AI "Apprenticeships" Creating Winners and Losers

Twenty thousand years ago a dad sat upon the earth... shaping flint with a piece of bone.

His son watched.

The boy took up the stone and struck it. The flint shattered.

He took a new stone and showed him how to do it again.

The years passed...

Until one day the young man shaped a point as fine as any his father had made.

His son grew into a hunter.

Stone Age Dad knew his days upon the earth would end...

But rejoiced knowing something of him would remain in the hands of his son.

Long before the age of AI, skills were passed from generation to generation.

For most of human history, that was how you became competent at anything.

From one human brain to another.

Eventually, we figured out how to save practical knowledge outside the brain...

Because writing allowed instructions to survive long after their authors died.

An early example...

In 1677, Joseph Moxon began publishing Mechanick Exercises.

It explained how to work with metal, construct things from wood, and use the tools of skilled tradesmen.

Later volumes even showed you how to set up a print shop to create more instruction books.

Complete with illustrations...

Aspiring craftsmen could learn techniques from an expert they'd never met.

Even one who'd been dead for centuries. But books had a limitation.

A book could explain where to strike a piece of flint.

It couldn't tell you that you'd chosen the wrong stone, the wrong tool, the wrong angle, the wrong pressure.

Today, a quick YouTube search for "flint knapping" produces page after page of videos that would make a Stone Age toolmaker's jaw drop...

But the video still can't see what you're doing as a mentor.

Until now...

That's the limitation that computer vision is about to change...

Your New AI Apprenticeship

The sons of tomorrow will have a very different apprenticeship.

AI that can see.

Imagine your son, off in college, standing in front of a broken light switch in his first place.

A screwdriver in one hand. A small AI device in the other.

The device has a camera, a microphone, a processor, and enough memory to run a compact AI model.

As we saw on Tuesday, it might look like this...

He aims it at the switch.

"Why isn't this working?"

It identifies the type, retrieves the wiring guide and helps him figure out what's wrong.

It guides him through the safe troubleshooting steps... and from home to the hardware store and back again.

A patient instructor, watching and explaining. Available whenever he needs it.

And when the job is done, he'll have learned something new.

Dad won't pass this knowledge along.

It'll be a device like this.

Your AI Mentors

Of course, today's AI isn't reliable enough (yet) to safely supervise every repair.

Electrical work, in particular, requires training and safety precautions.

But the direction is unmistakable.

AI is gaining the ability to see the physical world, understand natural language, and provide contextual guidance.

In fact, OpenAI and other technology companies are exploring AI-first hardware that could provide an always-accessible companion.

Some devices may use glasses. Others could be worn around the neck or clipped in a pocket. Others look like a portable donut. Some will have screens for feedback.

The final form is a robot.

But the real breakthrough is an AI that can see what you see... and become master to an apprentice.

The potential applications go far beyond repairing light switches.

Cooking. Car maintenance. Gardening. Learning musical instruments. Operating unfamiliar equipment.

Training factory workers. Helping inexperienced technicians perform more complicated jobs.

You might even have an AI that patiently teaches you how to knap flint.

Twenty thousand plus years later, we've come full circle...

But now it's passed from silicon to neuron... from sensor to eye.

It's hard to overstate this breakthrough...

We've created another way for an exponential set of learned patterns to travel from machine to human.

But there's a catch.

And it's a big one.

Knowledge Isn't Wisdom

There's a reason electricians typically spend four or five years (8,000 to 10,000 hours) learning their trade.

You can memorize the electrical code.

Have a great mentor (not all are).

Study.

And still get zapped right out of the apprentice pool.

But imagine what happens when AI can guide millions of people through thousands of tasks they previously wouldn't have attempted.

The economic implications are enormous.

For example...

Harvard's Joint Center for Housing Studies estimated homeowners would spend $518 billion on home improvements and repairs in 2026 alone.

What happens if AI makes even a small percentage of those projects easier for you?

That's potentially billions of dollars shifting between contractors, retailers, and manufacturers.

And consider the companies that charge people to learn something.

Why pay hundreds of dollars for instruction when an AI companion can watch you practice, answer your questions, and correct your mistakes?

But there's an opportunity here, too.

Businesses that help skilled workers complete more jobs, faster, could become considerably more valuable.

So...

AI could destroy billions in one industry while creating billions in another.

And three publicly traded companies are already showing us how this is playing out.

To be clear, these three are examples I'm watching for you, not official Moonshot Minute recommendations. Do your own research before you buy any of them.

CHEGG (CHGG): AI Destroys $390 Million in Annual Sales

Chegg built a business around helping students answer questions, complete assignments, and understand difficult subjects.

Then ChatGPT arrived. Suddenly, students could ask questions directly to a general-purpose AI.

They didn't necessarily need to pay for a specialized homework service.

The consequences have been brutal.

I've seen this first-hand.

My son showed me how he loaded all of his course material into ChatGPT...

And had ChatGPT quiz him ahead of tests.

He's not alone because in August, Chegg reported second-quarter 2026 revenue of $51.8 million.

Down 51% from the previous year.

And this isn't just a bad quarter. Chegg's annual revenue fell from $766.9 million in 2022 to $376.9 million in 2025.

That's $390 million in annual sales gone.

Not all of that decline can be blamed on AI.

But the company has specifically pointed to the loss of search traffic as Google began answering more questions using AI.

Chegg is now working to reinvent itself around workforce skills, and its smaller skilling business grew 2% last quarter.

But the broader lesson is difficult to ignore.

When your business depends on selling access to answers, AI can wipe you out.

And the same threat extends to countless businesses selling tutorials, technical instructions, or basic professional information.

Chegg stock was a great short.

It's down 99% from its high in 2021.

Now consider a very different company and how it's using AI...

LOWE'S (LOW): When AI Sells More Products

Lowe's sells lumber, tools, appliances, plumbing supplies, and practically everything else homeowners need to fix or improve their houses.

Including Romex.

DD told you copper was going to be precious.

Lowe's apparently took him literally.

My local Lowe's also locks up its Romex behind bars!

It's not the only one with copper jail, too...

Whether the thefts prevented outweigh the sales lost while customers hunt down someone with a key is a question for another day.

But credit where it's due.

DD called copper for you a long time ago... and Lowe's called the locksmith.

They also have something Chegg doesn't.

A physical product to sell after answering your question.

Mylow

Working with OpenAI, Lowe's developed Mylow, an AI assistant that helps customers plan projects and choose products.

There's also Mylow Companion, which helps store associates answer unfamiliar questions.

The early results are striking.

According to Lowe's, customers using Mylow convert into purchasers at 3X the rate of other online shoppers.

That doesn't prove AI caused the increase (I'll listen to its next earnings call for you to find out)...

But look at the size of this market it could disrupt.

In 2023, homeowners spent $340.5 billion on professionally installed improvements, versus $64.5 billion on DIY projects.

Just 1% of that professional spending equals $3.4 billion in revenue potentially up for grabs.

This scenario suggests something interesting.

When someone doesn't know how to fix a problem, uncertainty can block a purchase.

Give that person useful guidance, and suddenly the project is possible.

The customer buys the tools, the replacement parts, the supplies... and perhaps a 12-foot Halloween skeleton.

Chegg can lose business when AI provides the answer.

Lowe's can gain business when AI provides the answer.

Same technology. Completely different outcome.

And our third stock may have an even more defensible position...

SERVICETITAN (TTAN): When You Have the Data AI Needs

ServiceTitan provides business software for plumbers, electricians, HVAC technicians, and other skilled trades.

Think of everything that happens during a service call.

A customer calls, an appointment gets scheduled and a technician is dispatched.

The technician inspects the equipment, diagnoses the problem, recommends a repair, and prepares an estimate.

Parts may be ordered. At some point the job gets completed.

An invoice goes out. Payment arrives.

ServiceTitan connects these activities inside a single software platform...

And that creates something valuable that few companies have.

Data about work being performed in the real world.

Its software can store photographs, videos, equipment information, customer histories, estimates, and job records.

This gives ServiceTitan a lot of proprietary data.

It also has an AI assistant called Atlas, which can help technicians find equipment documentation, identify parts, and troubleshoot problems.

In September, ServiceTitan reported quarterly revenue of $292.8 million. Up 21% year over year.

Its net dollar retention exceeded 110%, meaning expansion among existing customers more than offset losses.

Those are encouraging signs of customer stickiness.

Plus, switching software platforms becomes difficult when years of customer records, pricing, dispatching, equipment histories, and billing are tied together.

Data Is the Moat

In the age of AI, proprietary data that you have the right to use may become a moat.

Imagine collecting thousands of service calls where technicians photographed defective equipment, diagnosed the problem, repaired it, and documented the outcome.

That could become an extraordinarily valuable dataset for training AI to understand practical work.

Having pictures is only the start.

The value is in connecting what the technician saw with what they did... and whether it worked.

Now, ServiceTitan isn't alone...

Nor does it automatically have the right to train general-purpose models on its customers' information.

Still, I think ServiceTitan should be collecting more opt-in repair videos, linking diagnoses to confirmed outcomes, and building vision tools that make technicians faster and more accurate.

It could even turn that into a new education product...

An AI master technician for onboarding and training.

At a hypothetical $50 a month, 100,000 technicians would bring in $60 million in annual subscription revenue.

ServiceTitan already offers recorded field conversations and coaching tools, so it wouldn't be starting from zero.

But there's another reason ServiceTitan could win.

Even if OpenAI, Google, or another company builds the world's best AI repair assistant, someone still needs to integrate good data into a contractor's business.

Scheduling. Equipment history. Part selection. Estimates. Invoicing. Payment. Photos. Videos. Outcomes.

That's where ServiceTitan already operates, but there are risks...

ServiceTitan still reports GAAP operating losses, and powerful AI agents could eventually make competing software easier to build and switch to...

But of our three examples, ServiceTitan may provide the most interesting combination of AI opportunity, specialized data, and switching costs...

So, I'm keeping my eye on developments for you.

The Next Twenty Thousand Years

Our old Stone Age dad was proud that something of him would remain in the hands of his son...

Because mentors give what they know... and rejoice in their students' success.

AI changes this.

Every technician who photographs a failed electric device, diagnoses it, fixes it, and logs the outcome...

Is now leaving something of himself behind, too.

Not in one pair of hands.

But in a massive dataset that can guide a million pairs of hands he'll never know.

Which brings us back to our three stocks.

Chegg found out what happens when AI provides free (or nearly free) answers.

The company loses value.

Lowe's found out what happens when the answer leads to a purchase. The customer converts.

The company becomes more efficient.

And ServiceTitan is sitting on something rarer than either... a record of what skilled people actually did, and whether it worked.

The company has what AI needs.

AI devices will usher in the biggest change in how humans learn since the first father handed his son a stone.

And it's only just starting.

Always be prospering,

Harry Seldon
Moonshot Minute

🔓 Premium Content Begins Here 🔒