops immediately, but the unlocked alpha doesn't become fully sellable at once. It doesn't become fully sellable. This is the delay on their cells. So if you did it right now, it would take you about 21 days to get 50% of whatever you want to dump as a subnet owner. After 3 months, you'll get about, you know, 95% of whatever it is you wanted to dump, but it takes you 3 months. That's plenty of time as a holder to react. It's going to be way more time to react, which is exactly we want as an ecosystem. The change is, you know, if you're doing bad behavior and you want to dump everything, well, it's going to it's going to become a little bit more difficult because it becomes harder to hide. We now have it on the blockchain. Do not trust verify.

[00:15:19]Conviction isn't some kind of morality machine. I don't want you to dog after um subnet owners for uh saying that they need to dump some of their emissions.

[00:15:27]Let's say, for example, a subnet owner needs to pay their team or maybe they deserve to make some money. I mean, let's be honest, if they're a subnet owner and they make a hyper profitable product, they do deserve to take some profits. So, I'm not saying that if anybody decides to sell some of their emissions as a subnet owner, they're bad people. you know, they're running a business at the end of the day. But if they do decide to maliciously dump everything, it's going to take some time. It's going to take some time and it gives a lot more time for people to react. It gives actual time. The system does not need perfect people. It needs incentives that make misalignment more expensive. Now, my friends, if you are going ahead and doing long-term trading, uh, one of the places that recently I've been looking into is a crypto IRA platform. That's an individual retirement account. Yes, my friends, you have to think about it at some point. I know a lot of you are maybe around my age, maybe younger. No, I'm kidding. I know a lot of you guys are around uh around the around the age where you should probably start thinking about that. And I myself have started stacking. It's something that as tax season came along, I realized that there are tax benefits for it. They are the biggest crypto IRA platform in America.

[00:16:37]And also, they offer stocks, they offer gold, so you kind of have that one-stop shop for all of that. They have a $100 funding bonus for clicking my link in the description. And also, if you're doing leverage trading, I currently have uh all of my leverage trading happening on BTCC. They have some great bonuses for people that are actively trading. I believe if you sign up, you'll get a 30% trading fee reduction at least for the first month that you sign up. So, take advantage of that. And let's get back to conviction because it's about to get real interesting. How do you keep ownership? How do you keep ownership?

[00:17:10]So, Talstat, which is where I got all of this conviction information. Shout out to Talstat. Says that the hotkey with the most conviction is the subnet king.

[00:17:20]Today, that has no label. Now, there are a couple things happening. Future discussion suggests it could matter for ownership, but today that label has no active power. Right now, it doesn't matter. I'm assuming it's because they're trying to figure out how to do this properly. Uh, but if implemented, owners may need to keep proving alignment to stay in control. So, they're going to have to keep that alpha stake to prove that they deserve to continue to own that subnet and that they're not just here for a quick buck.

[00:17:47]Now, there is something else that I found that I think is interesting. Um, a a bit complex, a bit complex, but I think is important. If you're holding towel, you should know about this. So, there is something coming, right? Uh, there was a hot fix to remove what's called initial owner alpha. What that means is that the amount of alpha that an owner initially needs to have to have a new subnet is actually less than it is right now. It's going to be less which means that the root proportion is going to be higher. So all you need to understand is that if you stake on root all that means is you're staking bit tensor you're staking to you're staking the coin. Now if you're staking alpha you're staking whatever the subnet is.

[00:18:30]That's that's all you need to understand. So if you're not invested to the subnet, you're staking on root. If you're staking right now, you're staking on route. It doesn't matter if you're on Kraken or wherever, you're staking. If you're staking, you are on route. And so new subnet launches become less owner skewed. So more of the early emissions actually end up routing through Routt.

[00:18:49]That means that the subnet owners are getting less rewards initially kind of to prove themselves and uh people that are sticking on root um they get more rewards as a result. So we should be expecting a slightly higher APYs on which which is like the amount that you're making per year. So instead of like 5% you might get a little bit more than that. Um we should be expecting slightly higher APYs. This is a bit of an explanation that I was able to create. So root proportion is calculated using tow and root alpha issued and tow weight. If less alpha is issued at launch, the root side of the split can stay larger. So this is kind of the the old split was 50 to 70% was root and then the rest of it was alpha exposure.

[00:19:37]Now it's going to be about 75 to 90% and uh the rest will be alpha exposure.

[00:19:42]Okay. So this is the split that you should watch if you're staking regular talon root because you might be making a little bit more because of that. Now that's how in my opinion bit tensor mature matures. I think this is great. I know a lot of people don't like conviction. I I' I've seen Al got on the timeline. He says, "Oh, but this is going to destroy the chain because who cares?" But I'm also like, why not? Yes, you know, submit owners deserve to make money. I agree with you. But also, immediate cash flow, there are ways to go around that. There are ways to go around that. There there are plenty of raises that happen all the time in the OTC markets that have vesting involved.

[00:20:21]That's over the counter. That's like it happens all the time. Even Bit Tensor did OTC rounds. All of these chains do OTC rounds with these venture capital funds. They give them a discount and say, "Hey, you have to hold for the next uh I don't know how many months or I don't know how many years. Maybe it's one year and then we'll give you your money over time and maybe you'll provide some value for me." And that's how they kind of uh conceptualize or rationalize that relationship and they're able to raise money. So there there are ways around this. There are ways around this.

[00:20:47]And so this is for me a definite win for the Bit Tensor community. Um, you know, I'm thinking this is going to be great.

[00:20:55]I I think that subnets need to mature. I don't want to see this world where they all just become pump fund memecoin garbage coins. We need to make sure that this AI work is useful. We need to make sure that the subnet owners are held accountable. And also, we need to prove resilience. Tutonic building out ADB is going to show new builders that the network can keep pushing after controversy. And so conviction doesn't solve everything. It doesn't make bad people good inherently. It might make it more difficult to be a bad person, but well, you could still be a bad person. A delayed exit is not the same as preventing bad behavior. They could still dump on you. It just takes them a little bit longer. And also uh the second risk is Subnet King is not fully live. Um so you can't get ownership just yet. Um so yeah, we'll have to see where everything goes. But this is this is a big this is a big moment for the Bit Tensor ecosystem if they're going to be providing a serious change for institutions to build it. We need to have easier ways to go ahead and verify commitment to building out these subnets. And I I don't want to be a person to trust, especially if I'm a big venture capital fund with $30 million under my wing. Well, I guess that's a mini one, but you know what I mean. Like maybe I've got a a billion dollars in assets under management and I'm going to be pouring some of that to Bit Tensor. I don't want to have to go ahead and trust these subnet owners. I want to just have this on chain. I want to have a fail safe so that I can get out before some of these bad actors decide that's uh a good opportune time to leave. The upgrade doesn't make the subnet economy perfect. I think this chain is far from perfect, but it does make alignment more visible, exit slower, and accountability easier for the market to price, and that's great. I think accountable AI infrastructure is absolutely necessary.

[00:22:46]I hope this video served as a great guideline serving surrounding conviction, the biggest change in recent Bitensor history. It's slated to go live, I believe, on a Wednesday, so pretty soon. I will see you all on the next one. Thank you so much for watching this video. If you're into AI content, crypto content, agentic trading, make sure you follow.