Introducing the Masa Bittensor Subnet
Masa Brings LLM-Powered AI Data Network to Bittensor, launching the First Dual-Token Reward Model (MASA-TAO)
Introduction
We are beyond excited to announce the launch of Masa’s AI Data Subnet on Bittensor. The launch of the Masa Bittensor Subnet unites the technology and community of two of the pioneering forces in decentralized AI. Masa is now bringing the world’s data to the Bittensor’s global brain.
Bittensor is a decentralized, peer-to-peer artificial intelligence network designed for the creation, sharing, deployment, and training of machine learning models. With its sophisticated TAO economic model that incentivizes the decentralized production of high-value artificial intelligence, Bittensor has become a force to be reckoned with. Since its launch in March 2023, Bittensor has grown into one of the first $10-billion AI ecosystems, with institutional validators — such as Digital Currency Group’s subsidiary Foundry and Polychain. Currently, there is 5.7 million staked $TAO in total, worth $1.48 billion.
Masa is the decentralized network for Fair AI, where people earn by contributing data. Masa empowers AI developers to build anything, anywhere with the world’s data. Masa is backed by Digital Currency Group, Anagram, Animoca, and incubated by Binance and Hashkey. Masa was the first AI project to debut on CoinList Launchpad in 2024, with a viral 17-minute sale. We are on a mission to create Fair AI, powered by the people.
Bittensor rewards miners, validators, and subnet owners with daily emissions of 7,200 TAO, equating to over $1.8-million dollars distributed across the subnets. The Masa Bittensor Subnet launch brings hundreds of our ecosystem AI developers to Bittensor. Masa is also the first live token project to launch a Bittensor subnet. To incentivize workers and validators to contribute to the Masa Bittensor Subnet, Masa is introducing the first dual-token reward model in the Bittensor ecosystem — giving everyone an opportunity to earn MASA & TAO double emissions. We aim to build one of the highest performing, most active, and highly utilized Bittensor subnets within the ecosystem.
Bittensor’s global brain is about to get a whole lot smarter, with the arrival of Masa’s global data network — ushering in the next generation of AI.
Introducing the Masa Bittensor Subnet
Data is the new currency of the AI economy. With many proprietary and open source options, general purpose LLMs are converging, becoming commoditized, and are in a race to the bottom. The most valuable AI applications will be the most specialized ones that leverage real-time data for up to date context and subject specialization. Specialized AI applications require huge amounts of specialized AI training data that is constantly up to date.
The Masa Bittensor Subnet provides real-time and static, structured, annotated, and vectorized data from a variety of data sources critical for AI development, such as X (Twitter), Discord, diarized speech (e.g. podcasts, YouTube, TikTok), gated web data (e.g. New York Times), and public web data (e.g. Google Search).
Real-time data can be used to build robust datasets or used directly in system prompts, providing up to date context to LLMs. Static data sets are constantly updated and stored by subnet workers for further processing into vectors to fuel Retrieval Augmented Generation (RAG) in AI agents. These data sets are processed and annotated using agentic data pipelines that employ fine-tuned LLMs trained on JSON and other formats to deliver high-quality outputs from volatile data inputs.
AI developers have been using Masa data for a wide range of use cases, such as capturing trading signals and building hyper-personalized AI companions.
The Masa Bittensor Subnet Architecture
Masa enables people from all over the world to contribute data and compute to AI development, without centralized control. The Masa Bittensor Subnet operates through three main types of participants:
- Workers (Users): Users supply data and bandwidth, powering AI training and processing. The Masa and Bittensor communities can participate using low-power devices to run a Masa worker node to scrape, structure, annotate (using LLMs), and vectorize data from laptops, servers, or, in the future, mobile devices by contributing compute and bandwidth from anywhere in the world.
- Validators: Validators maintain the integrity of the data network by validating transactions, data contributions, and AI outputs.
- Oracle Nodes (Developers): Developers utilize the AI data and services from the network to build applications. Masa allows AI developers to build anything, anywhere with the world’s data.
Similar to Bittensor’s incentive mechanism, Masa contributors — validators and workers — are rewarded based on the value of their contribution to the network, using game-theoretical frameworks that optimize a contributor’s utility on the network. This ensures an effective system that uses economic incentives to drive growth and the equitable expansion of Fair AI.
The Masa Bittensor Subnet Incentives: MASA-TAO Dual-Token Rewards
Participants in the Masa Bittensor Subnet are rewarded for their contributions with dual-token staking rewards in $MASA and $TAO, providing both economic incentives and rewarding high performance on the subnet.
$MASA is the first and only live token for any subnet in the Bittensor ecosystem. To enhance the incentive and reward mechanism of the network, validators, oracle nodes and node workers can earn dual-token staking rewards in both $MASA and $TAO.
Masa Foundation-owned TAO from operating the subnet will be used to support $MASA through buybacks or distributions as part of the emissions schedule.
The Mathematics and Technical Whitepaper Behind the Masa Bittensor Subnet
The Math Behind Masa
At the heart of Masa lies a robust mathematical framework that ensures its smooth operation and continuous improvement. Here are the key concepts:
- Staking Mechanism: Validators, Worker Nodes (users), and Oracle Nodes (developers) must stake tokens to participate. The amount of tokens staked determines their eligibility and influence within the network. The staking requirement adjusts based on performance, meaning better performers earn more while maintaining influence, creating an incentive for high-quality contributions.
- Performance Evaluation: Validators are scored based on their contributions to consensus activities, data verification, and evaluating the outputs of AI models. This scoring ensures that only the most effective participants remain active, driving continuous improvement.
- Reward Distribution: Participants earn rewards for their contributions. For Worker Nodes, this means providing high-quality data and computational power. The rewards are calculated using a combination of factors such as data relevance, coherence, factual accuracy, and response time of data and AI outputs.
- Slashing Mechanisms: To discourage poor performance or malicious behavior, participants can lose a portion of their staked tokens (slashing) if their contributions do not meet the required standards.
How Masa is Self-Improving
Masa’s design inherently promotes continuous improvement through several mechanisms:
- Incentive Alignment: By tying rewards and penalties to performance, Masa ensures that participants are motivated to continually enhance their contributions. This alignment of incentives fosters a culture of excellence and innovation.
- Adaptive Staking Requirements: The staking requirements adjust dynamically based on ongoing performance evaluations. This means that as participants improve, they can earn more while maintaining their influence, encouraging sustained high performance.
- Collaborative Environment: The network’s decentralized nature promotes collaboration and the sharing of resources, which leads to the collective enhancement of AI models.
- Mathematical Foundation: The use of advanced mathematical model ensures that the network adapts to changes and remains stable and fair over time.
To dive into the mathematics behind the Masa Protocol and the Masa Bittensor Subnet, please refer to the Masa Technical Whitepaper.
How to Participate in the Masa Bittensor Subnet
Join the CoinList x Masa Bittensor Subnet Incentivized Testnet
Masa Bittensor Incentivized Testnet is designed to test the Masa Bittensor Subnet’s capacity, resilience, and performance. Miners will provide data and compute resources to the Masa Bittensor Subnet.
Get Started Now > Join the Incentivized Testnet
The top-performing 256 miners at any given point in time will earn a share of the 3,000,000 MASA token rewards pool. The higher your performance, the more rewards you earn. Performance is based on the amount of data, quality of data, and compute processing power miners provide to the network. All rewards are calculated programmatically over the entire duration of the testnet campaign.
How to Participate:
- Open Participation: All participants are welcomed to join the Incentivized Testnet
- Performance-Based Rewards: The top-performing 256 miners at any point in time will earn a share of the 3,000,000 MASA rewards pool. All rewards are performance based and programmatically calculated
- Hardware Requirements: 2 vCPUs, 8GB RAM, 10GB storage, open ports 4001 (TCP/UDP) and 8080 (TCP); No KYC required
- Rewards: Final rewards will be published on a public Dune Dashboard on August 7, 2024. Rewards will be delegated to a validator and distributed at Masa Protocol mainnet launch, target Q4’2024
Key Milestones: The Masa Bittensor Incentivized Testnet runs from July 9th to July 23rd, 2024.
- July 9, 2024: Registration opens, testnet begins, and miners will start earning rewards
- July 23, 2024: Testnet concludes
- August 7, 2024: Masa Bittensor Subnet testnet final rewards will be published on a public Dune Dashboard
- Q4’ 2024: Rewards will be delegated to a validator and distributed at Masa Protocol mainnet launch, target Q4’2024
Get Started Now > Join the Incentivized Testnet
Trade MASA / wTAO Pair on Uniswap
$MASA is the first live token in the Bittensor subnet ecosystem. Users can now trade the MASA/wTAO pair on Uniswap, further integrating Masa into the Bittensor ecosystem and expanding its liquidity and reach.
In Conclusion
At Masa, we believe Bittensor will become the Ethereum for decentralized and Fair AI. The Bittensor ecosystem is at the early stages of growing into a trillion dollar market driven by decentralized and Fair AI. With the launch of the Masa Bittensor Subnet, Masa will exponentially accelerate the development of Fair AI with the world’s data.
Join us in our mission to create Fair AI, powered by the people. For more information and to participate, visit the Masa Bittensor Subnet Incetivized Testnet campaign on CoinList and follow us on X.
