This episode reveals Bittensor's strategic pivot towards a more sophisticated and competitive network, introducing sub-subnets for greater creative freedom and deregistration to enforce a new standard of performance and capital efficiency.
Introduction to Sub-Subnets: Expanding Subnet Capabilities
- Core Functionality: Sub-subnets allow a single subnet to run multiple, distinct incentive mechanisms, each with its own weight matrix. This turns a subnet into a "mixtures of experts" model, where different components can focus on different tasks (e.g., speed vs. accuracy).
- Transparency and Usability: This update makes the distribution of weights and emissions for each mechanism explicit on-chain. Developers can query the performance of each mechanism individually, which was previously obscured.
- Strategic Implications: This allows subnet owners to merge different functionalities, such as combining an open-source, non-monetized mechanism with a closed-source, revenue-generating one under a single token. Const notes, "We really think that this is going to open up a lot of creativity in how big subnets can be and what they can focus on."
- Implementation Details:
- The initial rollout will grant each subnet eight sub-subnets.
- While there will be multi-dimensional weight and bond matrices, the number of UIDs per subnet remains at 256.
- This change is expected to increase chain bloat, but the team believes the added functionality justifies the cost. It may also reduce demand for new subnet slots as teams can consolidate their projects.
Subnet Deregistration: Introducing a Performance-Based Filter
- The Problem: The long tail of inactive subnets creates conceptual clutter, consumes chain resources, and dilutes the attention of investors and developers.
- The Mechanism:
- Subnets will be deregistered based on the lowest price, determined by a moving average, similar to how low-priority miners are removed.
- When a new subnet registers, it will take the slot of the lowest-priced existing subnet.
- Holders of the deregistered subnet's tokens will have their holdings automatically converted back to the TAO in the subnet's liquidity pool.
- Strategic Goal: The primary goal is to create upward pressure on quality and performance. By filtering out the long tail, Bittensor aims to focus community attention and resources on well-run, active projects, thereby increasing the overall credibility and value of holding a subnet slot.
- Investor Impact: This fundamentally changes a subnet token from a potentially perpetual asset to one contingent on performance. New subnets will face a four-month immunity period before becoming eligible for deregistration.
Deregistration Mechanics and Economic Implications
- Incumbent vs. New Subnets: Const acknowledges that new subnets with lower initial liquidity will be more competitive and volatile. However, he argues that as the overall quality of the network increases, the average price floor for all subnets should rise, mitigating this risk over time.
- Registration Lock Cost: The registration lock cost will be returned to the owner of a deregistered subnet, minus any owner emissions they have already received. This decision was made to soften the blow for existing owners who registered under the assumption of a permanent slot. Const suggests that in the future, the lock may be fully burned, but not until the new dynamics are well-understood.
- New Registration Dynamics:
- The registration cost will start high (proposed at 2,000 TAO) and decay over a half-life period (initially proposed as half a week) if no new subnets are registered.
- The total number of subnet slots will remain at 128 for the foreseeable future, with increases contingent on the rising quality of the network.
Open Q&A: Decentralization, Yuma Consensus, and Value Accrual
- Decentralization and Governance: Const confirms that work on decentralization is a priority. A Proof-of-Stake devnet is currently running, with a full mainnet clone planned for testing. He anticipates a potential move to Proof-of-Stake next year. On governance, the team is designing a system involving subnet owners and a Senate to veto or approve changes, moving away from the current Triumvirate model.
- Yuma Consensus Explained: Community member Ref provides a detailed explanation of Yuma consensus versions.
- Yuma 2 introduced bonds to penalize validators who diverge from consensus, creating a stronger incentive alignment.
- Yuma 3 is presented as a superior implementation that fixes issues in Yuma 2, better incentivizes validators to recognize new top-performing miners, and more effectively combats weight copying. Ref states, "Yuma 3 is better than Yuma 2 in all dimensions."
- Value Accrual for Subnet Tokens: A recurring concern is how to ensure real-world revenue generated by a subnet flows back to its token holders.
- There is no on-chain enforcement mechanism. Value accrual relies on the subnet owner's strategy and the community's trust.
- Community member Fish, a subnet owner, outlines two primary methods: using external revenue for token buybacks or issuing distributions similar to stock dividends. He emphasizes that for now, this is a matter of trust and alignment between owners and stakers.
The TAO Halving and Future Economic Design
- Impact of Halving: Const explains that because the emission mechanism adds liquidity rather than directly manipulating price, the halving is not expected to affect subnet prices directly. However, it will reduce the amount of TAO entering liquidity pools, which could increase price volatility for all subnets, especially newer ones.
- Alpha Downscaling: A technical question reveals a key design choice in the DTO (Dynamic TAO Offering) model. The system scales down Alpha token injections but not TAO injections when the sum of prices is high. This provides a "cushion" or price support for subnets, acting as an incentive from the network to maintain a stable floor and encourage innovation.
Conclusion
This episode underscores Bittensor's evolution towards a more mature, performance-driven ecosystem. The introduction of sub-subnets and deregistration creates a dual dynamic of creative expansion and competitive consolidation. Investors and researchers must now assess subnets on their ability to innovate within this new framework while maintaining the capital efficiency required to survive.
