My first step in AI-driven indie development

Introduction

I recently developed a task management module called task-orchestrator. I created this module as a base for building an algorithmic trading framework. It can manage two types of tasks:

1. Continuous tasks that run in the background.
2. Event-triggered tasks that start when a specific event happens.

During the development process, I used AI tools for planning, coding, testing, and writing documentation.

Planning

At the beginning, I was not sure which architecture to use for the framework. I considered the actor model and the event-driven model. I asked ChatGPT, Claude, and DeepSeek for advice. They suggested a hybrid approach, combining both models. Based on this idea, I wrote the specifications and improved them through discussions with AI. Finally, I asked Claude to generate a basic code structure using my final specifications.

Coding

During implementation, I used Claude's code as a base. I also used aider to make necessary changes and add features. Some specific tasks I did:

• Making small adjustments
• Renaming the project
• Creating unit tests

For documentation, aider generated most of it, so I only needed to make small edits, like removing unnecessary links.

Conclusion

Through this process, I realized how useful AI tools are. They helped me write test code and documentation efficiently.

Here are the tools I used:

• Editor: Helix (with lsp-ai)
• AI tools: Aider (for coding, testing, documentation) and DeepSeek API
• Specification discussions: ChatGPT, Claude, and DeepSeek (Web-based)