The Quarter on the Football Field:

**The Quarter on the Football Field:

What a College Thought Experiment Taught Me About AI, Engineering, and Learning**

When I was a young engineer in college, someone handed me what I thought at the time was a strange, almost contrived mental game.

It sounded like one of those esoteric puzzles you talk about late at night in a dorm room — half philosophy, half math, half nonsense.

Here was the challenge:

“You can hide a quarter anywhere on a 100-yard football field.

I’ll find it using only four balls.”

Naturally, I said, “You’re kidding.”

He wasn’t.

The Game

The rules were simple:

1. Someone hides a quarter anywhere on the football field.

2. I throw four balls anywhere I want onto the field.

3. The person who hid the quarter tells me which ball is closest to it.

4. I repeat the process — throwing four new balls — but now I focus on the region around the “closest” ball.

5. Over and over again… until I find the quarter.

That’s it. No coordinates. No distances. No compass. No GPS.

Just one piece of feedback each round:

“That one is closer.”

At first, I thought it was ridiculous.

Then I tried it.

I Actually Did This (and It Worked)

Believe it or not, I went to an actual football field and tried it.

I hid a quarter.

I threw four balls.

I asked which one was closest.

Then I threw four more balls near that one.

And I kept going.

After enough rounds, the search area collapsed down to a few feet… then inches… and then there it was.

I found the quarter.

I remember standing there thinking:

“Okay… that’s not a trick. That actually works.”

Why I Thought It Was Contrived

At the time, I still felt uneasy about it.

Because here was my objection:

“Well, of course it works.

The person who hid the quarter is telling me which one is closer.

I’m just honing in on an answer I’m obviously going to get eventually.”

It felt like cheating.

It felt staged.

It felt like the outcome was baked in.

So I dismissed it as a clever but meaningless curiosity.

I was wrong.

The Insight I Didn’t Understand Back Then

Here’s what I didn’t realize as a young engineer:

That little game is a perfect demonstration of how learning and search actually work.

You don’t need exact answers.

You don’t need full information.

You don’t need coordinates.

You only need one thing:

Directional feedback.

“Closer.”

“Farther.”

“More like this.”

“Less like that.”

Each round collapses the search space.

Each piece of feedback adds information.

Each iteration narrows uncertainty.

That’s not cheating.

That’s intelligence.

Fast Forward 30+ Years: Hello, AI

Here’s the part that stopped me in my tracks recently.

That football-field game is exactly how modern AI prompting works.

When you talk to an AI and say:

• “That’s close, but sharper.”

• “More opinionated.”

• “Less corporate.”

• “More like my voice.”

• “That answer was better than the last one.”

You are doing the same thing I did with the four balls.

You’re not giving the final answer.

You’re giving directional constraints.

And the model converges.

Why This Matters (and Why It’s Not a Bad Thing)

Here’s the big philosophical point — the one I missed back then.

Yes… the system is “rigged” to converge.

Yes… you are honing in on an answer you’re eventually going to get.

Yes… feedback is steering the outcome.

But that’s not a flaw.

That’s the whole point.

That’s how:

• Humans learn

• Engineers debug

• Teams align

• Systems improve

• Models converge

• Understanding deepens

There is no intelligence without feedback loops.

There is no learning without constraint.

There is no mastery without iteration.

The Real Lesson for Engineers (and for AI)

The lesson isn’t that AI is magical.

The lesson is this:

Intelligence doesn’t come from knowing the answer.

It comes from being able to recognize when you’re getting closer.

That was true on a football field with a quarter.

It’s true in software architecture.

It’s true in product design.

It’s true in leadership.

It’s true in life.

And it’s true in AI.

Why This Fits Perfectly with Learn • Teach • Master

This is why I keep saying:

We don’t need “smart machines.”

We need learning machines.

Machines that:

• Accept feedback

• Adjust direction

• Narrow uncertainty

• Improve through iteration

• Stay steerable

• Stay corrigible

That little college thought experiment accidentally taught me the core principle behind modern AI — decades before AI existed in any real form.

The Ending I Finally Understand

So let me answer my younger self’s objection:

“Isn’t this just honing in on an answer I was always going to get anyway?”

Yes.

And that’s exactly why it works.

That’s not cheating.

That’s learning.

That’s engineering.

That’s intelligence.

Final Thought

Most people think the power of AI is in the model.

It isn’t.

The power is in the feedback loop.

And I learned that lesson a long time ago —

on a football field —

with four balls —

and a quarter.