A 10-hour experiment explores fully automated AI-driven software development, where all code, prompts, and project structures are AI-generated. It targets developers interested in leveraging AI for code creation, shifting the programmer's role towards AI control and direction rather than manual coding. The project aims to demonstrate a novel, potentially accelerated development paradigm by showcasing AI's capabilities in complex project execution.

How It Works

The core methodology employs an AI agent, specifically Claude Code, to iteratively develop projects based on a task.json file and track progress in progress.txt. The workflow encompasses environment initialization, task selection, AI-driven code implementation, automated testing (using npm run lint, npm run build, and playwright mcp), and committing changes. This approach is exemplified by the hello-nextjs application, which integrates full-stack features and external AI model APIs.

Quick Start & Requirements

• Primary Install/Run: Requires claude code and playwright mcp. Execution can be manual, via --dangerously-skip-permissions for autonomous task completion, or through the run-automation.sh script for unattended, looped operation.
• Prerequisites: claude code, playwright mcp.
• Setup Time/Footprint: Not explicitly detailed, though the core experiment spanned 10 hours. The hello-nextjs project involves front-end, back-end, database, and external AI APIs.
• Links: No direct links to official quick-start or demo pages are provided.

Highlighted Details

• All code and prompts (100%) are AI-generated, including project structure and task definitions.
• Features a 10-hour experiment demonstrating AI's capability in complex, full-stack project development.
• Includes a sample hello-nextjs application integrating various AI services (image/video generation).
• Explores advanced AI control methods, including a "dangerously skip permissions" mode for autonomous task completion.

Maintenance & Community

No specific details on contributors, sponsorships, community channels (like Discord/Slack), or roadmaps are present in the provided text.

Licensing & Compatibility

The README does not specify a license type or any compatibility notes for commercial use.

Limitations & Caveats

All code and prompts are AI-generated and require manual review before execution due to potential inaccuracies or unforeseen consequences. The automated execution script (run-automation.sh) is explicitly flagged as dangerous and prone to resource wastage. The project's premise relies on the assumption that AI can handle complex projects, which may not hold true for all use cases.