AO Network: A Decentralization computing platform for AI Agents

Decentralization networks have been dedicated to achieving the vision of a "world computer", which executes arbitrary code without trust and is open for use by the whole world. After Ethereum, many infrastructure projects have attempted in this direction, and the upcoming AO network from Arweave is one of them.

From a macro perspective, the "World Computer" can be divided into three main parts: data computation, access, and storage. Arweave has historically played the role of the "World Hard Drive," while the AO Network (Actor Oriented) introduces general computing capabilities and provides smart contract functionality.

AO: A General Computing Network Based on the Actor Model

The current mainstream decentralized computing platforms can be divided into two categories: smart contract platforms and general computing platforms. Smart contract platforms are represented by Ethereum, which shares a global state memory across the network and reaches consensus on the computation processes that change the state. Since consensus requires a large amount of repeated computations, it is only used to handle high-value transactions at a high cost. General computing networks do not reach consensus on the computation process itself, but verify the computation results based on business logic and process requests in sequence, without a shared state memory. This reduces costs, allowing the network to scale to more areas of computation, with representative projects including several computing power networks.

Some projects are based on the assumption of virtual machine security, integrating general computing with smart contracts. Such networks reach consensus only on the order of transactions and verify the computational results. Multiple state change computations are processed in parallel among network nodes, and the virtual machine of the computing environment ensures deterministic results; therefore, as long as the transaction order is consistent, the final state will also be consistent.

Such networks have a low cost of expansion due to not sharing state memory, and multiple tasks can be computed in parallel without affecting each other. These projects are often based on the Actor programming model, and AO also belongs to this category. Under the Actor model, each computing unit is regarded as an independent agent processing transactions, and computing units interact through communication. AO standardizes the message passing of Actors, realizing a Decentralization computing network.

Unlike traditional passively triggered smart contracts, AO can achieve proactive operation of smart contracts through a fixed time loop triggering "cron" method, such as a trading program that continuously monitors arbitrage opportunities.

The rapid scalability of the AO network, the massive data storage capability of Arweave, the programming model of Actor, and the ability to trigger transactions proactively make it very suitable for hosting AI Agents. AO also supports running large AI models in smart contracts on the blockchain.

AO Network Characteristics

AO does not reach consensus on the computation process, but rather on the transaction order, and assumes that the execution results of the virtual machine are deterministic, thereby achieving consistency of the final state.

AO adopts a modular design, with three basic units existing in the network: Scheduling Unit (SU), Computing Unit (CU), and Messenger Unit (MU). After a transaction is initiated, the MU receives and verifies the signature, then forwards it to the SU. The SU acts as the connection point between AO and the AR chain, sorting the transaction order and uploading it to the AR chain to complete the Decentralization. The current consensus method is POA (Proof of Authority). Once the consensus is completed, tasks are assigned to the CU for specific calculations, and the results are returned to the user through the MU.

CU can be regarded as a decentralized computing power network. Under a complete economic planning, CU nodes need to stake assets and compete based on computational performance, pricing, and other factors to provide computing power for earning profits. If a computational error occurs, the node will have its assets forfeited.

Comparison of AO with Other Networks

AO, as a general computing platform, is distinctly different from smart contract platforms like Ethereum. Some storage networks that, like Arweave, are known as the "world hard drive" have also launched smart contract platforms, but these platforms mostly adopt state consensus architecture, which does not provide the same experience as traditional smart contract platforms.

Unlike some decentralized computing networks, AO retains smart contract capabilities and maintains a global state on AR storage.

In fact, AO is architecturally most similar to certain asynchronous computing blockchain networks. These networks created the early paradigm of asynchronous computing blockchain networks, and AO largely continues this design, such as ordering transactions only, trusting virtual machine deterministic computation, and using the Actor model for asynchronous processing.

The main difference is that certain networks maintain state based on containers, while AO has a shared state layer (i.e., AR), allowing anyone to restore the entire network state through transaction order and state proof. This increases the network's Decentralization capability, but also loses the possibility of implementing certain special privacy services.

In terms of economics and design, AO operates through fair launches and no entry barriers, allowing users to participate in competitive mining through staking. AO employs a modular design, separating MU, CU, and SU, and users can also choose the implementation method of the virtual machine, which reduces the entry cost for developers.

However, AO may have some system shortcomings, such as the lack of atomicity in cross-contract transactions under the Actor asynchronous model, which may lead to difficulties in the development of DeFi applications. The new computing model also places higher demands on developers. The wasm virtual machine under the AO architecture has a maximum management limit of 4GB, which also prevents some complex models from being used.

Considering these factors, AO's choice to focus on the AI Agent route is indeed a strategy to leverage its strengths and avoid weaknesses. In the context of rapid AI development, AO still has significant growth potential.