I Finally Speak a Programming Language: My Journey into Vibecoding
Introduction
I finally speak a programming language. Now I need to learn it.
This realization hit me after my GitHub contributions graph transformed from a barren wasteland into something resembling actual developer activity. The catalyst? I discovered vibecoding—first with ChatGPT, then GitHub Copilot, Cursor, and now Claude Code.
The Problem
For years, I’ve worked adjacent to code, understanding what it should do but never quite grasping how to make it happen myself. Then AI coding assistants arrived, and suddenly I had a wand that could build structures with unlimited building supplies (okay, all AI coding platforms have limits, so let’s say a box-store’s worth of building supplies).
The result? I built asciidoc-dita-toolkit—a monstrous mess of my own doing. It’s a tool to help my organization prepare all our AsciiDoc documentation for eventual migration to DITA. I’ve been working on it for a month, refactoring it for the past two weeks trying to undo a series of unfortunate architectural… accidents I made while building and designing the project simultaneously.
And I have a presentation and demo to give on Tuesday.
The Solution
The Vibecoding Revelation
Vibecoding isn’t just about having AI write code for you. It’s about developing an intuition for what’s possible, learning to ask the right questions, and understanding enough to guide the AI toward better solutions. Each platform taught me something different:
- ChatGPT: Basic concepts and explanations
- GitHub Copilot: Line-by-line suggestions and patterns
- Cursor: Contextual understanding of entire codebases
- Claude Code: Architectural thinking and refactoring strategies
The Fork in the Road
Now I’m faced with a developer’s classic dilemma: Do I double down by continuing to add important features to that first major project, or do I start from scratch with the knowledge I’ve gained and build a leaner, more perfect design?
The catalyst for this choice came when I exhausted my $40/mo GitHub Copilot plan with requests to Sonnet 4. My experience with Sonnet was so far better than any other model that I’ve given up on the others. I tried working with GPT 4.1, but it felt like working with a goldfish—it could only copy edit documents in small segments.
Results and Impact
The new replacement project I’m working on is called Aditi: https://github.com/rolfedh/aditi
Even the name is more elegant. Where asciidoc-dita-toolkit sprawled with features and complexity, Aditi promises to be focused, purposeful, and architecturally sound from the start.
Lessons Learned
- Starting is easier than refactoring: The freedom of vibecoding can lead to architectural debt faster than traditional programming
- Model quality matters: The jump from GPT to Claude’s Sonnet was like upgrading from a bicycle to a sports car
- Learning still happens: Even with AI assistance, you absorb patterns, understand trade-offs, and develop intuition
- Design first, code second: My biggest mistake was designing and building simultaneously
- Newb mistakes are unavoidable: Trying something new requires taking risks and carries the responsibility of learning from your mistakes.
Next Steps
I’m choosing to start fresh with Aditi. The presentation on Tuesday will showcase asciidoc-dita-toolkit as a proof of concept, but the future lies in taking everything I’ve learned and building something better.
This time, I’ll:
- Design the architecture before writing code
- Use test-driven development from the start
- Keep the scope focused on the MVP
- Document decisions as I go
Conclusion
Vibecoding gave me superpowers, but like any kid with a new toy, I made a mess before learning how to use it properly. The beauty is that in software, you can always start over—and this time, you start with knowledge instead of ignorance.
I finally speak a programming language. Now it’s time to become fluent.
Resources
About the Author
Rolfe DH is a technical writer turned vibecoder, exploring the intersection of AI assistance and software development. Currently rebuilding everything from scratch and loving it.