...and no sooner than I got this project off the ground, the AI dev world once again morphed. Jules, while not officially deprecated, has been effectively superseded by Google's Antigravity, specifically Antigravity's Manager View. The nice thing about this is that you can now conduct asynchronous and autonomous code experiments locally without requiring a GitHub repo. This is good since you might want to keep many experiments you conduct under wraps. Positronikal maintains a corresponding private GitHub repo for those experiments, but Antigravity means neither the public nor the private repos are necessary any longer. For what it's worth, If you're a Project IDX (Google's cloud IDE) or Gemini Code Assist (for VS Code), Antigravity means you can ditch those, too. Such is the rapidly developing and ever changing world of AI we live in nowadays.
This repository is dedicated to autonomous Artificial Intelligence code research and experimentation utilizing advanced AI agents. The primary goal is to rapidly prototype, test hypotheses, and generate functional code snippets related to modern computing challenges, while leveraging AI to both develop and refine software, tools, methods, and solutions.
- Research/Experiments Architect: Claude Sonnet 4.5
- Experimental Researcher: Google Jules
- Code Reviewer: GitHub Copilot
- Documentation/Summaries: GPT 4.1 Nano
- Core Language: Python (enforced with Ruff formatting and linting rules defined in
pyproject.toml). - Methodology: All code is generated in response to specific research prompts.
- Contribution Process: The AI agent directly commits its work via Pull Requests (PRs) as long as the working repository enforces it, which serve as the primary audit log for each experiment.
⚠️ WARNING: This is a research environment. Code quality is secondary to experimental outcome. DO NOT use any code from this repository in production or high-stakes environments without rigorous human review, testing, and validation.- License: All code is published under the MIT License (see
LICENSEfile for details). - Discussion: Use the Discussions tab to propose new research questions, discuss experimental results, or provide feedback on the AI agent's performance.
- Seeking Contribution: If you wish to contribute test validation scripts or refined versions of the AI's output, please open a separate Pull Request targeting the specific experiment's folder.
The Documentation/Summaries agent (GPT 4.1 Nano) automatically generates project descriptions which are recorded in _summary.md in the target experiment's root home subdirectory.
A GitHub Action auto-triggers an update of the repository's README.md (this file) to add or update new target experiment information on every push to main.
To manually update the local repository that can then be manually pushed to GitHub, install Cog, then:
cd /path/to/repo
# Run cogapp to regenerate the project list:
cog -r -P README.mdThe embedded script in README.md automatically:
- Discovers all subdirectories in the repository root
- Gets the first commit date for each subdirectory and sorts by most recent first
- Checks subdirectories for a
_summary.mdfile - If the summary exists, it uses the cached version
- If not, it generates a new summary using:
# Includes prompt that creates engaging descriptions with bullets and links llm -m openai/gpt-4.1-nano - Creates Markdown links to each project folder on GitHub
- New summaries are saved to
_summary.mdto avoid regenerating them on every run
To regenerate a specific project's description, delete its _summary.md file and run cog -r -P README.md again.