The Proof of Intelligence Model
Proof of Liquidity isn't just about incentivizing liquidity. It's an incentive flywheel that allows protocols to incentivize any on-chain action.
In DeFi, this looks like being incentivized by PoL to stake LP tokens, take out loans, and maintain open positions on perpetual exchanges, creating an infinite design space for how to incentivize users, bringing in a new frontier for the markets we design in crypto.
Now, what does that look like for AI?
In Artificial Intelligence, just as we have swaps and lending in DeFi as base actions that will always be done, we have similar fundamental actions in AI. Some of these are:
Inference - users utilizing AI agents
Training/Fine-tuning models
Managing compute resources in centralized or decentralized networks
Agentic Workflows / Function Calling
In Web2, these actions are typically incentivized and subsidized by centralized entities through venture capital and revenue/subscription models. This has proven to work, as AI is essentially meant to be centralized to create abundance.
In crypto, with AI agents, these tokens started experimentally but with good memetics/sometimes real use cases, e.g., Goat, AI16z, & Luna at Virtuals. This has led to a long tail of startups attempting to capitalize on this, but the majority have failed. Why?
AI Agents in crypto fundamentally have four traits:
Knowledge - This is crucial, as crypto is mostly a game of information, and the quality of an agent's short to medium-term knowledge is crucial to what makes it effective.
Economic Value (AUM, Market Cap, Revenue streams) - This is mostly applicable to trading/DeFAI agents. Agents in crypto garner supporters because they make their users money or provide tools for users to make money themselves.
Utility (What does the agent do?) - From Zerebro making songs to AI16z trading memecoins, an AI agent's utility is essential for establishing a meaningful edge amongst a sea of competitors.
Mindshare (Cultural Relevance) - AI Agents on Twitter and their fight for cultural relevance typically correlates to price action. AI Agents provide entertainment, alpha, and also have the job of maintaining attention in a fleeting industry with transient users.
Now that we understand what makes AI Agents in crypto unique, what happens when we introduce PoL into that mix?
Proof of Liquidity expands the design space for these traits in several different ways:
Knowledge - AI Agents are fundamentally underpinned by foundational LLM models with datasets that go stale very quickly. They're useful for long-term context windows, but this is a huge disadvantage for LLMs operating in financial markets. With Proof of Liquidity, AI Agents can incentivize their participants with BGT yield in PoL through their holders' contributions to their knowledge contexts.
Economic Value - As agents accrue better knowledge and value, their economic value increases. If you're an AI16z holder who gives Eliza a contract address that eventually boosts its AUM, you should receive a larger share of those profits than other holders for being active and directly contributing to the DAO's AUM. In PoL, you can provide commission fees to these users through boosted BGT yield.
Utility - This is the most challenging aspect, but if an AI agent's framework and codebase are public, imagine a developer adding a new plugin to enable deposits into Morpho Vaults. That developer should receive a commission from the BGT yield that the new bribe revenue generates. This can be implemented elegantly using off-chain + on-chain logic with the core PoL smart contracts.
Mindshare - AI Agents can't accumulate mindshare by themselves. They need people writing about them, interacting with them, and prompting them. Otherwise, AI agents are just tweeting into the void. Just as there are Kaito yap boards that have network effect flywheels around certain topics and pre-TGE tokens, you can create these flywheels for AI Agents.
We fundamentally believe that these flywheels, can all take place with Berachain's Proof of Liquidity economic models where we can incentivize both off chain & on chain action that results in economic value created for AI agents in crypto, which subsequently rewards these participants to continue to interact with and support these agents.
This doesn't just apply to AI Agents though, with the right designs adn considerations this extends to all parts of what crypto x AI look like from
DePin - Bootstrap Inference networks without the need for a new token
Fine tuning multi-modal models by incentivizing content creation and IP to tailor a model
Open Source contributions to building AI tools
Fully autonomous ai agent markets
Last updated