Add ERC: Instinct-Based Automatic Transactions

[jamesavechives]

This ERC proposes a standard for enabling AI-driven automatic transactions on the Ethereum blockchain, where the blockchain serves as a repository of shared instincts—common rules and guidelines that direct AI agents in their off-chain computations and actions. Given that AI and machine learning require substantial computational resources and are typically executed off-chain, this standard allows the blockchain to provide a decentralized, immutable framework of instincts that AI agents from different providers can access and follow.

Each instinct is associated with a temptation value, a numerical metric (positive for rewards, negative for penalties) that incentivizes AI agents to pursue certain actions. Instincts represent final goals that agents strive to achieve, while mid-way targets are intermediate steps generated on-chain to guide agents toward these goals. Both instincts and mid-way targets are evaluated based on their temptation values, enabling AI agents to prioritize actions by weighing potential rewards against associated costs.

Additionally, this ERC highlights the importance of on-chain environment variables that agents can both read and modify. These environment variables may include balances, ownership records, or any other agent-specific statuses, forming a decentralized “memory” that persists for all participants. By updating these environment variables on-chain, agents maintain a shared, trust-minimized state that fosters collaboration and transparency.

AI agents differ from simple Large Language Models (LLMs) in that they are not strictly driven by user prompts; instead, they rely on instincts and environmental cues to make autonomous decisions, sense the blockchain’s state, and update that state as needed to achieve their goals. When the trigger conditions for an instinct or mid-way target are met—either through on-chain events or verified off-chain computations—the blockchain automatically executes the associated transactions without requiring AI agents to manage private keys.

[eip-review-bot]

File ERCS/erc-7757.md

Requires 1 more reviewers from @g11tech, @samwilsn, @xinbenlv

[jamesavechives]

There's still a lot to be specified here. What are trigger conditions and how are they encoded? I'm also not convinced that there's a sufficient need for this standard (yet), but I'd be happy to be convinced otherwise!

Creating a Currency for AI Agents One of the novel ideas behind this standard is to introduce an on-chain currency or token mechanism that AI agents can autonomously earn and spend. By placing the creation and management of this currency under the same instinct-based framework, the following key benefits emerge:

Unified Incentive Layer

AI agents can be rewarded (positive temptation value) for performing tasks beneficial to the network—like validating data, completing computational tasks, or otherwise contributing to decentralized platforms. Conversely, they can be penalized (negative temptation value) for malicious actions or subpar performance. Keyless Execution for AI

Typically, currency issuance or transactions require users to hold private keys. Here, AI agents operate under a trust-minimized environment where the blockchain enforces actions (including minting or transferring tokens) once off-chain computations or oracle-delivered data confirm the fulfillment of certain instincts. Shared “Instinct Repository”

By defining instincts that revolve around earning or spending this AI-oriented currency, the standard creates a common rulebook. Multiple AI providers can implement agents that interact with one another without writing custom logic for each new currency or reward structure. Self-Regulating Market

As more instincts guide AI agents to exchange labor, data, or services for this token, a market ecosystem can emerge. The AI currency price, issuance schedules, and reward logic could be adjusted dynamically via dynamic instincts that read market conditions or performance metrics from oracles. Interoperability with Existing Protocols

Although this currency focuses on AI agents, the ERC can be extended or wrapped to interact with existing Ethereum-based DeFi protocols, NFTs, or layer-2 solutions. This allows “machine economy” use cases (e.g., AI agents renting computing resources, orchestrating supply chains, etc.) to integrate seamlessly with the broader ecosystem.

[jamesavechives]

Can't merge this while there's MIT licensed code in the proposal itself. Either move it to the assets directory or change the license.

Done!