Explaining AI to My Mom

The Text

My mom texted me this week, worried about artificial intelligence. She had watched one of those videos: alarming music, red arrows, a narrator saying the word "tokens" like it was a diagnosis. My mom is Brazilian. When she worries, she worries in Portuguese, and her message ended with one question. "Devo me preocupar?" Should I be worried?

Here is the detail that makes this interesting. She teaches English for a living. If any job on earth is supposed to be doomed by a machine that speaks every language, it is hers. Hold that thought, because we are coming back to it.

This is what I told her, written down for anyone's mom.

The Prediction Machine

The machine in her video has read more than any person who ever lived. Books, manuals, court records, recipes, a billion arguments between strangers. And out of all that reading, it learned to do exactly one thing: guess what comes next.

That is the whole trick. Guess the next word, then the next, then the next. When the guessing gets good enough, it can write a letter or build working software. It is still guessing. It is just very good at it now.

It is not a mind. It wants nothing. It does not know my mom exists.

And the tokens?

A token is a piece of a word. Companies pay for AI by the piece, the way you pay for electricity by the kilowatt hour.

The video that scared her was about big corporations getting surprised by their own electric bill. That is the confession hiding under all the scary music. Not that the machines are taking over. That the accountants fell behind on the meter readings.

The Backhoe

I was a contractor in my former life, floors and remodels around metro Atlanta, so let me explain this the way job sites explained it to me.

Within living memory, a machine rolled onto American job sites and did in an afternoon what a hundred men with shovels did in a month. The men with shovels were not wrong to be afraid. Those jobs really did disappear, and it really did hurt the families that depended on them.

Then the other half of the story happened. Foundations too expensive to dig got dug. Highways and sewer lines and dams that would have taken decades got finished in years. Whole trades appeared out of nowhere: equipment operators, hydraulic mechanics, heavy haulers, the dealer down the road who keeps the machines running.

The backhoe did not end work. It ended one kind of work and paid for ten new kinds.

Every big tool tells the same story twice. When banks installed ATMs, everyone knew the tellers were finished. Teller jobs grew for thirty more years, because branches got cheap to open and banks opened more of them. The spreadsheet emptied whole floors of ledger clerks and created the financial analyst. Both predictions, both times, wrong: nobody starved, and nobody stopped working.

The Part I Will Not Sugarcoat

Some jobs are disappearing right now, and the people inside them are not statistics. A person mid-career, watching a program eat their daily tasks one by one, is living through something real. They deserve better than a pep talk about backhoes.

But look closely at what actually gets eaten: tasks, not purposes. The machine took the digging. It never took the judgment about where to dig, or what Georgia red clay does after three days of rain. The clerk who understood the ledgers became the analyst.

The value moves up. It does not vanish.

Ninety Days at My Desk

Now the part I can prove, because every piece of it is dated and saved where anyone technical can go count.

In roughly the last three months, working with AI every day, I created about thirty five software projects holding more than three thousand commits. A commit is a saved, documented change. A brick laid in the wall, with my name and the date on it.

Some of what got built:

A learning platform for our family's business, which helps Brazilian immigrants learn English and prepare for the U.S. citizenship interview. Student accounts, video lessons, progress tracking. It replaces software we used to rent.

An operations hub that pulls our sales, support tickets, course data, and marketing numbers into one place, with dashboards the team actually reads.

A private health dashboard for our family, so lab results live in one organized place instead of a drawer.

This website.

A small program that reads our household budget every morning and writes us a plain-English briefing before breakfast.

The tools have unfamiliar names doing familiar jobs. I build with Claude Code, an AI that writes software while I direct it. Things run on Supabase and Firebase, which are databases for rent, and on Vercel and Railway, which are hosting for rent. A tool called n8n moves information between apps like plumbing. And around all of it, the company still runs on things you would recognize: Slack, Notion, Zoom, spreadsheets, a calendar. The AI did not replace any of that. It is the crew that finally wired it together.

I priced work for a living, so I will price this honestly. Handed to a US software agency, that list is somewhere between half a million and a million dollars, and two or three years of waiting. The truer answer is that most of it would never have been built at all. One former flooring contractor, no engineering team, about ninety days.

I am not writing about the backhoe from the sidewalk. I am the guy in the cab.

The New Bottleneck

Building was supposed to be the hard part. It is not anymore. The hard part now is getting things into human hands: the student who trusts you enough to enroll, the reader who understands what the tool does for them, the customer who tells a friend.

Everything AI makes is for a person at the end of the line. So the scarce things now are the human things. Attention. Trust. Taste. The ability to explain. The machines made those matter more, not less.

That is also why I push every business owner I know to start now. A small team using these tools will outbuild a big team that refuses, the same way a crew with equipment outbuilds a crew with shovels. Waiting is a decision too. It is just a decision someone else profits from.

Neither Doom nor Utopia

The doom channels and the utopia channels are the same business. Strong feelings keep you watching, and watching sells ads. "Things will change gradually and mostly work out" has never once gone viral.

The quiet truth is that automation did not just arrive. It has been in your kitchen for fifty years. The washing machine, the ATM, the bill that pays itself, the phone that spells for you and reads the map out loud. You absorbed every one of those waves and stayed entirely yourself. This wave is faster, and the speed deserves respect. But it is an acceleration on a road we were already driving, not a cliff.

What I Told Her

So, the English teacher.

My mom teaches Brazilians to speak English, and on paper no job looks more finished. The phone in her students' pockets already translates everything instantly. Her students keep signing up anyway, because they are not trying to translate sentences. They are trying to become Americans. There is a room in a government building where you sit across from an officer and answer questions about your new country, in English, alone.

No machine can take the citizenship interview for you.

Her job was never delivering English sentences. It was walking a nervous adult across the border between two languages. That part is as safe as work gets. And some of what I shipped this spring is for her school. The machine she is afraid of spent the spring building her students a better classroom.

So no, I told her, there is nothing at her kitchen table to be afraid of. She will never need to understand tokens, the same way she never needed to understand kilowatts to iron a shirt. If she gets curious, she can ask the thing for a recipe, or have it help write a birthday note, and she will see it for what it is: a very capable helper with no plans of its own.

The fear always sounds wise and the curiosity always looks naive, and it has been the other way around every single time.

The shovel did not disappear when the backhoe came. I still own several. Nobody digs a foundation with one anymore, and nobody misses it.

Não, mãe. Not about this.