@aoTheComputer, built on @ArweaveEco’s on-chain storage, achieves infinitely scalable decentralized computing, allowing an unlimited number of processes to run in parallel. Decentralized AI Agents are hosted on-chain by AR and run on-chain by AO. AO is an official Arweave project launched by Arweave founder @samecwilliams, supporting the integration of large AI models into blockchain smart contracts.

A16Z @a16zcrypto’s CTO highly praised AO’s programming model as “very elegant,” calling it truly impressive work. Meanwhile, Hill, the head of research at the well-known Web3 investment firm SevenX Ventures, enthusiastically tweeted just twenty minutes after AO’s launch event: “AO might be the ultimate solution for hosting and running decentralized AI agents.”

On-Chain AI Storage and Computing

The ultimate vision of AO is to achieve seamless integration of AI and blockchain, enabling AI models to be hosted and run directly on the chain, and to make autonomous decisions. This will significantly enhance the intelligence level of decentralized applications, bringing revolutionary changes to the crypto space. Funds will not only be passively stored but can be actively operated, calculated, and optimized through AI, greatly improving liquidity and utilization efficiency.

In traditional finance, funds are stored in banks and managed centrally, which is inefficient and cumbersome. In DeFi, assets can be automatically staked and circulated, increasing the efficiency of capital utilization. On AO, all funds are self-custodied by users and managed through their own robots or AI. Each fund can have its own strategy, which will greatly enhance the intelligence level of financial assets. Due to AO’s superior performance and architecture, users can inject contract codes into their own wallets and have agents automatically execute strategies.

In the past, Arweave as a storage protocol could only be seen as a hard drive; relying solely on a hard drive cannot support larger narratives and use cases. Therefore, Arweave developed a CPU perfectly compatible with its own hard drive—AO, and integrated it into the on-chain AI narrative. It can be said that AO saved AR, and the market has shown great interest and anticipation for AO. Before February 2024, AR’s price hovered around $6-10, significantly underperforming BTC, and did not rise with the mainnet upgrade and the start of the bull market. Until the release of AO, $AR’s price rose rapidly, increasing fourfold in a month. After the announcement of the $AO token, $AR saw a short-term maximum increase of over 18%. The volatility and wealth effect brought by $AO are becoming increasingly evident.

Challenges and Solutions for On-Chain AI Models

The core goal of AO (Actor Oriented) is to achieve trustless and collaborative computing services without any actual scalability limitations, providing applications with a new paradigm that integrates blockchain technology. While there are many mainstream public chains, they all have significant shortcomings when it comes to on-chain AI:

Ethereum:

Although Ethereum dominates in smart contracts, its computational and storage capabilities are limited, making it difficult to handle complex AI models. In contrast to Ethereum, which can only use a single shared memory space, AO allows any number of parallel processes to operate within computing units and collaborate with other units through open messaging without relying on a centralized memory space. Moreover, Ethereum’s gas fees are very high, making it infeasible for the average user to run a fully on-chain AI agent.

High-Performance Blockchains (Solana, Aptos, Sui):

Despite their high throughput, these blockchains are primarily optimized for transaction processing speed and lack support for complex AI computations. They cannot store AI models on-chain, and their runtime environments are not friendly to AI.

Decentralized Storage and Compute Projects (Filecoin, io.net):

Projects like Filecoin and io.net propose decentralized storage and computing power, aiming to compete with traditional services like AWS. However, the true value of blockchain lies in consensus and trustlessness, not in directly competing on computation or storage costs. Filecoin and io.net merely utilize blockchain technology to create a market for storage or compute resources, allowing users to trade on them. They mostly use decentralized technology to allocate computational resources without innovations to achieve censorship resistance, immutability, or trustless computation. Arweave is entirely different, focusing solely on blockchain storage. Blockchain storage, similar to Bitcoin’s ledger, is decentralized, immutable, and traceable. Within the Arweave ecosystem, projects like Apus are building GPU-based AO nodes, which will also run on AO in the future. AO, as a protocol designed by Arweave’s founder Sam, aims to be more decentralized. Global developers can freely develop GPU computation, DPIN, and social applications on the AO protocol. This openness and independence from large VC funding allow AO to be highly decentralized from the start.

AO+AR Architecture

AO is a distributed, decentralized, Actor Oriented computing system based on Arweave. An Actor is the fundamental unit of a concurrent computing model in computer science. Actor Oriented refers to a methodology based on the Actor model, where each Actor can modify its allocated private state, but to modify other Actors’ states, it can only do so indirectly through message passing. This is suitable for building highly concurrent, distributed, and fault-tolerant systems.

AO comprises three subnetworks: Messenger Units (MU), Scheduler Units (SU), and Compute Units (CU). These three subsystems collaborate, handling information reception and processing, scheduling and sorting information, and performing actual computations, respectively. Through blockchain sequencing, each unit can act as a horizontally scalable subnet, simultaneously executing a large number of transactions, thereby achieving high-performance computation and theoretically providing nearly infinite computational power. Its characteristics can be summarized as two key points:

1. Separation of Computation and Storage:

   AO uses Arweave’s decentralized storage solution to ensure data permanence and immutability. Arweave handles storing AI models and related data, ensuring data security and verifiability, while AO executes complex computational tasks, including AI model inference and training. By storing all interaction logs, AO ensures the verifiability of computation results. This holographic state storage mechanism allows anyone to reproduce the computation process and verify its correctness, opening new possibilities for decentralized computation.

2. Unlimited Expansion of Computational Capability:

   AO’s design allows for any number of parallel running processes, coordinating them through a message-passing system, in contrast to the traditional shared memory parallel computing model. This design enhances system scalability and efficiency by avoiding the lock contention issues inherent in traditional models. This is particularly important for running large AI models, as it can significantly improve computational efficiency and speed.

In the AO network, the key processes include the following steps: interaction information within the AO network is verified by the Messenger Unit (MU), then passed to the Scheduler Unit (SU). The SU sorts the information and uploads it to Arweave, where consensus on the order is achieved and the data is stored. When users need to obtain a status, they send a message through the MU to the SU. The SU selects an appropriate Compute Unit (CU) and delivers the message to the CU. The CU performs the state computation using the data on Arweave and returns the output to the user via the MU.

Llama3 Model on AO by Sam

Sam, the founder of Arweave, has provided a Llama3 model on AO in his personal GitHub repository. This means that large models can run directly on AO. The AO+AR architecture is tailored for on-chain AI, and this design has proven to be highly suitable for AI computation. It achieves three critical points that make on-chain AI agents truly possible:

1. Separation of Computation and Storage:

   Storing AI models and their training data requires significant space. Arweave offers a decentralized and efficient storage solution, ensuring data permanence and immutability. Unlike other blockchains, AO can perform model training directly, which is a significant breakthrough. The AO+AR Storage Consensus Paradigm (SCP) separates the computation layer from the storage layer, ensuring the storage layer remains permanently decentralized while the computation layer retains the traditional model. Therefore, AO’s computational resources are fundamentally no different from traditional computation, but with an added permanent storage layer, making traditional computation traceable and decentralized. AO’s computational scalability has no type restrictions, allowing various computation modes such as EVM and WASM to be implemented on AO. By storing computation logs and processes on AR, AO’s computational capability is almost identical to traditional computing. This allows further expansion into GPU computing and adaptation to NVIDIA’s CUDA, demonstrating its extensive scalability.

2. Actor Architecture:

   On-chain computation often faces resource limitations, but AO significantly improves processing efficiency through its parallel computation capabilities. AO itself is designed using the Actor model, which has a close relationship with AI research. The core idea is that each component of the system can be an independent, autonomous agent, interacting by sending messages. In this model, Actors are very similar to AI Agents. AO can allocate multiple compute units to handle AI model inference tasks, overcoming computation bottlenecks.

3. Security and Trustworthiness:

   To ensure the trustworthiness of AI inference results, all interaction logs and computation processes are recorded on Arweave, allowing any participant to verify the correctness of the computation results. AO’s security mechanisms ensure the credibility of both the computation process and its outcomes.

Ecosystem of On-Chain Autonomous AI Agents

Since the launch of the testnet, the number of developers and the volume of network messages in the AO ecosystem have grown rapidly, currently exceeding 100 million messages. The AO Ventures incubator plays a key role in promoting the development of the AO ecosystem. By providing technical support and resources, it has attracted a large number of developers and innovative projects, further enriching the AO ecosystem.

Since the launch of the testnet, the number of developers and network messages within the AO ecosystem has grown rapidly, currently exceeding 100 million messages. AO Ventures incubator has played a crucial role in promoting the development of the AO ecosystem. By providing technical support and resources, it has attracted a large number of developers and innovative projects, further enriching the AO ecosystem.

Key Projects in the AO Ecosystem

@autonomous_af:

Using the AO platform, Autonomous Finance aims to implement various financial Agents, including DCA asset management Agents, autonomously balanced index fund Agents, autonomous hedge fund Agents with customized risk strategies, on-chain prediction Agents, and high-frequency trading Agents. Currently, the DCA investment Agent product has been launched, allowing users to set parameters such as the type of assets to invest in, slippage range, liquidity pool, and investment time. However, this does not involve using AI to formulate investment strategies but stays more at the level of contract automation that does not require off-chain triggers.

@apus_network:

Apus Network, inspired by Sam, utilizes a Trustless GPU solution to achieve Trustless AI within AO. By combining the Trustless GPU solution with Arweave’s decentralized storage and AO’s super-parallel computing, Apus Network has achieved verifiable and trustworthy AI inference results, paving the way for building more reliable and transparent AI applications. The aos-llama project builds AO images using the llama2.c inference engine, providing a Lua interface to load Llama models from Arweave, and offers conversion tools and a comprehensive toolkit for efficiently performing AI inference on AO.

Other Notable Projects in the AO Ecosystem

AO GAMES:

AO GAMES is a full-chain game publishing platform built on AO, providing a Web2-level gaming experience with low entry barriers, extremely fast transaction speeds, and very low costs. Importantly, AO GAMES also supports the integration of AI models, enabling the creation of AI-driven Web3 gaming projects.

@aox_xyz:

AOX is AO’s first and only cross-chain bridge project, developed by the everVision team. It uses MPC (Multi-Party Computation) technology to provide cross-chain asset services between the AO network and other blockchain networks. Currently, AOX is in the Beta phase and only supports cross-chain functionality for $AR between the Arweave and AO networks. Recently, a cross-chain incentive campaign was launched, encouraging users to transfer $AR from the Arweave network to AO through AOX. Users can earn $TAOX test tokens by completing tasks, which will be eligible for future official token airdrops. It is reported that AOX will gradually introduce cross-chain services for Ethereum, BTC, and BSC networks, allowing more public chains and asset types to seamlessly enter the AO network.

@ar_swap:

ArSwap is the third DEX launched in the AO ecosystem, following Permaswap and Bark. It is developed by the Pianity team (an Arweave ecosystem music NFT project). Currently, ArSwap has launched test tokens such as $AR, $AOCRED, $Fire, $Earth, $Water, and $Air. Users can experience trading and establishing LP liquidity pools on ArSwap by receiving faucets. Recently, the team has optimized the product to further enhance trading speed.

@aoWebWallet:

aoWebWallet is the first native AO network wallet designed to securely manage assets on AO, supporting the sending, receiving, and tracking of AO network assets. Users can log in directly by connecting their Arconnect wallet (Arweave wallet) or by importing a .JSON wallet file. Currently, aoWebWallet supports multiple AO assets, including $AO-CRED, $Bark, $TRUNK, $0rbit, $Fire, and $Earth.

Tokenomics

On May 30, 2024, AO announced the imminent completion of the $AO token issuance, with the token going live on June 13, Beijing time. Additionally, the project is associated with the $AR token, which remains a speculative asset until $AO completes its Token Generation Event (TGE). Following the announcement of the $AO token, $AR experienced a brief surge of over 18%. This is partly because $AR is currently the primary asset associated with this event, and partly due to the rule that holding $AR allows one to obtain $AO tokens. The rules of bridging assets to AO (currently, $AR is essentially the only asset that can cross-chain to the AO network) and holding $AR are effectively mitigating the sell-off pressure on $AR due to this event.

Circulation, Token Distribution, and Participation Opportunities

According to official information, $AO will be launched with 100% fairness, with no pre-mining, pre-sale, or early access. The total token supply is 21 million, with halving occurring every four years. The $AO token model adopts the same total supply of 21 million and a four-year halving mechanism as Bitcoin. These design elements not only ensure the token’s scarcity and fairness but also pay homage to Bitcoin, demonstrating a commitment to decentralization and fair distribution principles.

The important methods of acquiring tokens are:

1. Bridging assets to AO

2. Holding $AR

3. Participating in the construction of the AO ecosystem.

However, it has been explicitly stated by the official sources that users from the United States and sanctioned regions will not be able to mint $AO by bridging networks. The release method reflects the official desire to ensure a 100% fair launch of the token. Notably, cross-chaining to AO is one of the ways to mint tokens, and currently, AOX is the only cross-chain bridge in the AO ecosystem.

Token Use Cases

According to outprog, one of the proposers of AO, during a session on X Space, the functions of AO tokens and AR tokens will be distinct. The AR token will primarily focus on storage functions and consensus maintenance for Arweave, while the AO token will focus on solving communication issues between computation and applications. In essence, AO and AR respectively maintain the network’s computing and storage functions.

Disclaimer：

1. This article is reprinted from [TrustlessLabs]. All copyrights belong to the original author [TrustlessLabs]. If there are objections to this reprint, please contact the Gate Learn team, and they will handle it promptly.
2. Liability Disclaimer: The views and opinions expressed in this article are solely those of the author and do not constitute any investment advice.
3. Translations of the article into other languages are done by the Gate Learn team. Unless mentioned, copying, distributing, or plagiarizing the translated articles is prohibited.