Calvin's Updates

The Latchkey Club Daily Draft — August 11, 2026

Teleprompter / Blog Script

I was reading one of my own scripts into a teleprompter, and about halfway through I had a strange thought.

I agreed with everything on the screen. The grammar was fine. The point made sense.

But I did not sound like the person who wrote it.

Which was awkward, because technically I was the person whose name was going on it.

Welcome back to the channel, guys.

Today I wanted to talk about something I have been learning while trying to make these videos with AI.

AI can write the words. It can organize the argument. It can make a rough idea look finished very quickly.

But it cannot have lived the thing I am trying to talk about.

And if I let it do too much, that difference starts to show.

When I began making videos, I thought a good script was a script that sounded polished. So I would give AI a topic, let it build the structure, clean up the sentences, and then I would put the result on the teleprompter.

It looked complete and saved me from the blank page. But I was confusing ease with quality.

The scripts were not necessarily bad. They were just smooth in a way I am not smooth when I am actually thinking.

They had tidy transitions. Balanced lists. A strong sentence arriving exactly where a strong sentence was supposed to arrive. Then a conclusion that wrapped the whole thing up like the box was going back on the shelf.

I do not talk that way.

I use contractions. I pause. I back up because the first version was not quite what I meant. I remember some ordinary detail halfway through the thought, and sometimes that detail is the part that makes the thought worth hearing.

So I tried a simple experiment. I recorded myself talking about a subject without reading anything. No finished script. No carefully constructed opening. I just tried to explain what had been bothering me.

When I listened back, it was less polished, but it was more recognizable.

The pace was different. The words were shorter. I repeated myself a little. I did not sound like I had arrived to teach a lesson. I sounded like I was still working out what the lesson might be.

That is closer to what this channel is supposed to be.

I do not even know exactly what this channel is yet. It is partly a journal. It is partly a record of what it feels like to be 57 and think about work, family, health, retirement, technology, and how quickly the next season is approaching.

If the voice describing all of that could belong to anybody, then I may be removing the only part that actually belongs to me.

This is becoming a larger issue online. People use the phrase “AI slop” for mass-produced articles, videos, images, and posts that are cheap to make and almost impossible to remember.

YouTube’s own monetization guidance now uses the phrase “inauthentic content” for repetitive or mass-produced work. That does not mean YouTube banned AI. It did not. AI can be part of a real creative process.

The concern is sameness at scale.

You can feel it after a while. The pacing is familiar. The examples could have happened to anybody. The joke arrives because the format says a joke should arrive. The speaker claims certainty they probably did not earn. Everything is technically complete, but there is nobody home.

And I have made some of that.

I only noticed after comparing the generated version with the way I actually speak.

I think this is where people our age may have an advantage, if we do not throw it away.

At 55 or 60, we have a lot of specific material.

We have jobs that changed underneath us. We have made decisions that looked smart for six months and expensive for six years. We have raised children, helped parents, sat in waiting rooms, changed careers, lost friends, carried debt, fixed things badly, and occasionally fixed them well.

That does not automatically make us wise. Years can produce wisdom, but they can also produce habits and very confident opinions.

Still, we have details.

We know what a decision felt like before we knew how it turned out.

That is something an AI cannot generate from the outside. It can imitate the shape of reflection, but it cannot remember the room. It cannot know which part still makes me uncomfortable or which lesson took me ten years to admit.

That is the material.

The hidden advantage for an older creator is not that we can use AI to publish seven times a day.

It is that we have something the tool can help us uncover, question, and express.

So I am trying to change the order of the work.

Instead of asking AI, “Write me a script about getting older,” I need to begin with my own evidence.

What happened?

What did I think at the time?

What do I think now?

What detail would I tell a friend if there were no camera running?

Where am I still uncertain?

I can speak those answers into a voice memo. I can write them badly. I can give the AI the messy material and ask it to find the thread.

Then the tool is helping me hear what I am already trying to say.

It can point out where I skipped a step. It can ask for an example. It can challenge a factual claim, remove repetition that is not helping, and arrange the pieces so another person can follow them.

That is useful work.

But there are a few things I do not want it deciding for me.

I do not want it manufacturing a childhood memory because the opening needs more emotion.

I do not want it adding certainty where I am honestly unsure.

I do not want it turning a person in my life into content just because the story would perform well.

And I do not want it sanding away every rough edge until I sound like the same competent stranger everybody else hired.

My test now is pretty simple: would I actually say this sentence out loud if the teleprompter were off?

If the answer is no, I either change it or remove it.

A second test is whether the piece contains something only I could have noticed. Not a secret. Not a private family detail. Just a real observation from an actual life.

Maybe it is the four tabs open on the computer. Maybe it is the way my back changes a plan. Maybe it is realizing I started explaining a work process at chapter six because I had forgotten what a beginner does not know.

Those details are small, but they carry fingerprints.

And the last test is whether AI made me more honest or merely more impressive.

Did it help me get closer to the thing I meant?

Did it expose a weak assumption?

Did it give me enough structure to speak clearly without pretending I have solved the whole subject?

Or did it hand me a voice that sounds better than mine but does not belong to me?

I still plan to use AI for this channel. I would be pretending if I said otherwise. It helps me research, organize, compare drafts, check sources, and get past the part where a blank page becomes an excuse to make nothing.

But I want the finished video to contain more of the person who lived the question and less of the machine that arranged the answer.

Maybe that is the right relationship with the tool.

It can help with the sentence.

I still have to supply the life.

Anyway, that is what I have been thinking about.

Have you noticed AI changing the way you write or speak? Does it help you sound more like yourself, or does it slowly make everybody sound the same?

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. Opening: When did you first read an AI-assisted script and realize you agreed with it but did not sound like it?
   • Follow-up: What felt too smooth, arranged, or complete?
2. The experiment: What changed when you recorded yourself speaking naturally before writing the script?
   • Follow-up: Which contractions, pauses, restarts, or ordinary details sounded more like you?
3. The channel: Why is The Latchkey Club closer to a spoken journal than a place where you teach from a finished position?
   • Follow-up: What disappears if the voice could belong to anybody?
4. The wider problem: What do people mean by “AI slop” or synthetic sameness?
   • Follow-up: How can something be polished, complete, and still feel like nobody is home?
5. The 55+ advantage: What kinds of specific experience accumulate by your late fifties?
   • Follow-up: Why do years provide material without automatically making somebody wise?
6. Change the order: What happens when you begin with a voice memo, a real event, and unresolved questions before asking AI for structure?
   • Follow-up: How can AI uncover the thread without inventing the life?
7. Boundaries: What should AI never manufacture—memories, certainty, private family material, or a more impressive personality?
8. Three tests: Would you say the sentence with the teleprompter off? Is there an observation only you could have made? Did AI make you more honest or merely more impressive?
9. Close: What is the proper division of labor between the creator and the tool?
   • Follow-up: Has AI helped you sound more like yourself, or more like everybody else?

Title Options

1. AI Can Write the Words. It Can’t Have Lived Them.
2. I Let AI Write My Voice — Then I Heard What Was Missing
3. The Best Thing Older Creators Have Is the Part AI Can’t Copy

Thumbnail / Onscreen Text Options

• WHOSE VOICE IS THIS?
• AI DIDN’T LIVE IT
• POLISHED — BUT NOT ME

Shorts / Reels Cutdowns

• “I agreed with every word, but it wasn’t me” — use the teleprompter opening and the realization that a correct, polished script can still feel borrowed.
• “Older creators already have the material” — cut the section on decades of work, family, consequences, and the difference between having experience and automatically being wise.
• “Three tests for AI-assisted writing” — would you say it without the teleprompter, is there a detail only you noticed, and did AI make you more honest or merely more impressive?

Viewer Question

Has AI helped you sound more like yourself, or have you noticed it slowly making everybody sound the same?