How a coaching framework designed by AI led to a 10-year-old independently discovering real automotive engineering — and actually having fun doing it.

The Skittles Moment

My 10-year-old was supposed to do a science fair project.

Instead, he bought Skittles and melted them on a paper plate. He thought that was the experiment.

There was a 14-page assignment packet sitting in his backpack — testable question, hypothesis, controlled variables, data tables, graphs, tri-fold display board. The whole thing. He'd never fully understood what was being asked.

My son has always had a hard time focusing. Long assignments overwhelm him. He'd misunderstood the project months ago, and nobody caught it until the deadline had passed.

His teacher gave him a second chance. A day and a half to redo the entire thing.

What I Actually Asked AI to Do

I'm a mechanical engineer with an applied physics degree. I spent years working at an automotive startup. When it comes to science, I know my stuff.

And that was the problem.

My instinct was to just take over — to lay out the experiment, explain the science, and guide him step-by-step through a rigorous project. I would have done an excellent job. And my son would have learned almost nothing.

Instead, I opened Claude and uploaded all 14 pages of the assignment packet.

I didn't ask AI to write answers. I said: "Help me make this fun for my kid — and design a coaching strategy so I can teach him without taking over."

What I was really asking for was empathy. A way to stop seeing the assignment from my perspective and start seeing it from his.

6 Rounds of Iterative Refinement

Over six rounds of conversation, I gave AI more and more context:

Round 1: I uploaded the assignment. AI gave me 3 experiment options with full procedures, data templates, and a focus-friendly schedule with short work blocks.

Round 2: I wanted more Lego-based options. AI gave me 3 more, with a comparison matrix rating fun factor, cost, prep speed, and "mechanical energy relevance."

Round 3: I shared my engineering background. AI built a custom coaching playbook with word-for-word scripts and 8 teaching rules designed specifically for an engineer parent of a kid who struggles to focus.

Round 4: I listed my gear — two iPhones, iPad Mini, MacBook, Insta360 Link 2, Insta360 Flow 2 Pro gimbal. AI assigned every device a specific role.

Round 5: My son suggested building a Lego city as the test track. AI validated the idea and explained why the Lego street plates actually improved the science — consistent rolling surface as a controlled variable.

Round 6: I realized this was content-worthy. AI mapped the narrative arc and dual-camera capture strategy.

Three Word documents. Over 1,400 paragraphs. All custom to one 10-year-old.

The 8 Teaching Rules (Empathy-First Coaching)

AI didn't give me generic tips. It designed rules specifically for our situation — an engineer dad who knows too much and talks too long, teaching a kid who learns by doing:

1. Hands first, words second. Let him BUILD before explaining anything.
2. Two-minute lecture limit. If you're talking for more than 2 minutes, you've lost him.
3. His language, not yours. Don't say "kinetic energy." Say "how fast the car goes."
4. 30-second career stories. "At the car company, we tested this exact thing" — then stop.
5. Predict before every trial. "What do you THINK will happen?" on camera.
6. Talk-then-write hack. He talks, I type. He copies. Writing is the bottleneck — talking isn't.
7. Pomodoro blocks. 15 minutes on, 5-minute movement break.
8. Reward checkpoints. Every section = iPad time or a snack.

These rules weren't about managing behavior. They were about understanding how my son experiences learning — and adapting my teaching to match.

Building the Lego City (Making Science Invisible)

The empathy shift AI helped me make was simple but powerful: start with what he loves, not with what the assignment says.

My son already had a Lego car and street plates. So instead of saying "we need to redo your science project," I said: "Hey, want to build a city for your car?"

He was in immediately.

What he didn't know was that those Lego street plates were about to become his test track — and the consistent rolling surface actually made his experiment more scientifically valid than most 5th graders' projects.

A cutting board became the ramp. The Lego city was the landing zone.

9 Trials and a Wrong Hypothesis

He predicted: the tallest ramp will make the car go the farthest.

Makes sense, right? More height = more speed = more distance.

We set up three ramp heights:

• 7 inches (13 degrees)
• 11 inches (32 degrees)
• 12 inches (36 degrees)

Before every roll, I asked: "What do you THINK will happen?" And after every roll, he told me what he noticed — excitedly, in his own words, without being prompted.

"Dad, that one went WAY farther!" "Wait, why did that one go sideways?"

He was narrating his own learning without realizing it.

The Discovery That Changed Everything

The middle ramp won. The tallest ramp? The car went sideways. Traveled less distance.

Before I could even ask why, my son started explaining:

"Because the car scratches on the bottom and it creates the friction so it goes less faster and doesn't go straight anymore."

Let that sink in.

A 10-year-old — a kid who has a hard time focusing on worksheets — just independently identified breakover angle and ground clearance effects. A real concept in automotive engineering. Without anyone teaching him.

He figured it out by observing. By watching the car scrape the transition from ramp to track at steep angles. And he told me about it — excitedly, in his own words — before I could even ask.

The Quote That Made It All Worth It

When the whole thing was done — 9 trials completed, data tables filled in, averages calculated by hand, display board planned — he looked at me and said:

"That was a really fun experiment, Dad."

A kid who struggles to focus on schoolwork called the most rigorous project of his year fun.

Not because I dumbed it down. Not because AI did it for him. Because somebody finally made the effort to see the assignment through his eyes.

What This Means for AI in Education

We're having the wrong conversation about AI in education. The question isn't "should kids use AI?" The question is: how do we use AI to help parents see their kids more clearly?

AI's most powerful use case in education isn't student-facing. It's parent-facing. It's an empathy engine.

• AI as a parenting translator — converting complex assignments into coaching strategies tailored to YOUR kid
• AI as a mirror — showing parents their blind spots (mine: explaining instead of asking)
• AI as an empathy engine — because "how to teach a kid" is useless, but "how to teach THIS kid who struggles to focus, who loves Legos, from an engineer dad who lectures" is powerful

My son thought for himself. AI just made sure I let him.

Get the Framework

If this resonated with you — especially if you're a parent who "knows too much" about a subject and accidentally teaches the way you learn instead of the way your kid learns — I've packaged the complete coaching framework AI designed into a free guide.

Comment "SCIENCE" on any of my social posts this week, or reply to my newsletter, and I'll send it directly to you.

The teaching rules, the coaching scripts, the gear deployment plan, and the recording strategy that made my son feel like a YouTuber instead of a student — it's all in there.

Jonesy is a PMO Director, Chief of Staff, and Doctor of Strategic Leadership candidate, transitioning to AI leadership consulting through AI Mastery Academy. He writes about practical AI wins — the kind that change how you work, parent, and lead.