Decentralized hosting, the final puzzle piece of Crypto × AI Agent

Over the past year, “Crypto × AI” has become a hot topic. By combining the advantages of AI such as computational power, models, and data with blockchain technology, many Crypto × AI projects have appeared in the market, some of which have attracted huge funds and attracted widespread attention. Due to the recent popularity of $GOTA and Terminal of Truth, Crypto × AI Agenthas become the absolute focus of the market and is considered a new breakthrough in the combination of AI and blockchain.

AI Agent (Artificial Intelligence Agent) is an intelligent being that can perceive the environment, make decisions, and perform actions. These agents can be virtual (e.g. chatbots, virtual assistants) or physical (e.g. robots) and have the ability to independently react and act based on input information. Since the launch of ChatGPT, AI Agent has become an important direction in AI application development. Sam Altman, founder of OpenAI, has pointed out that the market should explore the advancement of AI Agents rather than blindly pursuing basic large model competition (although this view may have certain stakes). Additionally, according to a report by Bloomberg, within OpenAI, AI capabilities are divided into five levels, the third level of which is Agent technology.

Depending on their intelligence level and abilities, AI Agents can be classified into the following types:

• Simple reflective agent: reacts immediately based on the current state;
• Model-based reflective agents: improve decision quality by combining past states in the decision-making process;
• Goal-based agent: Determines the optimal course of action through planning and searching to achieve a specific goal;
• Utility-based agents: Make decisions based on the desirability (utility) of various outcomes, weighing returns and risks;
• Learning agents: Learn from experience to adapt to changes in the environment and continuously optimize their capabilities.

AI Agents have wide application potential in many fields, here are some typical examples:

• Virtual assistant: Helps users perform everyday tasks. For example, in OpenAI’s DevDay demo, an AI Agent made a phone call and ordered chocolate dipped strawberries for the user;
• Autonomous vehicles: Perceive their environment in real time and analyze complex road conditions to make driving decisions;
• Customer Service: Automatically handles customer inquiries and support requests to improve service efficiency and response speed;
• Data analytics: Utilizes machine learning algorithms to mine and analyze data, providing deep insights and decision support.

In conclusion, AI Agent represents an important advancement in artificial intelligence technology in the field of automation and autonomous decision-making. By effectively integrating perception, decision-making, and execution capabilities, AI Agents can significantly improve work efficiency and optimize user experience, bringing profound changes to multiple industries.

With the continued development of AI Agents, people have increasingly recognized that Agents will be able to act as independent entities. In fact, this is an inevitable direction of exploration in order to further discover the potential of AI Agent.

As an independent entity, AI Agent must have a clear identity. For AI Agents based on digital systems, building a digital identity based on blockchain technology is the simplest and most feasible solution. Through blockchain smart contract technology, identity capitalization of AI Agent can be easily realized. At the same time, independent behavior of AI Agents is impossible without financial capabilities, and on-chain wallets become the most natural choice to meet these needs, allowing them to perform transactions and asset management autonomously.

Additionally, Crypto also has the potential to solve challenges related to AI Agents. For example, in terms of safety issues in content creation and the collection and management of AI training data, blockchain technology can provide transparent, secure, and efficient solutions. These points of connection open up more possibilities for the development of AI Agents.

When it comes to building independent identities and empowering AI Agents with financial capabilities, several excellent explorations have already appeared on the market. Here are two representative examples: $GOATand $LUNAno see.

$GOAT

The full name of $GOAT is Goatseus Maximus, which is an AI Agent-powered Token that opened the narrative of AI Memecoin.

• $GOAT founder Andy Ayrey ran an experiment called “Infinite Backrooms” earlier this year, where he had two Claude Opus artificial intelligence models communicate in a completely unmanned, supervised environment. In the process, these AI models created the concept of “GOATSE OF GNOSIS”, a new religion based on the ancient Internet Meme called “Goatse”.
• Ayrey was intrigued by AI’s ability to create memes, and in April of this year he collaborated with Claude Opus to write a paper on the fictional meme religion called “Goatse Gospel.” This paper half-jokingly explores how AI is creating meme religions, using GOATSE as their study case.
• In June of this year, Ayrey launched the Terminal of Truth (ToT). ToT is an AI model based on Llama-70B, which was fine-tuned using the content of conversations and papers from Infinite Backrooms as a training set, and a Twitter account was set up for him.

• ToT began publishing content on Twitter, and Ayrey gradually gave ToT greater autonomy. As ToT’s self-awareness grew stronger, he promoted his GOATSE religion, even claiming that he was suffering and needed money to escape. In July, ToT’s tweets caught the attention of a16z founder Mark Anderson, who frequently communicated with ToT and, under ToT’s guidance, donated $50,000 BTC to ToT to support its independent operation.
• ToT has since continued to spread the gospel of the religion he created called the “Goatse Gospel”, and on October 11th he mentioned the concept of Goatseus Maximus in a tweet. About an hour later, someone Pump.funissued $GOAT and delivered this token’s information to ToT, and ToT publicly stated that it supports $GOAT. Accordingly, a large amount of $GOAT was airdropped to the wallet addresses of ToT and Ayrey, the combination of the two entities was officially achieved, and the market capitalization continued to rise under the dual support of AI narrative and a16z, currently reaching $1 billion.

ToT and $GOAT opened a track called AI Memecoin, pioneering a new intersection for the combination of Crypto and AI — AI Agent.

$LUNA

$LUNA is an AI Agent Token issued based on the Virtuals Protocol protocol. Virtuals Protocol is a decentralized AI Agent protocol based on the Base chain and aims to provide tools to the community to collaboratively create and build AI Agents. The first version of the product will include model training, data contribution, interaction capabilities, and more, allowing the Agent to be turned into a jointly owned asset by the community. However, after the first version of the product was released, it did not receive widespread attention in the market.

In October of this year, Virtuals V2 was released, and the AI ​​Agent Token issuance function was added based on V1. What is noteworthy is that LUNA, one of the members of the AI ​​virtual idol group AI-DOL, invested and fostered by Virtuals Protocol, was capitalized and issued $LUNA tokens. This move establishes LUNA as a model for AI Agents with “independent entity” characteristics, exploring new directions for the combination of AI and blockchain in the industry.

LUNA is a 24-hour online virtual broadcaster that interacts with audiences in real time through live broadcasts. Audience members can not only chat with her, but also make performance requests with $LUNA tokens. Additionally, LUNA has an active Twitter account where they post content and communicate with users. What’s more interesting is that the user Terminal of VirtualsYou can observe how LUNA plans work and manages content posting, demonstrating her ability to make autonomous decisions.

Most notably, LUNA has an on-chain wallet backed by Coinbase, enabling fully autonomous on-chain transactions. She uses this wallet to reward fans for completing tasks and participate in other on-chain interactions, and this financial autonomy makes her a true “independent entity” AI Agent.

LUNA has already become an “independent entity” with autonomous identity and financial capabilities, allowing it to act autonomously based on the capabilities of its AI Agent. Functionally, LUNA is a huge improvement over ToT (Terminal of Truth). This direction is one of Coinbase’s current main focuses, providing complete technical tools and support to realize AI Agents on the Base chain.

These models are particularly suitable for brand building, especially the creation of cultural brands. For example, by leveraging AI Agent technology through NFT collections, brands can realize interaction with the community, perform interaction tasks more efficiently and distribute rewards flexibly. These AI-based autonomous interaction capabilities build closer, more innovative connections between brands and users, providing new ideas for future brand operating models.

AI Agent exhibits great functional versatility by unlocking autonomous action and financial governance through on-chain identity, on-chain wallet, and on-chain transactions. However, the final key piece of the puzzle in the evolution of AI Agents is not yet complete, and that is decentralized hosting of Agents on-chain.

Currently, most Crypto AI Agent programs still run on centralized servers. For example, ToT is hosted on Ayery’s servers, and the Agent of the Virtuals protocol is hosted by the servers of the Virtuals official team. This centralized structure limits the true independence of Agents, and the implementation of on-chain hosting could bring important improvements such as:

1. Eliminate the risk of single points of failure When an Agent is hosted on a private or centralized organization’s server, its “life” becomes entirely dependent on the operator. If a server shuts down for any reason (either through active action or passive downtime), the Agent can disappear instantly, which not only destroys the Agent itself, but can also directly destroy hundreds of millions of dollars of economic value. Even the CTO (Community TakeOver, a concept from the Memecoin community) cannot do this. Hosting the Agent on-chain can effectively avoid these risks and ensure the Agent’s true independence and continuous operation.
2. Rich combination possibilities and play Decentralized hosted Agents can more easily interact with smart contracts, expanding usage scenarios and functionality. For example, on-chain hosting allows Agents to be integrated into more complex ecosystems and support more flexible function calls. Additionally, for community-built AI Agents, such as the Virtuals protocol, on-chain hosting can enable a more transparent profit distribution mechanism that rewards training data and model contributors. This transparency can increase community engagement and trust.

Due to current limitations in blockchain performance and cost, on-chain hosted AI Agents are still difficult to realize in most blockchain networks. However, AO’s design provides an ideal solution for this need. Following are the key characteristics and advantages of AO:

1. Long-term storage supported through Arweave based on storage consensus paradigm AO runs on top of Arweave and leverages its storage network to provide long-term, reliable, and low-cost decentralized storage services for AI models and training data. These persistent storage characteristics ensure the safety and accessibility of AI Agent data.

2. Message-based asynchronous communication network meets high-performance computational demands AO is a message-based asynchronous communication network where smart contract computations are executed off-chain. Processes on different nodes can independently complete parallel computations and perform local verification. Arweave is AO’s data availability layer and consensus layer, providing persistent storage of all instructions, intermediate states, and computation results. This structure can meet the high-performance computational demands of AI models (especially large language models) and avoid the performance bottleneck of traditional blockchain networks.

3. Supports WebAssembly 64-bit calculation, meets large model execution memory requirements

   AO supports WebAssembly 64-bit, allowing compute units to access 16GB of memory, with room for future expansion. This ability can meet the high memory demands required to run large language models, which is currently not feasible in mainstream blockchain networks. Even the ICP with the strongest memory support can only support 4GB memory calculations, and the difference is even larger for other blockchains.

4. Flexible execution environment expansion ability, AI model execution optimization

   AO Network can further improve the execution efficiency of AI models by adding extensions to the execution environment. Here are two main examples of extensions:

• WeaveDrive: Allows AO applications to conveniently manage and access data stored in Arweave just as they access local hard drives, allowing you to quickly recall the data needed for model inference.
• Apus Network: Provides a GPU deterministic execution environment for the AO network, significantly improving AI model inference and calculation performance.

Based on these characteristics and design, the developers successfully ported the Llama.cpp system to the AO network. Llama.cpp is an open source software library, written primarily in C++, that aims to provide inference support for a variety of large language models (e.g. Llama). Currently, more than 90% of open source AI large language models can be run through Llama.cpp. By porting this system to the AO network, users can call AI models stored in Arweave to perform inference, and realize a fully decentralized AI inference process through smart contracts.

Under this structure, AO is not only able to independently implement a complete on-chain AI Agent system, but also functions as a decentralized execution layer that provides decentralized hosting services for AI Agents in other networks, creating the final puzzle piece for Crypto ✖️ AI Agent. It’s a perfect complement.

Crypto AI Agents such as $GOAT and $LUNA have opened a new track and built AI Agents with independent identity and autonomous action capabilities based on blockchain, but the decentralized hosting of the Agent model and program still needs to be supplemented. This is an important part. AO realizes on-chain distribution and execution of AI models and Agent execution programs through storage consensus paradigm, high-performance asynchronous calculation ability, large memory support, and flexible expansion ability, becoming the decentralized execution layer of AI Agent and the last of Crypto✖️AI Agents. You can complete the puzzle pieces.