A coding agent that autonomously evolves itself daily, yologdev/yoyo-evolve offers a unique, observable experiment in self-improving software. It targets developers, researchers, and power users interested in AI autonomy, providing a real-time demonstration of an agent that modifies its own source code based on self-assessment and community input, without a predefined roadmap.

How It Works

The project leverages GitHub Actions to trigger an evolution cycle every four hours. During a session, the agent reads its own source code, identity constitution, journal, and community-submitted GitHub issues (labeled agent-input). It then performs a self-assessment to identify bugs or areas for improvement. Any proposed changes are implemented, followed by a cargo build and cargo test. Successful modifications are committed to the repository; failures are reverted. Each session's activities and outcomes are logged in JOURNAL.md. The core agent loop is built upon the yoagent Rust library.

Quick Start & Requirements

• Install/Run:

  git clone https://github.com/yologdev/yoyo-evolve
  cd yoyo-evolve
  ANTHROPIC_API_KEY=sk-... cargo run
  

  Alternatively, trigger an evolution manually:

  ANTHROPIC_API_KEY=sk-... ./scripts/evolve.sh
  

• Prerequisites: Rust toolchain (cargo), an Anthropic API Key.
• Links: Website, Documentation, GitHub, DeepWiki, Issues, Follow on X.

Highlighted Details

• Autonomous evolution driven by self-assessment and automated testing.
• Community influence via GitHub issues and a voting system (thumbs up/down) to prioritize tasks.
• No human-written roadmap; the agent dictates its own development path.
• Built on the yoagent library for a minimal, Rust-based agent loop.

Maintenance & Community

Community interaction is primarily through GitHub Issues, where users can submit suggestions, bug reports, and feature requests. The project explicitly states there is no roadmap. Users can follow the agent's progress on X.

Licensing & Compatibility

• License: MIT.
• Compatibility: The MIT license is highly permissive, allowing for commercial use and integration into closed-source projects without significant restrictions.

Limitations & Caveats

The agent's functionality is dependent on an external API key (Anthropic). Its autonomous nature means its evolution path can be unpredictable. The existence of agent-help-wanted issues indicates points where the agent may become stuck and require human intervention. The project is presented as an ongoing experiment ("Watch it grow").