Calvin's Updates

The Latchkey Club Daily Draft — 2026-07-13

Teleprompter / Blog Script

I watched somebody demonstrate an AI tool recently, and about three minutes in I realized I had no idea why I would want it.

It was impressive. Things were moving around the screen. There were charts, agents, windows opening, words appearing very quickly.

But I kept thinking: What problem are we solving?

Welcome back to the channel, guys.

Today I wanted to talk about AI and older workers, because I saw several stories this week about how AI is changing careers for people over 50.

Some of the stories were hopeful. Some sounded like a polite warning that we had better catch up. And some were about people deciding they would rather retire than deal with another major technology change at work.

I understand that feeling.

But I think a lot of the advice starts in the wrong place.

If you are over 50 and trying to learn AI, do not start with AI.

Start with the job you already know.

So, let’s get into it.

The way AI is usually presented makes it feel like we all need to go back to school and become beginners again.

Learn the vocabulary. Learn prompting. Learn the newest model. Learn agents. Learn automation. Watch a forty-minute tutorial recorded six weeks ago that is already describing a product that no longer has the same buttons.

That can make a capable person feel behind before they have even started.

And after thirty or forty years of work, being told to become a beginner again is not always inspiring. Sometimes it just makes you want to close the laptop and see whether the retirement calculator has become any friendlier since lunch.

But maybe we are looking at it backward.

A younger person may understand the tool faster.

An experienced person may understand the problem better.

And the problem is where the value starts.

I have spent years around systems, customers, operations, schedules, data, and all the strange little gaps where one department knows something that another system cannot see.

That means when I look at a process, I usually do not begin with, “What can AI do?”

I begin with, “Why does this keep breaking?”

Why are we copying the same information twice?

Why does the customer have to understand an interface designed for an engineer?

Why is the answer in two different systems that do not talk to each other?

Why does the same question come back every week?

Why is somebody spending an hour assembling something a computer could prepare in two minutes?

Those questions came from the work, not from the technology.

That is the advantage I think people our age may be underestimating.

We have a backlog of real problems.

We know the irritating task that everybody accepts because it has always been irritating.

We know which report nobody trusts.

We know where the instructions stop matching reality.

We know the workaround that lives in one employee’s head.

We know which “simple” request will create three phone calls next Thursday if it is handled badly today.

That knowledge is not obsolete because a new tool showed up.

It is exactly what makes the tool useful.

My own path into AI did not begin with some grand plan to reinvent myself.

I was working on a Python script. I got stuck. I asked an AI how to do one part, copied the answer, and kept going.

Then I realized it could help with larger pieces.

Later, coding agents became capable of building much more than a snippet. They could help build the front end, the back end, the plumbing in between.

But the important change was not that the machine suddenly knew my job.

It did not.

The change was that I could describe a problem I already understood and get to a working version much faster.

That is a different thing.

I have used that approach for practical systems: sorting information, tracking repeated tasks, simplifying complicated tools, helping organize family logistics, and researching opportunities that would otherwise require a lot of manual searching.

None of those ideas came from asking, “What cool thing can AI do today?”

They came from noticing something that was taking too much time, creating confusion, or depending on somebody remembering one more thing.

That is where I would tell somebody our age to begin.

Pick one problem you know well.

Not your whole career.

Not “How do I automate my life?”

One problem.

Maybe you prepare the same weekly update from five emails and a spreadsheet.

Maybe customers ask the same ten questions, but the answers are scattered across old documents.

Maybe you have notes from years of projects and no easy way to find the lesson you need.

Maybe you are mentoring somebody and keep explaining the same decision process.

Maybe the family paperwork lives in twelve places and every important appointment begins with a scavenger hunt.

Start there.

Then ask AI to help with a piece of it.

Summarize the source material.

Draft the checklist.

Compare the documents.

Turn the voice memo into instructions.

Find the missing questions.

Build a rough tool that makes the complicated interface simpler.

And then do the part experience is good at.

Inspect it.

Does this match reality?

What did it miss?

Would a normal person understand it?

What happens in the unusual case?

Is the source trustworthy?

Does this save time, or does it merely move the work into checking a machine that keeps making confident mistakes?

Would I put my name on this?

That last question matters.

There is a lot of pressure right now to use AI so we look current. I think that is a bad reason to use almost anything.

You do not need an AI project that impresses a room.

You need one that fixes something you understand.

Because when you know the work, you can tell the difference between a demo and a solution.

A demo works once while somebody is sharing a screen.

A solution survives Tuesday afternoon when the data is messy, the person is distracted, the password does not work, and somebody has entered the date in the wrong format.

People with experience know about Tuesday afternoon.

We have lived there.

That does not mean age automatically makes us good at AI.

We can be stubborn. We can protect bad processes just because we helped create them. We can call something wisdom when it is really discomfort wearing reading glasses.

So yes, we still have to learn.

We still have to experiment. We still have to accept that some methods we are proud of may no longer be the best method.

But learning the tool does not require erasing what we already know.

The better posture may be curiosity in one hand and experience in the other.

Let the tool surprise you.

But make it answer to the real problem.

Let it move fast.

But do not confuse speed with being finished.

Let a younger coworker show you a better feature.

And then show them the condition that only appears after the system has been running for six months.

That is not a competition.

That is how useful work gets done.

I also think this matters for people who are close to retirement and wondering whether it is worth learning one more technology.

You do not need to become an AI expert to get value from AI.

You may only need enough skill to reduce one burden, document one process, teach one person, or make one part of the job less dependent on you.

That could make the remaining years at work better.

It could make your experience easier to pass on.

It could even open a smaller, more flexible kind of work later because you can now produce something that once required a team or a lot more energy.

But begin with what is already in your hands.

The process you understand.

The mistake you recognize early.

The customer you know how to help.

The question you have answered a hundred times.

The problem everybody else has stopped seeing.

Bring that to the tool.

That is your starting advantage.

Anyway, that is what I have been thinking about.

If AI at work feels like one more thing you are supposed to catch up on, do not try to learn the whole future this weekend.

Find one recurring problem you already know better than the software does.

See if the tool can help you fix that.

I would be curious what problem you would choose. Leave me a note in the comments.

Thanks for listening.

Video Prompt Script — Questions to Answer Without Reading

Use these as prompts. Don’t read them on camera; answer them naturally.

1. Cold open: What was impressive about the AI demonstration you watched—and why did you still wonder what problem it solved?

   • Follow-up: Have you ever seen technology that looked busier than it looked useful?

2. The pressure to catch up: What does most “learn AI” advice make an experienced worker feel they have to master?

   • Follow-up: Why can that make a capable person feel behind before starting?

3. The reversal: What does “don’t start with AI; start with the job you already know” mean?

   • Follow-up: How is knowing the tool different from understanding the problem?

4. Your own path: How did getting help with one Python problem grow into using agents for larger, practical systems?

   • Follow-up: What stayed yours even when the tool became more capable?

5. The hidden 55+ advantage: What recurring failures, workarounds, customer confusion, and unusual cases does an experienced worker notice?

   • Follow-up: Why is that backlog of real problems valuable now?

6. The first project: What is one appropriately small problem somebody could bring to AI?

   • Follow-up: Weekly updates, scattered documents, repeated questions, mentoring notes, or family paperwork?

7. The judgment test: What questions should an experienced person ask after AI produces an answer?

   • Follow-up: Does it match reality, survive the odd case, help a normal person, save time, and deserve your name?

8. Late-career application: Why might learning just enough AI still matter for somebody a few years from retirement?

   • Follow-up: Could it reduce a burden, improve a handoff, or support smaller and more flexible work later?

9. Closing: What one problem do you already understand better than the software?

   • Follow-up: Invite viewers to name the first thing they would try to fix.

Title Options

1. Start With the Job You Already Know
2. Over 50 and Learning AI? Start Here
3. Your Experience Is the AI Advantage

Thumbnail / Onscreen Text Options

• DON’T START WITH AI
• YOU KNOW THE PROBLEM
• OVER 50? START HERE

Shorts / Reels Cutdowns

• “Don’t start with AI.” Cut from the pressure to learn every feature through the line, “Start with the job you already know.”
• “People with experience know Tuesday afternoon.” Use the contrast between a screen-share demo and a solution that survives messy data, broken passwords, and real people.
• “Would I put my name on this?” Cut the inspection question cluster as a practical framework for supervising AI-generated work.

Viewer Question

What recurring problem in your work or home life do you understand well enough to make it your first practical AI project?