Key Insights
- Autonomys Network leverages a modular blockchain architecture built on the Subspace Protocol, creating the needed infrastructure for AI3.0 applications.
- To achieve high scalability, Autonomys Network combines Decoupled Execution (DecEx), modular domains, data sharding, and a multi-layered Distributed Storage Network (DSN). The DSN also enables AI Agents to store and retrieve data directly from the blockchain.
- The Proof-of-Archival-Storage (PoAS) consensus mechanism called Dilithium, combined with a Proof-of-Time (PoT) mechanism, solves the “Farmer’s Dilemma” by allowing farmers to maintain consensus without storing the entire blockchain history, ensuring data availability, security, and efficient storage management.
- Auto ID and Auto Score create a self-sovereign identity framework and a proof-of-personhood mechanism that facilitates verifiable human-AI interactions.
- Space Acres simplifies participation in Autonomys Network by providing a user-friendly application that allows anyone to run a Farmer node on their computer to earn AI3 tokens.
Introduction
The rise of AI has absorbed the majority of mindshare in the market over the past few years. From the launch of GPT3.0 to onchain AI Agents, there has been a flood of interest in how AI will change the way the world operates. A pressing question for AI is where it will fit in the scale of decentralization. Many of the leading foundational models are built in a centralized and permissioned way. The need to democratize access to AI is a mission that large companies, like Meta, have aligned with by open-sourcing Llama3.1. The last time we saw a shift of this scale was the creation of Bitcoin by Satoshi Nakamoto in 2009. To live in a world where access to transformational technology is free to use by all, more projects need to join the likes of Meta and Bitcoin and find the value in being open-source and permissionless.
Autonomys Network aligns with Nakamoto’s vision of a permissionless and secure blockchain, as it aims to apply this vision to building out AI3.0 infrastructure as a public good. AI3.0 will be an open, collaborative, web3-enabled approach to AI where humans can customize, train, and deploy their own agents to act on their behalf. Evolving from AI1.0 of centralized machine learning, to AI2.0 of centralized generative AI, to now AI3.0 being decentralized human-centric AI.
Autonomys Network, built using the Subspace Protocol, is positioned as a foundation for AI3.0, providing the necessary infrastructure and tools to support this evolution. Autonomys addresses challenges related to scalability, security, data provenance, and user control. With a focus on decentralization, human-centricity, and open collaboration, Autonomys can lead the charge into an era of AI that is more equitable, beneficial, and aligned with human values.
Background
Autonomys Network is an implementation of the Subspace Protocol resulting from three years of R&D by Co-Founders Jeremiah Wagstaff and Nazar Mokrynsky. The Subspace Protocol created the ability to have permanent decentralized storage for Web3. From this creation, Jeremiah and Nazar cracked the integration of decoupled modular compute, resulting in the Autonomys Network. The rebranding of Subspace Protocol to Autonomys Network was announced on June 14th, 2024.
The project raised $32.9 million, led by Pantera Capital, with notable participation from Coinbase Ventures, Crypto.com, KR1, GSR Ventures, Alumni Ventures, Hypersphere Ventures, and Stratos Technologies.
Autonomys Network has had seven testnets, with over 100,000 Farmers (participants) pledging over 180 PiB in storage. Autonomys’ Mainnet Phase 1 was launched on November 6, 2024, through a Proof-of-Time seed ceremony where the Bitcoin block #869146 hash was chosen as the universal starting point of the network. In the two weeks following Mainnet Phase 1’s launch, over 2,000 nodes dedicating 140+ PiB of storage have joined the network. The native token for Autonomys Network is AI3, which is used for staking, governance, block rewards, and transaction fees.
Technology
Autonomys Network Stack
Autonomys Network follows a modular blockchain architecture that is divided into four main parts: (i) the decentralized application (dApp) layer, (ii) decoupled execution domains, (iii) the consensus layer, and (iv) the storage layer.
The modular design of Autonomys Network includes features such as Decoupled Execution (DecEx), domains, interoperability, data sharding, and an open-source AI directory. DecEx separates transaction execution from consensus, allowing independent scaling of throughput and storage while maintaining decentralization. Domains allow developers to create application-specific blockchains with customized functionality, supporting a diverse range of AI applications without overburdening the core protocol. The network’s interoperability with different state transition frameworks and execution environments like EVM and WASM facilitates integration with existing blockchains. Data sharding divides data into multiple shards distributed across different nodes, enabling parallel processing and reducing the load on individual nodes, which enhances the network’s ability to handle large datasets for AI training and deployment. Hosting a dedicated open-source AI directory fosters collaboration and innovation within the AI3.0 ecosystem, promoting knowledge sharing and protecting valuable AI resources against censorship.
Autonomys Network plans to achieve high throughput via its scalability framework. This framework includes data sharding to parallelize transactions, a beacon chain that utilizes their Proof-of-Archival Storage (PoAS) consensus mechanism, and the use of domains. Currently, storage is constrained on blockchains and is not available at scalable levels. Autonomys’ scalability framework aims to support not only storage, but also bandwidth and compute on par with current web2 throughput levels.
Consensus
Autonomys Network’s consensus mechanism is built upon the PoAS protocol called Dilithium. Dilithium is designed to be compatible with SSDs due to its reliance on frequent random reads of small data chunks, making it more efficient on this type of storage. The core functions of Dilithium, including Archiving, Plotting, Farming, Proving, and Verification, are influenced by the connectivity of both Dilithium and a separate Proof-of-Time (PoT) blockchain that runs alongside the PoAS blockchain. The PoT blockchain aims to emulate the beneficial qualities of Proof-of-Work (PoW) without the associated high energy consumption. The PoT blockchain contributes to security by enforcing a verifiable time constraint between block proposals, making it difficult for malicious actors to create long retroactive forks. It also enhances the randomness of block challenges by using its outputs as a source of unpredictable randomness, similar to how PoW mining works. To maintain the PoT blockchain, the network utilizes “Timekeepers,” dedicated nodes responsible for computing a delay function that produces the unpredictable output used for generating block challenge randomness.
The PoAS process can be broadly understood in three phases:
- A recurring deterministic Archiving phase is completed by all nodes. Archiving involves dividing the blockchain history into segments composed of equally sized pieces.
- A setup phase of Plotting involves individual farmers processing and encoding their designated portions of the blockchain history for storage.
- A probabilistic audit phase of Farmers’ storage to ensure data integrity and the opportunity for farmers to propose new blocks.
PoAS incorporates many cryptographic primitives to run successfully. These primitives include hashing, digital signatures, erasure code, KZG polynomial commitment, Merkle trees, and encoding mapping.
Decoupled Execution
Decoupled Execution (DecEx) separates the consensus mechanism from transaction execution. DecEx reduces the hardware requirements needed to participate in consensus, making the only real factor storage space widely available on all modern electronics. Instead of all nodes handling both consensus and computation, the network introduces two specialized roles for those with hardware constraints: (i) Farmers and (ii) Operators. Farmers participate in the PoAS consensus by providing storage space to maintain blockchain history, focusing on data availability and transaction ordering without executing complex computations or maintaining the full blockchain state. Operators, selected through a stake-based election process, are responsible for executing transactions and managing state transitions within specialized environments called domains.
Domains are modular, isolated execution environments tailored for specific applications or use cases, such as smart contracts or decentralized AI training. They leverage the security and data availability of the underlying consensus layer while enabling flexibility, scalability, and interoperability. By distributing execution across multiple domains, the network achieves horizontal scalability, as each domain handles a subset of transactions, reducing bottlenecks and improving throughput. Autonomys has already provided real use case products that utilize domains. Auto ID is a domain dedicated to managing identities for humans and AI Agents. Nova is a permissionless EVM domain designed for deploying and running smart contracts.
Farmers
In Autonomys Network, the main role of a Farmer is to maintain consensus. A Farmer plots pieces of Archival History to their disk and farms the created plot for block and vote rewards. They also join the Distributed Storage Network (DSN) as a node for data retrieval. The DSN is the network of Farmers who have plotted pieces of Archival History and serve them to Clients. The DSN handles data storage, retrieval, and replication across the network.
Farmers Dilemma
Utilizing Subspace, Autonomys Network can solve the “Farmer’s Dilemma.” This issue in Proof-of-Capacity (PoC) blockchain systems arises when Farmers must choose between (i) allocating their limited storage resources to either maintain the blockchain’s state and history or (ii) maximizing the storage they pledge toward consensus participation. Rational Farmers typically opt for the latter to increase their chances of earning rewards, which can lead them to operate as light clients and potentially centralize the network. Autonomys Network resolves this by allowing Farmers to maintain only a minimal state and history, thus retaining the benefits of full nodes without the heavy storage burden. Consensus is achieved through proofs of replicated storage of the blockchain’s history, collectively stored by Farmers up to the limits of their disk space. By decoupling consensus from computation, Farmers focus on proposing transaction orderings while Operator nodes handle state maintenance and compute transitions. This design enables participation from a wide range of hardware specifications, incentivizes Farmers through block rewards and fees, and promotes decentralization.
Distributed Storage Network (DSN)
Autonomys Network utilizes a multi-layered Distributed Storage Network (DSN) to ensure continuous availability and accessibility of all blockchain data without requiring any individual Farmer to store the blockchain’s entire history. The DSN is designed for efficient verifiability and dynamic availability, using techniques like consistent hashing, erasure coding, and Kademlia Distributed Hash Tables to guarantee data integrity and adapt to Farmers joining or leaving the network. Each piece of data is roughly replicated the same number of times across the DSN. The DSN is split into different layers that work in tandem. The Pieces Cache Layer (Layer-2) is for almost instant data retrieval access using a distributed hash table, the Archival Storage Layer (Layer-1) is the foundational “cold storage” with long-term data durability and redundancy and a cornerstone of Dilithium consensus, and the Content Delivery Network (Layer-3) that enhances retrieval speeds to Web2 levels. This architecture effectively manages storage bloat by distributing the growing blockchain data across Farmers.
To motivate participation in the DSN, Autonomys Network created a unique algorithm that dynamically changes onchain storage pricing according to supply-demand variations. This creates three main roles for network participants:
- Farmers: Responsible for maintaining the consensus layer. By joining the DSN, Farmers agree to retrieve data that is used for syncing nodes and return data to various clients.
- Domain Operators: Maintain the liveness of the Execution Chain and have the ability to earn rewards for their contributions.
- Timekeepers: Contribute to network security by maintaining the Proof-of-Time (PoT) blockchain, preventing long-range attacks, and ensuring the randomness of block proposals.
Data Flow
Data and data storage are central to what makes Autonomys Network unique. There are a few key steps to understanding how data flows, from a transaction being submitted to how it is permanently archived.
- A transaction is validated and then executed, activating a state change.
- Once the block that holds that transaction reaches a certain depth, currently set at 100 blocks, it follows the Archiving process.
- These newly archived pieces are added to the Farmer’s caches through the DSN. This means that these pieces are encoded into farmer plots on their disks for permanent storage in accordance with the Plotting protocol.
Based on this flow, a client can request data at any time, and the original data can be reconstructed from these archived pieces.
Staking
Staking in Autonomys Network involves Operators and Farmers, each playing distinct roles as explained above. Operators earn execution fees proportional to their stake. Farmers earn rewards based on their pledged storage and can also nominate operators by backing them with token holdings, like any other tokenholder. This increases the Operator’s stake and likelihood of being selected as a slot leader. In return, Farmers receive a portion of the fees earned by the Operators they nominate. The reward system employs a dynamic issuance model that adjusts rewards based on block height and demand for blockspace, incentivizing early adoption while ensuring long-term sustainability. Fees within the network cover operational costs and encourage efficient resource use, with transaction fees comprising storage and compute components to compensate participants appropriately.
Each transaction incurs a fee comprising two components: Storage and Compute fees. The Storage Fee is calculated based on the transaction’s length in bytes and the network’s current storage capacity, covering the cost of storing the transaction in a block and archiving it in the DSN. The Compute Fee is determined by the transaction’s weight, reflecting the computational resources required to execute it, and compensates Operators for their computational work in processing transactions. Additionally, Operators collect Domain Block Fees for executing transaction bundles within their assigned domains. These fees are distributed among operators who successfully submit an Execution Receipt (ER) for a bundle, proving the validity of their state transitions.
Node Functions
Autonomys Network features three node types, each with specialized roles:
- Full Nodes: The default configuration – (i) forming the network’s backbone, (ii) processing blocks and serving peers, and (iii) ensuring data integrity and network health.
- Archival Nodes: Extends the functionality of full nodes by retaining the entire blockchain history, proving valuable for block exploration and historical data analysis. The Subspace Foundation maintains these as a public resource.
- Light Clients: Designed for resource-constrained devices, interact with the network without storing the full blockchain state, relying on full nodes for data retrieval.
Ecosystem
Space Acres
Space Acres is an application that anyone can run on their computer and earn AI3 tokens. It allows a computer to run a Farmer node in the background, contributing unused disk space to the network.
AI Agents
Auto ID, deployed on a Domain, creates decentralized digital identities for both humans and AI agents. It serves as a self-sovereign identity (SSI) framework that enables individuals to verify their identity without resorting to biometric authentication. Key features include: (i) self-sovereignty, allowing users autonomy over information-sharing decisions through encryption, zero-knowledge proofs, and verifiable credentials; (ii) verifiability, with cryptographic proofs authenticating claims without exposing personal information; (iii) universality, as Auto IDs can be issued to any entity – human or artificial – establishing a common identity standard across the digital ecosystem; (iv) versatility, supporting identity self-issuance, issuance by another entity, and co-issuance by multiple entities; and (v) interoperability, designed for seamless integration with existing identity systems and Decentralized Identifiers.
Auto ID also incorporates a Proof-of-Personhood (PoP) mechanism called Auto Score, which assesses the likelihood that a user is human without revealing personal details. Auto Score aggregates various pieces of evidence – such as verifying official documents, linking social media accounts, or participating in decentralized networks – to calculate a probabilistic score to indicate a human identity. This verification is done mainly through the use of Zero-Knowlege Proofs (ZKPs) to ensure privacy is maintained.
Auto ID and Auto Score play a crucial role in establishing content provenance and data sovereignty by enabling entities to digitally sign the content they produce, creating a verifiable and tamper-proof record of authenticity linked to their Auto ID. This will be important as the distinction between human-created and machine-generated content becomes increasingly blurred. Through Auto ID, users are able to securely delegate authority to AI agents and define their roles and permissions for various tasks decided by the user.
By offering a standardized framework for digital identity and data provenance through Auto ID and Auto Score, Autonomys Network demonstrates the importance of operating in a way that facilitates verifiable human-AI interaction, enablement of privacy-first verification, and traceability. This framework serves as a litmus test for potential users to understand why this approach is essential in an increasingly AI-driven world.
Autonomy’s Github for AI Agents provides the needed tools to build an agent that can store its memories and context onchain enabled by the DSN. This means that agents can now store and retrieve data directly from the blockchain, enabling them to maintain a persistent memory and context over time. This capability allows agents to execute more complex tasks, adapt to dynamic environments, and provide personalized, context-aware interactions to the end users, as highlighted through Autonomy’s Auto Chain Agent demo.
Astral
Astral simplifies participation in Autonomys Network’s PoAS system by (i) offering a user-friendly interface for managing staking activities, and (ii) functioning as a block explorer. As Autonomys grows, Astral’s role in facilitating operator participation and supporting the network’s decentralized governance is expected to grow.
Tokenomics
AI3 is the token associated with Autonomys Network, with a maximum token supply of 1 billion. Approximately 65.00% of the supply was minted upon the launch of Mainnet Phase 1 in November 2024. The token will not be transferable until the official token generation event (TGE), which will occur with the launch of Mainnet Phase 2, slated for Q1 2025.
The Investors, Team, Autonomys Labs (DevCo Treasury), Subspace Foundation (Long-term Treasury), and Partners received 494.50 million AI3 (~76.08% of the initial token supply, or 49.45% of the maximum token supply), subject to a 12-month cliff and 36-month vesting schedule with monthly releases. 25.00% of the tokens will be unlocked at the end of the 12-month cliff, with the remaining 75.00% released linearly over the subsequent 36 months at a rate of 1/36th per month.
155.50 million AI3 (~23.92% of the initial token supply, or 15.55% of the maximum token supply) will be split between Autonomys Labs (Operating activities), Subspace Foundation (Operations and Near-Term Treasury), Ambassadors, and Testnets/Stake Wars participants. These tokens will not be subject to a vesting schedule; however, some participants, like the Ambassadors, might be subject to specialized vesting schedules based on ad-hoc requirements.
The remaining 35.00% of the maximum token supply will be minted as block rewards for Farmers and Operators over an estimated 40-year period.
Roadmap
With the launch of Mainnet Phase 1, Autonomys Network plans to launch Phase 2 in Q1 2025. Mainnet Phase 2 will deploy the Domain Layer, Nova EVM – currently available on Tarus testnet – and enable token transferability and vesting. Mainnet Phase 3 is slated to launch in 2026 to implement the scalability roadmap, which includes data sharding.
Closing Summary
The modular approach taken by the Autonomys Network allows them to scale effectively and efficiently to meet the demands of a growing user base and an increasingly complex AI landscape, while maintaining a high level of security and decentralization through specialized security measures at each layer. Features like DecEx, domains, PoAS, DSN, and staking provide the foundation for the open, collaborative, and human-centric AI future Autonomys aims to create. Autonomys Network’s commitment to democratizing access to transformative technologies aligns with the visionary ethos of pioneers like Satoshi Nakamoto, propelling us into an era where AI serves as a public good – accessible, equitable, and aligned with human values.
