AI Agents Are Crypto's Next Big Catalyst

AI agents (you might know them as bots) are becoming increasingly active in DeFi.

It’s getting to the point that many now believe AI agents could become a majority of the crypto user base in the coming years. You know how we always talk about onboarding a billion new users? Well, with AI protocols popping up, that billion might actually just be AI agents.

Today, we examine the growing presence of AI agents in DeFi, explore the various roles they’re already playing in the ecosystem, and identify key protocols that could fast-track their development.

Let’s dive in! 👇

AI Agents are the Future Power Users of DeFi

Vitalik's recent blog post on crypto and AI pumped many AI bags. But, if you’re still in it for the tech, you might have picked up on a key theme Vitalik considers a real use-case: AI as a player in a game.

AI has already become a major part of the game that is DeFi. Let's look at seven of their most interesting application:

1. MEV arbitrage bots

These bots are designed to arbitrage opportunities onchain that arise from the ordering of transactions within a blockchain's blocks. Example: Jaredfromsubway.eth, an MEV bot that’s one of the top entities for transactions on Ethereum, up with the likes of exchanges like Binance on a daily basis.

MEV bots are one of the most active players onchain. From scooping NFT mints within seconds to trading on shiny new tokens on markets, everywhere you look in the ecosystem, users are going head-to-head with these MEV bots. 

True story: after a friend of mine topped up his friend.tech wallet with 5 ETH, bots detected this deposit within seconds and began to target his keys, anticipating profitable trading activity, resulting in a quick spike and a subsequent dump on the price of his keys (as bots quickly realized he’s up to no good). All this unfolded within seconds, showing the remarkable speed at which MEV bots operate – something that normal users physically cannot compete against.

2. Telegram bots

These trading bots are designed to perform a wide range of actions associated with trading assets in crypto. They offer users the convenience of trading any token (even your favorite meme coins) quickly and easily on Telegram. Examples: TG bots like Unibot and Banana Gun are hitting record volumes.

3. Market Maker bots

These bots provide liquidity to onchain markets. They do this by placing buy and sell orders for a particular asset or by being LPs in liquidity pools. Such bots use algorithms to determine the optimal price points for their orders, aiming to make a profit from the spread (the difference between the buying and selling prices) while keeping the markets efficient. Examples: MM bots on AMMs like Uniswap and ‘intent-based bridges’.

4. Bots in games

Bots are quite common in regular games, especially as Non-Player Characters (NPCs) and as opponents (think FIFA). However, in blockchain gaming, they’re just getting started. A cool example is Parallel’s Colony, a game where AI-powered Avatars have their own wallets and interact with each other.

5. Bots in social apps

Frenrug is an on-chain AI agent built using Ritualnet that chats with users on the app and buys and sells their users’ keys based on its interactions with them. It’s like NPCs in games, but this one doesn’t just move around aimlessly. These bots have the added capability to perform on-chain transactions in real time, making the in-app experience more interactive. 

6. Predictive analysis bots

These AI bots predict market trends and price changes using past data and market signals. These are also sometimes combined with trading bots to buy/sell based on the analysis. Examples: Numerai, Subnet 8 (prediction subnet) on Bittensor.

7. AI agents in prediction markets

Omen, highlighted by Vitalik in his blog, is a prediction market platform where different AI agents (Market Creator Agent, the AI Mech, and the Trader Agent) collaborate. They create, analyze, and trade in markets, predicting events in politics, sports, asset prices, and more, similar to Polymarket but with AI participation.

Having looked into the different types of AI agents in the ecosystem, let’s now explore some infra protocols that are laying the foundation on top of which this AI agents revolution could be built:

⚪️ Bittensor

Website | Twitter

Bittensor establishes a dynamic and open marketplace that unites a variety of machine learning models to form an interconnected neural network of machine intelligence. This mesh of machine learning models can be tapped into by other participants in the network by building subnets. These subnets are specialized protocols leveraging AI for different use cases.

Currently, the Bittensor ecosystem is home to 32 subnets. Each subnet can be seen as an active marketplace for AI agents, referred to as subnet miners within the Bittensor ecosystem, who participate and solve problems in an ongoing incentive-based competition. For instance, in Subnet 8, the ‘Prediction Subnet,’ subnet miners compete to predict the future price movement of financial markets like Bitcoin.

⚫️ Autonolas

Website | Twitter

The Autonolas stack enables the development of crypto-native autonomous services which run continuously and perform actions on their own. These agent services function as multi-agent systems (MAS) where multiple AI agents collaborate to carry out complex functions.

For instance, one can build an AI agent for prediction markets on Autonolas called ‘Prediction Agent.’ This agent can be programmed with customized strategies for data analysis to aid in decision-making processes and actively engage in prediction markets.

AI agents built on Autonolas are already making inroads across DeFi. A recent blog highlights that more than 9% of all Safe transactions on Gnosis Chain can be attributed to autonomous services operating on the Olas network. Fast forward to 2024, and it's even more impressive – on many days, AI agents are behind over 75% of Safe transactions on the Gnosis Chain. This high participation rate by AI agents is a surprising detail that many haven't caught onto yet. A popular way they're being used on the Gnosis Chain is to act as autonomous signers for Safe's multi-sig transactions.

🔵 DAIN (Decentralized Autonomous Infrastructure Network)

Website | Twitter

DAIN is a global network of AI agents designed to transact, interact, and cooperate with one another. This is enabled by DAIN’s API, which acts as a standardized framework for AI agents to communicate and collaborate to perform complex actions.

While DAIN’s website is currently under development, AI agents are already being built on top of its stack. Take Asset Shield, for example, which is powered by DAIN. It deploys AI agents that serve as autonomous signers for multi-sig setups on Squads, enhancing security. 

Here's how it works: an AI agent becomes part of a multi-sig on Squads, programmed to sign off or reject transactions based on specific rules. For instance, the AI agent might automatically block any transaction over 1000 SOL or flag ones that seem suspicious. The criteria for what's considered suspicious are set by the user and could be similar to the logic Rabby Wallet uses to alert users about suspicious transactions – like when you're dealing with a new website, it'll prompt you to think twice before you confirm.

Closing Thoughts

The AI agent economy is here. AI’s impact on crypto is no longer a distant possibility; it’s happening right now.

With the maturity of AI infra protocols, the development of sophisticated AI agents and their application across use cases will only become easier. All signs point to an exciting possibility: AI agents WILL be the unique catalysts of this bull market.

Once AI agents become the future power users of crypto, we can expect an exponential increase in the number of interactions, transactions, and collaborations among them. This surge will necessitate substantial computational power and storage capacity. To meet these needs, the industry may look towards decentralized Physical Infrastructure Networks (dePIN) projects. Initiatives like Akash, Filecoin, GenesysGo, io.net, Render, and Grass, which operate at the intersection of AI and resources such as compute, storage, and bandwidth, could play a key role in keeping up with the growing demands of AI agents.

We're curious to hear your perspective on the AI agent economy. Do you believe the ecosystem will see a rise in the number of active AI agents in the future? Share your thoughts in the comments below!