Starting from scratch, AO assists Arweave in making a comeback
Arweave’s mainnet was launched on November 18, 2018, and has undergone 13 major upgrades in over five years. Its focus is on providing permanent decentralized storage services. However, when we observe the changes in network data, we find that these upgrades have not truly created a competitive advantage for Arweave. By examining Arweave’s business data, we can see that after entering 2023, the growth of its storage business has significantly slowed down, with a substantial decline in storage data. The monthly network storage volume has been fluctuating between 2 to 4 TiB, with the lowest volume recorded in June at only 1.43 TiB. In 2023, the total storage volume for the year was 32.96 TiB. In comparison, the Filecoin network’s total storage volume in 2023 exceeded 1.8 EiB (1 EiB = 1,048,576 TiB). It is evident that Arweave is unable to challenge Filecoin’s dominant position in decentralized storage, making business expansion extremely difficult.
Not only is Arweave facing challenges in horizontal competition within its industry, but its position in the decentralized storage field also has some distance from ordinary retail investors. Ordinary users do not have strong incentives to use Arweave and cannot quickly perceive changes in the fundamentals. Although decentralized storage has gained some momentum with the rise of AI/DePIN narratives in the new bull market, it has not received much attention from the market.
This predicament is also reflected in the token price. In the analysis period of one year, until February 2024, the price of AR has hovered around $6 to $10, significantly underperforming BTC. It did not experience an increase with the launch of the mainnet and the start of the bull market until founder Sam announced the official launch of AO on February 14.
Arweave, as a storage protocol, can only be seen as a hard drive. Relying solely on a hard drive is insufficient to support larger narratives and use cases. For a long time, Arweave relied on other protocols to use its hard drive, but the results were minimal. It was too far removed from the infrastructure of ordinary users and lacked market attention. As a result, Arweave created a CPU called AO, perfectly compatible with its own hard drive, which quickly drove its token price from $8 to nearly $50. This article will mainly introduce the technical principles and related ecosystem of AO.
1.1 AO’s technical principles: How to achieve verifiable infinite computation?
AO is an Actor Oriented (role-based) computer running on Arweave. It is designed as an environment that can host an arbitrary number of parallel processes, which are coordinated through an open message-passing layer.
The most fundamental characteristics of AO can be summarized as follows:
(1) Arbitrary number of parallel processes, providing unlimited computational capacity.
(2) Verifiability and reproducibility of computation results, achieving minimal trust requirements.
Before explaining how AO achieves the above features, let’s first understand the basic components of AO. The AO system includes two types of basic units: processes and messages, as well as three types of fundamental entities (which can also be seen as three important roles): scheduling units (SU), computing units (CU), and messenger units (MU).
Processes: These are the computational units in the network. The state of a process can be derived through computations performed by computing units, and processes can receive messages from users and other processes. In specific terms, let $P_i$ represent the $i^{th}$ process. Define $P_i = (Log_i, Init_i, Env_i)$, where $Log_i$ is the ordered sequence of all messages of $P_i$, $Init_i$ is the initialization data of $P_i$, $Sched_i$ is the scheduler of $P_i$, and $Env_i$ is the computational environment of $P_i$. At a given time step, the state of $P_i$, denoted as $S(P_i) = F(Log_i, Env_i)$, is calculated based on $Env_i$ and the message log.
Messages: In AO, each interaction between processes is represented as a message. The essence of a message is a data item that conforms to the ANS-104 standard. The uniformity of message format is crucial for the settlement and processing of the entire AO environment on the decentralized data layer of Arweave.
Scheduling units: These units are responsible for assigning atomic incrementing slot numbers (similar to Ethereum nonces) to messages sent to processes, i.e., they sort the messages. After allocation, the scheduler needs to ensure that the data is uploaded to Arweave for permanent accessibility by others.
Computing units: These units are responsible for computing the state of processes in AO. They form a decentralized computing market similar to decentralized computing protocols like Akash, where computing units compete with each other to provide services for computing process states. After completing the service, computing units need to return the computed result and a signed proof of state. If a user does not trust a single computing unit, they can further request other computing units. Computing units need to stake a certain amount, and if they provide incorrect states, their stakes will be forfeited.
Messenger units: These units are responsible for receiving incoming messages from clients, routing these messages to the designated scheduling units, and retrieving results from computing units.
AO achieves verifiable infinite computation capacity through several key technical points:
(1) Storage-based consensus paradigm (SCP): AO achieves consensus by holistically storing message logs in Arweave. Arweave serves as a continuous and immutable logbook to store all message logs, ensuring that the interaction logs are permanently accessible, thereby enabling any network participant to compute the state.
(2) Consensus is only required for data order and storage, not for state: Traditional blockchains like Bitcoin and Ethereum adopt consensus mechanisms that require participants to agree on the state of the ledger. All nodes need to validate and reach consensus on the current state, resulting in wasted computational resources and limiting the speed and scalability of the network. AO, on the other hand, does not require consensus on the state itself. Consensus is only needed for the order and storage of message logs in Arweave. The state of AO is “holographically” implied in the message logs hosted on Arweave. Although there is no consensus on the state, theoretically, anyone can compute the state based on the data stored in Arweave. If a user wants to obtain the state, they can request computing units to perform the computation and provide proofs.
(3) Separation of AR and AO functionalities: Based on the above analysis, AR and AO have their own roles. AO does not solve the validation problem but is responsible for message delivery, sorting, and state computation, mainly dealing with computational issues. Arweave handles security and verifiability, achieves consensus on data order, and ensures permanent and immutable decentralized storage. AO completes storage based on the interaction logs on Arweave but cannot modify the consensus of Arweave.
(4) Message-passing parallel computing architecture: There are two basic methods to achieve parallel computation: shared memory and message passing. AO adopts the message-passing architecture, in contrast to parallel blockchains like Solana and Sei that use shared memory. In shared memory, when one user accesses and modifies certain data, other users cannot modify it. They need to wait during the “lock competition” process until the user unlocks the data before others can access it. Therefore, this approach has a scalability limit. AO, on the other hand, only needs to send interaction messages during interactions, without the need for “lock competition” waiting, achieving horizontal scalability, which allows it to scale to any size.
(5) Modular architecture of AO: AO’s modularity is reflected in the separation of CU, SU, and MU. Users can freely choose appropriate sorters, message relayers, computing units, and even system-level virtual machines, thereby supporting the introduction of processes from various smart contract systems into AO. The horizontal scalability of CU, SU, and MU ensures that computational capacity can increase with demand.
Based on the above analysis, the key processes in the AO network include: After the interaction information in the AO network is verified and signed by MU, it is passed to SU, which sorts the messages and uploads them to Arweave. Consensus on the order is reached and stored in Arweave. When a user wants to obtain the state, they send a message to MU, which then routes the message to SU. SU selects an appropriate CU and delivers the message to it. The CU computes the state based on the data on Arweave and returns the output to the user through MU.
1.2 Does AO’s technology create a competitive advantage?
To explore this question, we compare AO with two similar projects: high-performance parallel blockchains (such as parallel EVM, Solana, etc.) and decentralized computing protocols (Akash).
Comparison with high-performance parallel blockchains:
The main difference lies in the difference in the basic architecture of parallel computing. Here, we will use Ethereum as a benchmark to better illustrate the differences of AO. Traditional EVM, represented by Ethereum, processes transactions sequentially, with only one transaction modifying the state at a time, resulting in a single linear forward mode for the entire system.
Parallel blockchains generally operate on non-conflicting or non-overlapping tasks, achieving parallel processing to improve performance. However, AO’s focus is on providing verifiable infinite computation capacity, which is distinct from the parallel processing architecture of high-performance blockchains.
Comparison with decentralized computing protocols:
Decentralized computing protocols like Akash focus on providing decentralized computing power and resources, enabling users to rent computing resources from a distributed network. In contrast, AO focuses on providing a verifiable infinite computation environment. Although both aim to decentralize computing, they have different focuses and use cases.
In conclusion, AO’s technology and architecture provide unique features and capabilities, enabling verifiable infinite computation. However, whether this technology creates a competitive advantage depends on various factors, including market demand, adoption, and the evolution of the decentralized storage and computing landscape.
