Did I Over-Engineer This? Refactoring AI Code from Monolith to Data-Driven

In this video, I break down the journey of building a SOAP architecture workflow viewer using Google Gemini—starting from a single 800-line static HTML prototype in Gemini Canvas to a fully data-driven, modular app using VS Code, Gemini CLI, and Lit-HTML. Along the way, I dive into the reality of coding with AI: managing token limits, structuring prompts, separating the data plane from the control plane, and why frequent Git commits are non-negotiable when working with LLMs. I also share thoughts on software engineering philosophy: the expansion and contraction of code, the limits of YAGNI ("You Ain't Gonna Need It"), and why exploratory engineering is worth doing even when you end up writing more code. *What's covered:* Prototyping diagrams and UI with Mermaid.js in Gemini Canvas Iterative refactoring with the Gemini CLI tool in VS Code Converting hardcoded layouts into an XML-driven page builder Architecture vs. feature delivery in modern software development Surviving token caps and building healthy AI development habits 💬 Let me know in the comments: Do you think refactoring AI-generated code to be extensible is worth the token burn, or should we stick to quick, monolithic prototypes? Links https://github.com/chrisseanhayes/soa... --- Timestamps `00:00` - Always Running Out of Tokens `00:48` - Why I Built a SOAP Workflow Viewer `01:35` - Prototyping in Gemini Canvas with Mermaid.js `02:59` - Moving to VS Code and Gemini CLI `04:26` - Turning a Monolithic Page into a Data-Driven App `06:58` - App Walkthrough & Stepper Navigation `08:58` - Was This Over-Engineered? (The Reality of YAGNI) `12:13` - Why Maintainability is 95% of Software Engineering `17:08` - The Natural Expansion and Contraction of Code `19:55` - Separating Data Plane from Control Plane with AI `22:07` - Why You Must Commit Constantly When Working with LLMs `24:19` - Reviewing the Git Commit History & Late-Night Token Burn `28:06` - Final Thoughts: The Value of Exploratory Engineering --- Hashtags `#SoftwareEngineering` `#GeminiAI` `#WebDevelopment` `#CodingWithAI` `#Refactoring` `#Programming` `#TechCareer`