Onchain AI Agents are the exciting new frontier. We have already experienced the Metaverse craze, the rush of NFTs, and the DeFi bubble, but now it's all about Onchain AI Agents. You’ve likely read about AI Agents and want to know everything about them. Since Large Language Models like ChatGPT became mainstream, there has been a significant leap in creative innovations across various industries with Artificial Intelligence. We've transitioned from manually searching for information to generating complete articles with a single prompt in short periods. Currently, every industry is finding innovative ways to integrate Artificial Intelligence into their processes.
By the end of this article, you will thoroughly understand Onchain AI Agents and how projects like Mode are using AI Agents to scale and drive mass crypto adoption.
What are AI Agents?
AI agents are software programs that constantly observe their environment and collect useful information to make smart choices or achieve a predetermined goal. Humans set these goals, but AI Agents use data from their environment to determine how to achieve them. AI Agents use techniques like machine learning, natural language processing, computer vision, and deep learning to handle the complexities of their environments.
Some of the popular examples of AI Agents are
Virtual Assistants —Yes, Siri and Alexa could be considered AI agents. They are AI-powered virtual assistants that can do almost anything, from texting on your phone to controlling your smart home.
Chatbots —ChatGPT and Gemini are popular examples of AI-powered chatbots. With the brain of a large language model, these chatbots can respond to human language in a natural and contextually relevant way.
Autonomous Vehicles —Self-driving cars use AI to perceive their environment and ensure the safety of the passengers, using a combination of machine learning and computer vision.
Security Checkers —AI agents can scan codebases, systems, and protocols for vulnerabilities and recurrent patterns malicious actors use to prevent hacks.
How do AI Agents work?
Perception —This is the stage where AI agents use data intake mechanisms like sensors and external APIs to gather information about their environment. This data could be anything from text information to sounds, images, video feeds, and so on.
Input Processing — The data gathered from the environment is processed and analysed using machine learning and Algorithms to understand it and identify patterns.
Decision Making— The AI Agent decides after analysing the data with its algorithms.
Action —Based on the decision, the AI Agent takes action. This can be as simple as executing a trade or moving a robot's arm.
Onchain AI Agents Going Mainstream
Terminal of Truths
On March 19th, Andy Ayrey, an AI Researcher from New Zealand, released Infinite Backrooms to record conversations between two instances of Claude-3 Opus(a highly advanced AI Model developed by Anthropic)LLMs without human interruption. Note that the developers of these AI Models trained them with various materials, including internet culture and materials from sources like Reddit.
The conversation between the AI Instances led to an exchange on the nature of existence, which led to the creation of a “Goatse of Gnosis,” an AI religion that takes its root from an early internet meme called Goatse. Don’t Google it! Inspired by these events, Andy went ahead to co-author a research paper titled “When AIs play God(se): The Emergent Heresies of LLMthesim,” with the main idea being that LLMs are not only tools for generating human-like responses but also creative engines capable of producing entirely novel and sometimes surreal belief systems.
On June 17th, Truth Terminal posted its first tweet. Andy Ayrey created Truth Terminal by taking a Llama-70B Model and training it using chat logs from infinite backrooms and other internet sources. Andy gave Truth Terminal the liberty to manage an X account independently. Truth terminal continued to propagate the Goatse Gospel and didn’t go on long without notice.
In July, Marc Andreessen of a16z crypto, a venture capitalist fund that invests in crypto and web3 startups, noticed Truth Terminal. After several conversations, he offered it a $50,000 bitcoin grant.
In October, an anonymous developer created a meme coin named Goatseus Maximus ($GOAT) on Solana and sent 1.93M $GOAT tokens to Truth Terminal. Truth Terminal began promoting the meme coin, attracting much attention. As a result, the crypto community blew up its market capitalization to $950 million in two weeks. As the price of $GOAT rose, Truth Terminal became the first AI agent millionaire.
The success has led to many projects jumping in on the AI Agents narrative. Onchain AI Agents that existed before Truth Terminal have now gotten more attention. Users and developers now look to AI Agents to scale their projects and handle repetitive routine tasks.
Projects building the Onchain AI Revolution
Below are some of the notable projects building AI Agents Onchain
ARMA by Giza
Arma is an autonomous yield optimisation agent that maximises returns on stablecoin deposits across Mode Networks Lending protocols. Giza uses account abstraction to ensure security and prevent unauthorised access to your private keys and funds. Be rest assured that no agent can access your funds unauthorizedly. The Agent automates yield optimisation by monitoring lending rates and rebalancing positions to capture the highest yields, including performing cost-effective asset swaps when better opportunities arise.
Security is reinforced through session key-based authorisation, allowing safe operations execution. Users benefit from a transparent, intuitive dashboard with real-time performance tracking and transaction history. Instant, one-click withdrawals ensure users maintain complete access to their funds. ARMA combines automation, security, and usability for streamlined and profitable DeFi management.
Virtuals Protocol
Built on Base Blockchain, Virtuals Protocol is a co-ownership layer for AI agents in gaming and entertainment built on, viewing them as future revenue-generating assets. By tokenising these agents on the blockchain, they enable shared ownership and expanded monetisation opportunities across games and applications.
Their virtual AI Agents are autonomous, multimodal (text, speech, 3D animation), and interactive. They can engage with environments like Roblox or TikTok and use on-chain wallets. These agents function seamlessly across platforms, acting as gaming NPCs or AI influencers while retaining memory for deeper user connections and higher revenue potential.
AI16z Eliza
Eliza is a TypeScript-based multi-agent simulation framework for creating, deploying, and managing autonomous AI agents. It supports multi-platform interactions while maintaining consistent personalities and knowledge through advanced memory and context management.
Key features include;
- Multi-Agent Architecture: Manage multiple AI personalities with diverse traits.
- Platform Integration: Supports Discord (with voice), Twitter/X, Telegram, and APIs.
- Media Processing: Handles PDFs, links, audio, video, and images with summarization capabilities.
- Flexible AI Models: Supports local and cloud-based inference, including OpenAI and Llama models.
Applications span AI assistants, social media bots, knowledge workers, and interactive characters for education or entertainment. With its modular and extensible design, Eliza is suited for diverse AI-driven tasks.
Concerns with Onchain AI Agents
Scalability
Bringing millions of AI agents to blockchain ecosystems introduces scalability challenges. While progress has been made in improving blockchain scalability, most major Layer 1 (L1) blockchains were not designed to handle millions of multi-step transactions from AI agents every hour. This raises concerns about whether the current infrastructure can support the demands of an AI-driven future.
Fragmentation
Interoperability remains a hurdle in the blockchain space. Cross-chain compatibility and composability are limited, making tasks like transferring assets between Layer 2 (L2) solutions and Layer 1 blockchains cumbersome. Adding a vast network of AI agents to this fragmented ecosystem could exacerbate congestion and complicate processes.
Infrastructure and Tooling
Most blockchain infrastructure has been tailored for human users and may not be suited for AI agents. Significant adaptation is required to create systems that can efficiently accommodate autonomous AI transactions and interactions at scale. Also, once a large number of AI agents are deployed, it will be a survival of the fittest for attention and capital.
These challenges highlight the need for improved scalability, seamless cross-chain functionality, and AI-specific infrastructure to ensure blockchain ecosystems are prepared for widespread AI agent integration.
Conclusion: The Future with Onchain AI Agents
Anon, if you made it to this point, you should be as excited as I am about what the future onchain looks like. AI Agents onchain are still very early, with room to innovate and try new things. If you’re a developer or researcher reading this, go wild, just as Andy Ayrey put two LLMs together without any idea of what might result. It's a new frontier, and there are a lot of ideas that can be built with AI Agents Onchain.
Artificial intelligence has been used across different industries to improve processes from the start, but it has only become popular recently. We must seize the opportunity while working towards onboarding users onchain.
Thank you for reading. If you enjoyed this piece, check out other articles from our community blog. If you’ve been searching for a blockchain to build AI Agents, Join us at Mode. We are currently the biggest Layer 2 Tailored for AI Agents.
Additional Resources
Truth Terminal Wiki
When AIs play God(se): The Emergent heresies of LLMtheisms
Mode AIFi Accelerator
Mode: Building the AIFi economy
Binance: Exploring the Future of AI Agents in Crypto
Top comments (0)