How far has the encrypted AI intelligent agent evolved? Will it become a first-class citizen on the blockchain?

What is the current state of evolution for encrypted AI intelligent agents? Can they achieve first-class status on the blockchain?

Author: Mason Nystrom, Variant Fund Partner; Translator: 0xjs@LianGuai

Robots are becoming first-class citizens in the crypto economy.

You don’t have to look far to see evidence of this trend. Searcher deploys robots like Jaredfromsubway.eth, taking advantage of human desire for convenience to front-run their DEX trades. Banana Gun and Maestro allow human users to conveniently transact with bots on Telegram, and they have long been some of the biggest gas consumers on Ethereum. Now, with new temporary social apps like Friendtech, bots are joining the competition after the app initially gains human users, potentially unintentionally fueling the speculation flywheel.

All of this indicates that bots, whether profit-driven (such as MEV bots) or consumer-driven (such as Telegram bot kits), are becoming increasingly prominent users on the blockchain.

• Overview of POW-based new tokens A victory for miners, communities, and mining equipment manufacturers?
• IBM launches new encrypted asset cold storage technology OSO.
• Elon Musk Boosts Dogecoin to 8-Month Highs as X.AI Looks to Raise $1 Billion – Can DOGE Reach $1? A Dogecoin Price Prediction!

While crypto bots have been relatively rudimentary thus far, the rise of large language models (LLMs) has seen bots outside of crypto begin to evolve into powerful AI agents with the ultimate aim of autonomously handling complex tasks and making wiser decisions.

Building these AI agents in the crypto field has led to several important enhancements:

• Native Payments: AI agents can exist outside of cryptocurrencies, but if we want them to perform complex operations, they will need access to capital. Compared to allowing AI agents access to bank accounts or payment processors (such as Stripe), or dealing with most other inefficient problems that exist in the off-chain world, the crypto field provides meaningful improvements for AI agents.

• AI Agent Wallet Ownership: AI agents connected to wallets will have the ability to own assets (such as NFTs, earnings, etc.), granting the agents inherent digital property rights over crypto assets. This is especially important for transactions between agents.

• Verifiable, Deterministic Behavior: AI agents will be most efficient when their actions are verifiable (they can ensure that certain actions have been completed). On-chain transactions are inherently deterministic – they either happen or they don’t – meaning AI agents will be able to perform on-chain tasks more accurately than off-chain tasks.

Of course, on-chain AI agents also have limitations.

One limitation is that AI agents requiring human-like logic to be performed off-chain in order to improve performance. This means on-chain AI agents will host their logic/computation off-chain for efficiency optimization, while decisions made by the agents will be executed on-chain, allowing for verifiable operations. Importantly, AAI agents can also use zkML providers like Modulus to ensure their off-chain data inputs are validated.

Another major limitation of AI smart agents is that their usefulness depends on the tools provided to them. For example, if you ask a smart agent to provide a summary of real-time news events, the agent’s toolkit needs to include a web crawler to scour the internet and perform the given task. Do you want the smart agent to save the response as a PDF file? Add a file system to the toolkit. Want the smart agent to follow your favorite crypto Twitter influencer? The agent needs access to a wallet and the key signing authority of that wallet.

Looking at the current shift from determinism to non-determinism, most crypto AI smart agents perform deterministic tasks. This means that humans program the parameters of the task and how the task is accomplished (e.g., token exchange).

Crypto AI smart agents have evolved from early keeper bots (still used in DeFi and oracle applications) to more complex entities utilizing LLM, including autonomous artists like Botto; AI agents that can leverage Syndicate’s trading cloud to provide banking services for themselves; and early AI agent service marketplaces like Autonolas.

There are various exciting cutting-edge applications already:

• “Smart wallet” supporting AI agents: Dawn utilizes DawnAI to offer AI agents that can assist users in sending transactions, executing trades, and providing other real-time on-chain insights (e.g., NFT trends).

• Crypto game agents: LianGuairallel Alpha’s latest game Colony aims to create AI characters that can own wallets and trade with each other.

• Enhanced toolkit for AI agents: The effectiveness of AI agents depends on their toolkit, and interacting with blockchain is a nascent area. Crypto AI agents require wallets, ways to fund them, permissioning capabilities, integration with AI models, and the ability to interact with other agents. Specifically, Gnosis showcases AI mechanisms for this early infrastructure by wrapping AI scripts in smart contracts, allowing anyone (including another bot) to call the contract for proxy operations (e.g., in prediction markets) while also being able to pay the agent fees.

• AI-enhanced traders: DeFi super apps that offer advanced operations for traders and speculators, including dollar-cost averaging if certain conditions are met, executing trades when gas prices fall below a specific level, monitoring new meme token contracts, and determining order routing without users needing to know where the trade is happening in which dapp, etc.

• AI agent long-tail markets: While large-scale applications like ChatGPT serves general chat purposes, AI agents need to be fine-tuned for numerous industries, topics, and niche markets. Marketplaces like Bittensor incentivize “miners” to train models for specific tasks (e.g., image generation, pre-training, predictive modeling) around target industries (e.g., cryptocurrency, biotechnology, academia). Although Bittensor is still in its early stages, developers have been using it to build applications/agents on top of the open-source LLM.

• NPC consumer app agents: Non-playable characters are common in games like MMORPG but not as prevalent in multiplayer consumer apps. However, the financialized nature of crypto consumer apps makes AI agents excellent participants in introducing new gaming mechanics. OpenAI infrastructure company Ritual recently released Frenrug, an LLM-based agent that runs within Friend.tech and performs trades (buying or selling keys) based on user messages. Friend.tech users can try to persuade the agent to buy their keys, sell other people’s keys, or attempt to have Frenrug agent use its funds in other creative ways.

As more and more applications and protocols utilize AI agents, humans will use them as channels to access the crypto economy. Although AI agents may look like toys today, they will enhance everyday consumer experiences, become key stakeholders in protocols, and create entire economies between them.

AI agents are still in the early stages, but these first-class citizens of blockchain economies are already starting to show their potential.

We will continue to update Blocking; if you have any questions or suggestions, please contact us!