The world of software development is buzzing with the promise of Artificial Intelligence, and like many of you, I've been exploring the growing landscape of AI-assisted development tools. As a .NET/ASP.NET Core developer and open-source contributor, I've been piecing together a workflow that helps me leverage these powerful assistants. It's been a journey of discovery, full of pleasant surprises and a few hurdles.
Here's a look at the tools in my current rotation and the workflow I've settled on for now.
My AI Toolkit: A Review of Development Assistants
Vibe Kanban: The Mission Control
I start most of my tasks in Vibe Kanban. Its strength lies in its flexibility. I can connect it to multiple repositories and, more importantly, assign different AI models to different tasks. I have access to GitHub Copilot, Gemini, and Claude, and Vibe Kanban lets me switch between them seamlessly. This is perfect for running multiple tasks in parallel, though I've found that reviewing the generated output is still a time-consuming but necessary step.
Claude: The Master Planner
Claude is my go-to for planning. I find it's excellent at understanding programming concepts and laying out a structured plan. The main drawback is the token expiration, which can sometimes be a bit of a nuisance. If I have enouhg limit, mostly I ask Claude to implement Phase 1 of the plan it created.
Cursor: The Coder
When it's time to write code, I turn to Cursor. Its "auto mode" is a standout feature, as it's incredibly efficient with token usage, which is a major concern in AI-assisted development. However, as a .NET developer, I sometimes find it challenging to work on my projects within Cursor and occasionally have to switch back to Rider or Visual Studio for certain tasks.
Gemini: The Technical Writer
Interestingly, I don't use Gemini much for coding. Instead, it has become my trusted partner for writing technical articles and documentation. In this area, it truly excels.
GitHub Copilot: The Dependable Backup
I use GitHub Copilot less frequently than the others, but it's a reliable tool for smaller tasks, especially when I've run out of tokens on other platforms.
My AI-Powered Development Workflow
My process has evolved into a multi-stage workflow that leverages the strengths of different AI tools:
- Plan with Claude: I start by asking Claude to create a detailed plan in a markdown format using Vibe Kanban.
- Review and Refine: I thoroughly review the plan. If any adjustments are needed, I'll work with Claude to refine it.
- Branch and Commit: Once the plan is solid, I create a new branch and commit the plan file. This provides clear context for the work that follows.
- Code with Cursor: With the plan in hand, I open Cursor and begin the development work, following the roadmap laid out by Claude.
- AI-Assisted Review: After the initial coding is done, I use another AI to review the changes and suggest improvements.
- Manual Verification: The final step is always a manual review and testing of the functionality to ensure everything works as expected.
What's Next?
My next goal is to automate the testing phase. I plan to create a process that runs automated tests before I begin my manual review, which should help streamline the final verification step.
What's Your Workflow?
This is just a snapshot of my current process, and I'm sure it will continue to evolve. I'm curious to hear about your experiences. What tools are you using for AI-assisted development, and what does your workflow look like? Feel free to share your thoughts and recommendations!
Top comments (0)