Calvin's Updates

The Latchkey Club Daily Draft — September 16, 2026

Teleprompter / Blog Script

I listened back to one of my earlier videos, and nothing in it was exactly wrong.

The sentences were clean. The transitions worked. It had an opening, a middle, and a conclusion. If you handed it to somebody as a writing sample, they might say it was good.

But I could hear myself reading it, and I had this uncomfortable thought.

I don’t really talk like that.

Welcome back to the channel, guys. Today I wanted to talk about using AI to make things without letting the tool slowly sand off the parts that sound like us.

I use AI every day, so this is not going to be one of those speeches about throwing the computer away and going back to a legal pad. AI helps me research, organize ideas, build software, check work, and get past the blank page. This channel would probably be harder for me to maintain without some of that help.

But easier production and better work are not automatically the same thing.

When I started making these videos, I thought the hardest part was getting a complete script. So I would start with an idea, let AI help turn it into something polished, load it into the teleprompter, and read it.

That solved the blank page.

It also created a different problem. The script could be smoother than I was. It could have better transitions than I would normally use. It could arrange everything into three balanced points and give me a final line that sounded like it belonged on a coffee mug.

Then I would say it out loud and feel like I was borrowing somebody else’s shirt. It fit well enough, but it wasn’t mine.

The thing that helped was very simple. I recorded myself talking naturally about one of the topics. No attempt to make every sentence perfect. I just tried to explain what I was thinking.

When I compared that recording with the written draft, the difference was obvious. I used contractions. I shortened the neat lists. I repeated myself a little while trying to get the thought right. I added an ordinary story from home that wasn’t in the script. And when I reached the end, I didn’t deliver a large conclusion. I just said what I had been thinking and asked people what they thought.

That version was less polished.

It was also more honest.

This matters beyond videos. AI now helps write email, presentations, family messages, music, and almost anything one person communicates to another. If it can make the result clearer and faster, the question becomes which part we should let it do.

YouTube recently made its labels for realistic AI-generated or heavily altered content more visible. It is also using automatic detection for some of that material. That makes sense. If a video shows a real person doing something that never happened, viewers should know.

But YouTube does not generally require that kind of disclosure when AI helps with an outline, a title, an idea, or a script. Those uses are treated as production assistance.

That is probably a reasonable policy line, but it does not answer the personal question.

The platform is asking whether the viewer may be deceived by what appears on the screen. The creator still has to ask whether the process is changing the thought underneath it.

A recent story about Hank Green caught my attention for that reason. As I understand it, the issue was not that AI secretly wrote his scripts. He had been using it to help locate research and papers. That sounds responsible compared with asking a chatbot to invent the whole video.

But he said the speed and convenience had affected how he found his way into a subject. He was getting useful material quickly, but some of his own wandering, reading, and discovering had been crowded out.

I understand that because I can feel the same thing happen.

If I ask AI for the angle too early, the first organized answer becomes a set of rails. Now I’m improving its structure instead of finding my own. The tool may have saved me from wasting time, but sometimes the so-called wasted time is where the personal thought was hiding.

That is especially important for people our age.

At 57, my advantage is not that I can publish faster than a machine. I cannot. It is not that I know every new platform or editing trick. I do not.

What I have is a lot of specific life already lived. Work problems that looked simple and were not. Technology that arrived with big promises and then created new chores. Family responsibilities that changed how a decision had to be made. A body that occasionally gets a vote in the schedule. Mistakes I would prefer not to repeat.

AI can imitate the category. It can write about career change, retirement, parenting, back pain, or learning new technology.

It did not live my Tuesday.

And it did not live yours.

That may be the hidden advantage for older creators. We do not need to compete by producing the largest amount of information. There is already more information than anybody can use. We can bring the part that has a cost attached to it: judgment formed because something worked, failed, hurt, helped, or mattered to somebody we love.

A recent Irish survey found that audiences placed a high value on creator content being human-generated, and most respondents wanted disclosure when AI substantially created or changed something. That is one national sample, not a universal law.

But it lines up with what people seem to be saying all over the internet. They are not only asking whether an image is fake. They are asking whether anybody was actually behind the work.

Did a person notice something? Did they risk an opinion? Did they include the inconvenient detail that does not fit the neat argument? Or did they type a topic into a system and publish whatever came back wearing their name?

I do not think the answer is to hide the AI or pretend I do everything alone. I also do not think every spelling correction needs a warning label.

For me, the better boundary is this: AI can help me carry the thought, but it should not be the reason the thought exists.

I can use it to search for sources, but I need to open the sources and decide what they mean. I can use it to organize a draft, but the opening should begin with something I actually noticed. I can let it point out repetition, but not remove every repetition that sounds like a person working something out.

And before I read a script to anybody else, I should read it out loud to myself.

Would I actually say this? Do I know what this sentence means? Is there a real example behind it? If somebody asked me to put the screen down and explain the point, could I still do it?

If the answer is no, the draft may be finished, but I’m not.

Maybe that is the practice I need: speak first sometimes. Record the rough thought. Tell the story badly. Let AI help clean up what I meant after there is actually something of mine on the page.

The goal is not to preserve every awkward sentence. Editing helps. I just do not want efficiency to remove the person I was trying to express.

This channel is supposed to be a record of what I was thinking in this part of life. If the videos become smoother but less recognizable to the people who know me, that is not an improvement.

So I’m still going to use AI. I’m going to let it help with the research, the loose ends, and the structure. But the story, the judgment, and the final responsibility have to remain mine.

Anyway, I’m curious about this one. When you read, watch, or listen to something online now, what makes you feel like there is a real person behind it? 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: What did you hear when you listened back to an early, polished script?
   • Follow-up: How can something be correct and still not sound like you?
2. Your AI use: Where does AI genuinely help you research, organize, build, or get started?
   • Follow-up: Why is this not an argument for abandoning the tool?
3. The first warning sign: What did those early scripts do too neatly?
   • Follow-up: Which transitions, lists, or big closing lines felt borrowed when spoken?
4. The voice test: What changed when you recorded yourself speaking naturally and compared it with the draft?
   • Follow-up: Which contractions, repetitions, domestic details, and simpler ending sounded more honest?
5. The policy line: What does YouTube label, and which uses of AI count as ordinary production assistance?
   • Follow-up: Why can a platform rule address deception without answering a creator’s personal question?
6. The research shortcut: What did the Hank Green discussion reveal about getting useful information too quickly?
   • Follow-up: How can an early AI answer put a creator’s thinking on rails?
7. The 55-plus advantage: What can decades of specific work, family, health, and technology experience contribute that generic information cannot?
   • Follow-up: Why is lived judgment more valuable than trying to publish at machine speed?
8. A practical boundary: What does it mean to let AI carry a thought without becoming the reason the thought exists?
   • Follow-up: Which parts can it help with, and which parts must remain yours?
9. The read-aloud test: What questions should you ask before putting an AI-assisted script in front of a camera?
   • Follow-up: Could you explain the point with the screen turned off?
10. Closing: What signals tell you there is a real person behind something you watch or read?

Title Options

1. Don’t Let AI Sand Off Your Voice
2. My AI-Written Scripts Didn’t Sound Like Me
3. The Part of Creating I Won’t Hand to AI

Thumbnail / Onscreen Text Options

• THAT DOESN’T SOUND LIKE ME
• KEEP YOUR VOICE
• AI MADE IT TOO SMOOTH

Shorts / Reels Cutdowns

• “Borrowing somebody else’s shirt” — a polished script can fit well enough and still feel wrong when the words are not yours.
• “It did not live my Tuesday” — the 55-plus creator advantage is not speed; it is specific experience with real consequences attached.
• “Could you explain it with the screen off?” — a simple test for whether AI helped carry your thought or quietly replaced it.

Viewer Question

When you read, watch, or listen to something online now, what makes you feel like there is a real person behind it?