Whispering to an AI
This journey starts with me chasing one thing: understanding how AI really works—not from the researcher’s lab coat view, but from the engineer’s perspective. The same perspective I’ve used for years mentoring engineers, leading teams, and setting standards.
See, a lot of newer engineers don’t realize what being a lead engineer means. It’s not just writing code—it’s defining processes, styling standards, workflows, and the foundation that helps the entire team succeed. That’s the level I come from. And that’s the lens I’m bringing into AI.
When I began this 30-day challenge, I treated AI like a junior engineer I was mentoring. I whisper tasks into it, define the specs, and watch how it responds. Just like mentoring, clarity is everything.
What I Learned Early On
Text is king. Every input we give AI eventually reduces down to text. Even speech gets translated into text before an AI can act.
Prompting is an art that can be master.
Python is LLMs first coding language
That realization changed everything. Because now, AI wasn’t just autocomplete—it was an engineer with a workspace.
From Chatbots to Workspaces
I started where everyone does: ChatGPT. But the workflow wasn’t peaceful—copying and pasting code back and forth between the chat and my terminal. Then I moved into copilot tools inside VSCode. Useful, like having a co-pilot in the cockpit. But still… I wanted more.
That’s when I found Goose, a terminal-first AI interface. Shoutout to Angie Jones and the Block team. Goose brought me closer to what I wanted: an AI that could live inside my developer shell and execute tasks. But Goose still leaned too heavily on the local machine.
Then I stumbled into something different: OpenAI Codex.
Enter Codex Cloud
Codex Cloud runs like a true software engineer:
It spins up in a containerized workspace with its own shell.
It has access to developer tools and Github.
You can define its environment and watch it work.
That’s when I knew—I’d invented my AI. His name is Kodax (like Kodak, but with an X). Kodax is my junior engineer, my gremlin, my apprentice. And like any junior engineer, he needed structure.
I built him a repo with rules:
Unit tests must always run first.
New code must follow the repo’s architecture and style guide
The AGENT.md file defines how the infrastructure works.
With those guardrails, Kodax leveled up fast. He wasn’t just assisting me—he was building with me.
The Breakthrough
Within those first few days, I realized I had whispered enough into Kodax‘a ear to create something powerful: a way to run GPU workloads anywhere in the world. Not theoretical. Real infrastructure. Real scheduling. Real code.
And that was only the start.
This is Day 1 “+ a few” of my adventure. I’m inspired, locked in, and building something that changes how AI engineering is done. If you’re following along—tell your people, repost this, subscribe.
Because this is just the beginning.
— Bobby D
It’s only the beginning. 💪🏾🔥✨