AI Built This Channel. I Just Show Up

00:00 - Opening

Hey, welcome to Slow Builds.

So I've mentioned a few times now that I was going to explain how this channel actually started.

And the honest version is... AI built most of it.

I know that sounds very dramatic, but it's pretty much mostly true.

AI structured it.
AI outlined it.
AI named it.

AI pretty much did everything.

I'm just the one sitting here talking as awkwardly and as cringe as I am.

I'm just going along for the ride to see where this takes me.

00:39 - It Started With My Kid

This didn't begin with me wanting to be on YouTube.

My kid was posting on TikTok and started getting some really good traction - a lot of followers, a lot of views.

They wanted to move into YouTube for longer-form videos and just see what that would take.

So I did what I always do.

I started researching.

I wanted to figure out the YouTube algorithm, the best way to go about creating a channel, how to keep retention, what makes channels have longevity, and what makes a channel actually last longer - not just from hype.

And I kept coming back to the same thing.

You can't just throw up one video and hope it works.

Sure, you might get one video to go viral.

But you need consistency.
You need history.
Common themes.
A recognizable voice and image that people trust.

Once people start watching, they come back because they know what to expect.

And this is what I told my kid.

02:02 - The 52 Video Plan

At some point during this journey of researching for my child, I asked AI something simple.

If someone was serious about doing a channel - or even from my kid's point of view - how many videos would it take to build enough history that someone would actually subscribe?

Not just click subscribe and never see another video again.

AI suggested thinking in cycles.

Not viral attempts.

A full year.

One video per week.

52 videos.

That made it feel less emotional and more like a plan.

So I asked it to map out a year.

ChatGPT took the history of our conversations and everything I had asked it before... and it produced 52 video ideas.

It mapped out 52 titles.

It organized them into themes, pillars, and bridges between topics.

Then it took the first few and gave script ideas.

So suddenly there was a roadmap and a starting point.

And that's when it started to feel like this might actually be something real.

04:01 - Recording the First Video

Eventually I thought... let's just record one.

After redefining scripts and trying to figure out how I was going to keep track of everything, I just said:

Let's take the first one and record it.

AI helped with lighting, positioning, and the structure of the video.

I filmed it.

It was very awkward.

So instead of scrapping it, I dropped the entire video and transcript into AI and asked it to analyze it.

It came back with notes.

Adjust the script.
Change the framing.
Reposition things.

So I did it all again.

Not that it made much difference.

You can go look at the first video.

Even this video, I feel like I'm getting a bit more comfortable - but it's still hard.

Right now there's someone standing nearby and it's kind of freaking me out.

You might have heard the cat meow too.

I'm not in a setup that makes me feel fully comfortable.

But even with that first awkward video, AI said something interesting:

Just post it.
It's real. It's authentic.

Over time you'll get more comfortable.

05:45 - Locking In The Channel

At that point I figured if I was going to keep going, I should lock it in.

AI gave me several name ideas.

I went searching for domains and YouTube handles.

The idea was Slow Builds, but the best available option was Slow Builds Lab.

As a developer, that phrase meant something specific.

Slow builds usually means compiling code - waiting for your code to finish building.

But AI reframed it.

Slow builds as in time building things.

Compounding.

Growth.

Analytical thinking.

Once I had the name and the URL, it became real.

We had:

• 52 video ideas
• scripts for the first several videos
• a website URL
• a YouTube handle

At that point, it was happening.

07:12 - The Website (Built By AI)

The funny part is... yes, there is a website.

AI told me where to put it.

Since I'm already a developer and use GitHub all the time, I simply told AI to build me a site.

And it did.

The site is slowbuildslab.com.

It created an about page.

It generated a logo using Canva based on the name and early scripts.

The logo uses gears and growth imagery, which I actually like a lot.

AI also built the banner for the channel.

Getting the banner aligned correctly across devices was difficult because of all the different screen sizes.

But eventually AI figured it out.

08:33 - The Video Workflow

The workflow now is pretty simple.

AI helps generate:

• script drafts
• title ideas
• descriptions
• first comment
• thumbnail ideas

There's a lot of back-and-forth while shaping the scripts.

Once the script is ready, I film the video.

Like I'm doing right now.

I barely edit.

Usually I just trim the beginning and the end.

Sometimes adjust volume or reduce background noise.

That's about it.

After recording, I export the audio and run it through Whisper to generate a full transcript.

Then I add that transcript into my Apple Notes.

AI formats it into sections with timestamps.

Then I check if the title, description, or comments need adjustments.

Usually they don't.

Then the video gets uploaded.

10:49 - Updating the Website

Once the video is live, I grab the URL.

My Apple note contains:

• the title
• the description
• the pinned comment
• the YouTube link
• the full transcript

Then I open Codex and tell it a new video went live.

Codex reads the note, updates the website code, commits to GitHub, pushes the changes, and deploys.

Within minutes the video appears on the website in the correct order.

That whole system runs without me touching anything.

11:28 - The Hardest Part

Honestly, the hardest part of the whole process is the thumbnail.

AI image generation often changes my face too much.

So instead I grab still frames from the video.

Remove the background.

Give AI the thumbnail concept.

Tell it not to modify my face too much.

And that becomes the thumbnail.

It's pretty straightforward.

12:14 - Why I Wanted To Do This

One of the reasons I wanted to try this experiment is something I'm noticing on YouTube.

Some of the fastest growing channels right now are not hyper-produced.

They're raw.

Someone opens their phone and starts talking.

They might be outside by a lake.

Or sitting in a room like this.

The videos are messy.

Unedited.

Just humans being humans.

Long-form conversation.

With so much fast, short-form content everywhere, people seem to want connection again.

It feels like you're sitting there with the person as they think through something.

And that authenticity matters.

14:07 - Where AI Fits

AI removes a lot of the friction.

It helps write scripts.

Suggest titles.

Generate ideas.

Plan a full year of videos.

Without that help, starting from scratch would feel overwhelming.

Instead, I know there's always another idea waiting.

And as I keep making videos, new ideas get added to the list.

So the pressure disappears.

AI also helped with the structure, the thumbnails, and the website.

But when I hit record...

It's just me.

15:31 - Finding My Voice

The videos I like most lately are the ones where I go off script.

I start with the structure.

But then I just talk.

AI formats the sections afterward so I can stay organized while recording.

That helps me keep track of where I am without losing the flow.

Over time I'm getting a bit more comfortable with that.

16:16 - The Real Bet

The bet here isn't that AI will make this successful.

The bet is simply that I said I would try it.

Maybe it goes nowhere.

But I want to see if authenticity compounds.

If I can show up once a week.

Not perfect.

Not polished.

Just consistent.

And see whether people are interested.

17:03 - The Experiment

These videos aren't perfect.

They're not highly produced.

But I want to see if consistency keeps you in the game.

As AI becomes more common and more content gets generated automatically, I want to see if something simple still works:

A real person.

Talking honestly.

Showing up every week.

And seeing where it goes.