This episode reveals Bittensor's strategic pivot towards a curated, competitive ecosystem, introducing sub-subnets for creative flexibility and deregistration to prune the network and focus value.
Introduction to Sub-subnets: A New Layer of Complexity
- Agrippa, a former video game producer and new OpenTensor team member, facilitates an AMA with Const to discuss major upcoming changes to the Bittensor protocol. The first topic is sub-subnets, a feature designed to allow a single subnet to run multiple, distinct incentive mechanisms simultaneously. Const explains that while this was technically possible for power users before, the new implementation makes it an explicit, transparent, and user-friendly feature on the chain.
- Initial Resistance: The feature initially faced pushback from advanced users who could already replicate the effect mathematically. However, the decision was made to prioritize ease of use and on-chain transparency for the broader community.
- Transparency and Creativity: Sub-subnets allow anyone to see the specific weights and emissions for each mechanism within a subnet. Const believes this will unlock creativity, enabling subnet owners to merge mechanisms from other subnets or run experimental and monetized versions in parallel.
- Technical Implementation: Each subnet will have a multi-dimensional weight and bond matrix, with individual emission and incentive terms for each sub-subnet. However, they will all share the same set of 256 UIDs. Const notes, "we really think that this is going to open up a lot of creativity in in how big uh subnets can be and what what they could focus on."
Sub-subnet Mechanics and Strategic Implications
- The conversation details the initial rollout and potential impact of sub-subnets on the network's structure and economics. This change introduces a new dynamic for how subnet owners can structure their projects and how investors evaluate them.
- Initial Limit: The protocol will start by allowing eight sub-subnets per subnet, a number chosen to provide flexibility without causing excessive chain bloat.
- Impact on Chain Bloat: While sub-subnets will increase the amount of data per epoch, Const suggests they might reduce the overall demand for new subnet slots, as teams like Templar could consolidate their multiple subnets into one.
- Strategic Use Cases for Investors: Const highlights a key advantage for investors: the ability to consolidate focus on a single token. A team can run a non-monetizable, open-source mechanism alongside a profitable, closed-source one, with the latter driving the token's value while the former sparks innovation. This is analogous to a corporation's R&D department being funded by its core business.
- Future "Calculus for Subnets": This is part of a larger vision to create a flexible "calculus" for subnets, including future operations like merging, forking, and destroying subnets.
The Introduction of Subnet Deregistration
- The discussion shifts to the more contentious topic of subnet deregistration, a mechanism to remove underperforming or abandoned subnets from the network. Const frames this as a necessary "garbage collection" feature to improve the health and focus of the ecosystem.
- Problem Statement: The network has a long tail of inactive projects, creating a secondary market for subnet slots and diluting attention. Deregistration aims to combat this "graveyard of subnets."
- Core Goal: The mechanism is designed to increase the quality and credibility of active subnets by focusing community attention and resources on fewer, better-run projects. This shifts the network's scaling strategy from horizontal (more subnets) to vertical (more depth within subnets via sub-subnets).
- Economic Rationale: Deregistration prevents the long tail of low-price subnets from diluting the emission share that goes to top-performing projects, ensuring capital is directed more efficiently.
How Deregistration and Liquidation Works
- Const outlines the mechanics of how a subnet will be selected for deregistration and what happens to the assets held by its stakeholders. The process is designed to be price-driven and automated.
- Selection Criteria: The subnet with the lowest price, calculated using a moving average, will be targeted for deregistration. This is analogous to how the lowest-priority miner is removed from a subnet.
- The Trigger: Deregistration occurs when a new subnet registers, taking the slot of the lowest-priced incumbent.
- Liquidation Process: When a subnet is deregistered, all outstanding tokens (held by owners, miners, and stakers) are automatically converted back into the TAO held within that subnet's liquidity pool. The TAO is then returned to the token holders' wallets.
Deregistration Economics and Fairness Considerations
- The community raised several questions about the fairness and economic impact of this new mechanism, particularly for new entrants and early subnet owners.
- Immunity Period: New subnets will have a four-month immunity period before they are eligible for deregistration, giving them time to establish themselves.
- Registration Lock Cost: The TAO locked to create a subnet will be returned to the owner upon deregistration, minus any owner emissions they have already received. Const explains this is a fairness consideration for early teams who registered under the assumption that slots were permanent. He states, "we don't think it's fair uh for us to not return that to registering subnets." In the future, this may change to a full burn.
- Registration Cost and Rate: The registration cost will start high (e.g., 2,000 TAO) and decay over a half-life of approximately half a week. This slow initial rate is intended to allow the ecosystem to adapt to the new dynamics without rapid churn.
- Incumbent Advantage: Const acknowledges that new subnets with lower liquidity will face a more competitive environment. However, he argues that the overall increase in network quality should raise the average price floor for all participants over time.
The Upcoming TAO Halving's Real Impact
- Agrippa asks about the anticipated effects of the first TAO halving in December. Const clarifies that the DTO (Dynamic TAO) mechanism was designed to be resilient to this event.
- Price vs. Volatility: The DTO mechanism adds liquidity at the current price rather than directly pushing the price up. Therefore, the halving (which reduces the amount of TAO available for liquidity) is not expected to directly impact subnet prices.
- Expected Outcome: The primary effect will be reduced liquidity in the pools, which could lead to increased price volatility in both upward and downward directions, especially for newer subnets. The fundamental price-setting mechanism remains unchanged.
Bittensor's Path to Decentralization
- The conversation covers the ongoing efforts to decentralize Bittensor at both the protocol and governance levels.
- Chain-Level Decentralization (Proof-of-Stake): A Proof-of-Stake (PoS) devnet is currently active. The team is building a full mainnet clone to stress-test the PoS implementation, with a potential launch next year. This is seen as crucial for the network's political and mimetic strength.
- Governance Decentralization: The plan is to evolve beyond the current Triumvirate model. A future system will empower subnet owners and a Senate to propose and veto changes. The immediate next step is to enable the cycling of Triumvirate keys, followed by implementing delay and veto powers for the other governing bodies.
The Yuma Consensus Debate: Yuma 2 vs. Yuma 3
- A deep dive into the Yuma consensus mechanism reveals a technical debate about the best approach to ensure fairness and prevent exploits like weight copying.
- Yuma 3 Explained: Ref, a community expert, is brought on stage to advocate for Yuma 3. He argues it is a superior implementation that fixes flaws in Yuma 2, works better to fight weight copying, and is fairer to smaller validators. He claims the only reason to use Yuma 2 is if a large validator wants to "disproportionally punish the small validators."
- Counterpoint on Yuma 3: Fish, another developer, offers a dissenting view, arguing that Yuma 3 "makes validating like a game" and opens up new potential exploits. He points out that validators can bypass EMA (Exponential Moving Average) filters in some subnets to gain an advantage, even if it goes against the subnet owner's intent.
Enforcing Real-World Value Accrual
- A critical question for investors is how to ensure that real-world revenue generated by a subnet flows back to its token holders. The speakers agree there is no simple on-chain solution, and different teams are pioneering their own models.
- The Challenge: Bittensor's design abstracts value, making it impossible to enforce revenue buybacks at the protocol level. The responsibility falls on subnet owners and the investors who back them.
- Fish's Approach (Subnet 51): Fish outlines two primary strategies: using external revenue for token buybacks or issuing dividend-like distributions. He notes that for now, this relies on trusting the subnet owner, but smart contracts could automate this for newer teams. He also mentions Stripe's beta for stablecoin payouts as a potential tool to streamline fiat-to-crypto buybacks.
- Ref's Approach (Subnet 89): Ref describes a more automated system where mined Bitcoin is converted to TAO via a CEX API and a DCA (Dollar-Cost Averaging) script, then staked into a wallet that never unstakes, effectively acting as a burn.
Conclusion
This discussion reveals Bittensor's evolution towards a more mature, economically robust network. The introduction of sub-subnets and deregistration provides powerful new tools for fostering quality over quantity. For investors and researchers, the key takeaway is the increasing importance of evaluating a team's specific strategy for value accrual and navigating the competitive landscape.
