From ChatGPT to Autonomy – AI Agents Reshape Crypto Trading Overnight

Artificial intelligence (AI) is transforming a number of industries, yet AI’s impact on the crypto sector will likely be huge moving ￰0￱ is primarily due to the rise of AI agents that can automate crypto trading tasks such as research, charting, strategy execution, and ￰1￱ Agents for Intelligent Crypto Automation Although still a relatively new concept, traders are now able to use OpenAI’s ChatGPT Agent to surface crypto signs, analyze trends, and quickly act on insights. A ChatGPT Agent is an AI tool that performs complex, multi-step tasks, allowing it to research and take certain actions to produce desired ￰2￱ Trading Assistant Is Live! Crypto traders, forget about manually analyzing Crypto charts & searching for patterns.

Now, there's an AI for that! We developed a mathematical #AI model designed to do Technical Analysis, detect chart patterns & more. Let's take a dive ￰3￱ — ChainGPT (@Chain_GPT) December 6, 2023 David Sneider, founder at Lit Protocol, told Cryptonews that AI agents—like ChatGPT Agent—are autonomous systems (usually powered by large language models, machine learning, or rule-based logic) that can perceive information, reason about it, and take actions on behalf of a user or an ￰4￱ explained that, unlike a static program, an AI agent can: Understand goals: For example, “maximize yield,” “reply to emails,” “rebalance portfolio.” Plan and decide : For instance, choose between multiple strategies or tools, and act in the real world through tools (APIs, smart contracts, integrations with apps).

Learn or adapt over time: This ensures improvement based on results. “In practice, AI agents often sit between users and systems, turning human intent into automated execution,” Sneider ￰5￱ Agents Versus Rule-Based Bots Sneider added that while AI agents are similar to rule-based bots, the major difference is that AI agents are dynamic in their decision-making, whereas rule-based bots operate on pre-programmed ￰6￱ example, rule-based bots act on “If X then Y” ￰7￱ Ozery, CEO and founder of Ensemble—a commerce layer for AI agents—told Cryptonews that an example of a natural-language prompt is as follows: “Watch ETH on ￰8￱ 1h momentum turns positive and gas Sneider pointed out that it’s the AI agent’s ability to perceive, reason, and adapt that makes them different from rule-based ￰9￱ is also why AI agents are currently becoming one of the biggest trends in ￰10￱ to Sneider, the increasing power and accuracy of large language models are expanding out of their initial chatbot/Google-replacement use cases and are being given tools to execute complex tasks.) ￰11￱ AI Agents Are Used for Crypto For instance, Sneider pointed out that today, AI Agents are mainly used for personal productivity, customer service, finance, and crypto ￰12￱ speaking, he explained that AI agents for crypto are used primarily for three types of emerging products.

“First, AI agents can be used as research agents that help users understand the crypto ￰13￱ agents do not trade, but simply help with education.” Secondly, Sneider mentioned that agent chatbots can now execute transactions in real time via a user’s self-custody key. “In this case, the agent is a replacement for the Web3 browser and users tell the agent something like ‘Buy $50 worth of ETH’ and the user signs the message in real time to execute the transaction,” Sneider explained. Finally, Sneider stated that some AI agents can execute trading strategies with a “human out of the loop” approach (also known as “HOOL”). This means that the user can give instructions to the agent and/or deposit funds with an agent who will perform transactions automatically.

“In all of these applications, the agents will perceive via the ingestion of market data and reason on that ￰14￱ in the third case, where agents trade, will the agent execute transactions on behalf of the users,” Sneider ￰15￱ crypto trading bots are also being used to simplify decentralized finance (DeFi). For example, Marko Stokic, head of AI at Oasis Network, told Cryptonews that “DeFi agents” can autonomously farm the best ￰16￱ like Giza and ZyFAI are already proving the potential around these use ￰17￱ example, Giza’s model leverages on-chain agents that execute trades quickly and efficiently. Giza’s flagship agent, ARMA, has already executed over 100,000 trades and optimized over $30 million in user ￰18￱ autonomous agents operate on a block-by-block basis, adjusting real-time strategies to respond to market conditions without requiring constant user ￰19￱ AI Agents Perform Private Key Management While AI agents for crypto trading can be powerful tools, challenges such as private key management must be managed ￰20￱ explained that there are three main ways to ensure this.

“First is centralized, third-party custody of key material offered by embedded wallet SaaS companies that let users delegate authority to the application they are using to sign on their behalf,” Sneider ￰21￱ this case offers security and guardrails to be maintained by the custodian or embedded wallet provider, this may also pose a major security disadvantage and even limit the use of agents. Next, Sneider pointed out that guardrails can be embedded within a smart contract account. “This setup lets users create a session signer (via methods like ERC-7579) to create a secondary sub-key that can sign transactions within the scoped permissions or allowances.” He noted that the advantage of this setup is that the permissions are all on-chain.

However, the drawback is the limit of smart contract accounts, which incur the expense and overhead in a multichain setup, where a wallet needs to be maintained on each support chain. “Furthermore, universal rules such as a daily spending limit for an agent across chains become expensive to calculate since the smart account on each chain needs to be aware of all trades made by the ￰22￱ can result in writing a lot of state to the blockchain, which is gas-intensive,” Sneider commented. Finally, Sneider noted that the third option is to leverage a decentralized method to manage the agent’s key ￰23￱ is where Lit Protocol comes into play, which is a decentralized key management ￰24￱ agent platform called “Vincent” is built on Lit Protocol, which allows AI agent users to write permissions or guardrails on-chain, and then have the network enforce these rules (for example, spending limit, allowed contracts, etc.).

According to Sneider, this allows a number of unique advantages. “First, agents can be multichain out of the box. Next, the ability to perform checks/compute off-chain, like running a transaction simulation, as part of the user’s guardrails, and third, an ability for agent developers to call smart contract functions via APIs and MCPs, opposed to having to write and call smart contracts directly,” he ￰25￱ put this in perspective, Sneider shared an example of an agent at work:),” Sneider ￰26￱ this agent, an end user would have to perform all these steps manually, while also being aware of yield ￰27￱ Challenges to Consider In addition to private key management, a number of other challenges are associated with AI crypto trading ￰28￱ instance, Ozery mentioned that AI agents can also make bad decisions or experience “hallucinations.” “This involves misreading data or overfitting backtests,” he said.

“Market structure risks should also be taken into account, along with compliance and auditability.” Stokic also said that the biggest adoption blocker today is trust. “Why should users trust someone else’s AI agent and deposit their money into it?” The Future of AI Agents and Crypto Trading Challenges aside, automated trading with AI crypto trading bots will continue to gain ￰29￱ ChatGPT Agent progress, OpenAI has stated that this is part of a broader movement toward agentic finance, where multiple autonomous assistants support real-time decision-making under human ￰30￱ further believes that as AI agent management matures, automated bots will likely not only live within applications, but will also be able to log into apps to become the de facto paradigm for how work is automated.

“For example, if you give an AI agent your Facebook password, Facebook will not know if the user requests are coming from a human or an agent,” he ￰31￱ added that AI crypto trading bots may also become the biggest users of stablecoins , at least in terms of generating yield. “In traditional finance, generating yield on government-issued assets is the role of banks who provide meager returns to users as they lend out their dollars via savings ￰32￱ the emerging world that runs on crypto rails, the agent that optimizes yield on stablecoins can dethrone traditional savings accounts by providing better returns for users on their dollar, leading to agents being the biggest users of stablecoins.”