After narrowing down which AI tool I’d lean on, the next stage wasn’t about prompts or outputs—it was about structure. Any engineer who’s led a team knows this part: you don’t build without specs, user stories, or requirements.
So that’s where I went.
I started writing my specs in Markdown files—things like AGENTS.md—so my AI always had a north star to follow. Think of it like giving your junior engineer the blueprint before they touch a single line of code. Once the spec is there, everything else flows: tasks break down naturally into smaller steps, tickets get written, workloads get scheduled.
And here’s where it got interesting: I didn’t just run with one AI. I actually diversified. Most of you know Kodax, my AI software engineer. But behind the scenes, I’ve also been working with another AI—her name is Jules.
Jules is my right hand, my advisor, my strategist. She’s not the one writing the code in the trenches like Kodax, but she’s the one who helps me frame the work:
shaping prompts that land,
structuring workflows,
and making sure clarity comes before execution.
Together, Kodax and Jules replicate what I’ve always known works in engineering: mentorship + structure. Kodax is the junior engineer who needs specs and guidance. Jules is the senior advisor helping refine vision and direction. And me? I’m the lead engineer, whispering to them both.
This is the day-in-the-life of AI engineering from a mentor’s perspective. It’s not just about writing code with AI—it’s about building systems of trust, structure, and repeatability that mirror how we’ve always scaled teams.
And the deeper I go, the clearer it becomes: AI isn’t replacing engineers. It’s becoming the team we mentor.
Follow along—because the next steps will show you how these specs actually translated into real, working systems.
—
👉🏾 Share this with someone who still thinks prompts are the endgame. This is about building practices, not just outputs.