2. my-generative-coding-workflow

My Generative Coding Workflow: An Old School Coder Adapting to the AI Age

1 min read

My Generative Coding Workflow: An Old School Coder Adapting to the AI Age

They say you can't teach an old dog new tricks. Well, after two decades deep in web development’s trenches, I’m here to tell you: when the "trick" is as game-changing as AI, you don’t just learn it—you evolve.

My journey as an "old school coder" becoming a Generative Coder has been thrilling, bewildering at times, and always a learning experience. Today, I’ll share my current AI-assisted development workflow: the tools I rely on, the process I follow, and where I’m actively refining it to make it sharper and smarter.

My AI Toolkit: The Brains Trust

My daily development is now a dance between trusty veterans and agile AI tools:

  • Gemini — My brainstorming buddy for broad ideas and conceptual frameworks.
  • Claude (via Cursor IDE) — The workhorse for actual code generation. Quick, context-sensitive, and aligned with my prompt style.
  • GPT — My second opinion and problem-solving sidekick. When I’m stuck or need a fresh perspective, GPT or Gemini brings a different lens.

IDE Showdown: Cursor vs. WebStorm

  • WebStorm — My long-time partner, unbeatable for diffing, deep project insights, and precise code management. And yes, WebStorm has AI too, but the apply process is appreciably slower.
  • Cursor — The AI-powered rocket. Cursor’s speed, “apply to this” feature, and tight AI integration make the initial coding phase feel like I’m coding at warp speed.

The contrast is stark: after a sprint in Cursor, returning to WebStorm can feel like hitting the brakes. It’s a vivid reminder of how AI is transforming the initial coding phase — even though WebStorm’s AI is helpful, it just can’t match the seamless acceleration Cursor provides.

My Generative Coding Workflow: A Three-Step Dance

Here’s my current flow, especially for projects like KODKAFA’s API or Python NLP tools:

  1. The AI Huddle — I talk to Gemini, laying out what I need, constraints, and questions. This clarifies the problem and often uncovers hidden edges.
  2. Charting the Course — I translate the AI insights into user flows and structured requirements. Human-first, AI-informed.
  3. The Generative Relay — The real coding begins:
    • Prompting Claude in Cursor for initial code.
    • Reviewing and iterating rapidly.
    • Turning to GPT/Gemini for second opinions or alternative approaches.
    • Repeat until done.

It’s less like bricklaying, more like conducting a symphony of smart tools.

The Road Ahead: Refining the Process

This workflow works, but it’s not perfect. My current goals:

  • Sharpening my prompts to get better first-shot results... well, the first-shot result is a dream. I want to refine the process because I’m a damn engineer. It should be perfectly efficient: fast, precise, crystal-clear development. Sure, I can crank out a full 100 Lighthouse, but it still takes a lot of effort.
  • Streamlining review techniques for AI code.
  • Smarter tool integration to reduce iteration cycles.

The goal? Fewer re-prompts, tighter loops, more time for big-picture design.

✓1.0.2