---
id: "concept-parenting-ai"
type: "concept"
source_timestamps: ["01:25:50"]
tags: ["prompt-engineering", "mindset"]
related: ["concept-ai-agent", "framework-parenting-ai", "quote-parenting-ai", "action-build-tone-mirror"]
definition: "The approach of training an AI model through context, examples, and iterative feedback, rather than relying on rigid, technical programming commands."
---
# Parenting AI (vs. Programming)

A paradigm shift in how to interact with Large Language Models, from [[entity-lior-weinstein|Lior Weinstein]].

## The Argument

LLMs are essentially **prediction engines** — guessing the next most likely word based on vast training data. Treating them like traditional software that requires rigid "programming" or complex coding is the wrong approach.

Instead, treat AI like a **highly capable but inexperienced human assistant or a child**. You *parent* the AI by:

1. providing rich context about who you are,
2. giving clear examples of what good output looks like,
3. providing iterative feedback when it gets things wrong.

This human-centric approach yields far better, more personalized results than chasing a "magic prompt".

## Captured Slogan

> [[quote-parenting-ai|"Don't program it. Parent it."]]

## Operationalized

The [[framework-parenting-ai|Parenting AI Framework]] gives the three concrete steps. The [[action-build-tone-mirror|Tone Mirror prompt]] is a worked example of parenting in action — feeding the AI your style and refining via feedback.

## External Caveat

Iterative "parenting" still risks producing outputs that look polished but mask shallow understanding ("illusion of mastery"). Don't let beautiful output substitute for your own thinking.

## Definition

> The approach of training an AI model through context, examples, and iterative feedback, rather than relying on rigid, technical programming commands.