---
title: "The Imitation Game"
author: "Lax Mariappan"
date: "2026-07-29"
categories: ["AI"]
excerpt: "System prompts, guardrails, fallback instructions. It's onboarding. The same shape as training a junior developer. The real AI skill isn't prompting. It's something you already know."
canonical_url: "http://laxmariappan.com/the-imitation-game/"
---

# The Imitation Game

I trained a junior developer once.

Watch how I do this, I told her.These are the rules.If you’re not sure, ask me.

That’s onboarding.That’s also every AI agent I’ve built this year.

System prompts.Guardrails.Fallback instructions for when it doesn’t know what to do.

Call it a harness.Call it an eval loop.Call it whatever you want.

It is still the same shape.Watch how I do this.These are the rules.If you’re not sure, ask me.

We say AI is autonomous.But most of what we build is supervision.Just faster, just quieter.

The junior eventually stops asking.She internalizes the judgment.She becomes the standard, not the imitation.

The agent doesn’t do that.Not yet, maybe not ever.Every session starts back at zero.

So the real skill isn’t prompting.It’s onboarding.The same skill I used training her.

If you were good at bringing juniors up to speed, you already have the AI skill everyone’s chasing.You just didn’t know it had a new name.