AI with Kyle Daily Update 185

I produce around twenty pieces of content a day with one business partner and a Mac Mini.

It’s (deep breath)…

Livestream. Newsletter. Five to eight short-form videos. Carousels. LinkedIn post. DM giveaway PDF. YouTube long-form. Reel covers. Thumbnails. And more.

All whilst running a business. And just two of us. (Hi Harms!) Two years ago, this would have needed a five-person media team. It now needs about one hour of my actual human time and systems that runs around it.

Audience is now sat at about ~300,000 across TikTok, Instagram, YouTube, newsletter. The system has driven over half a million in product revenue in the last year alone.

That's the topic today. How the hell to do all of this.

Why You Need This At All

Why become a one-person media company?

Audience is leverage. Full stop. For sales, for trust, for reach, for partnerships, for everything that comes after.

If you have something to sell - yourself, a product, a service, a workshop, a course, an idea, a charity, a candidate, anything - it is vastly easier to sell it to 100,000 people who already trust you than to find 100,000 new people to sell to. And trust takes time to build. You need to ideally be doing this BEFORE you have anything to sell!

Two years ago I had basically nothing. Today the audience is closing in on 300k and the business has done £500k+ in product sales off the back of it. Not from cold ads. From people who'd been watching me yap and chat for free, every day, on the internet.

Now, this is not about being an influencer. Not about dancing on TikTok. Not about being beautiful and going to five-star hotels (sweet gig if you can get it, not the gig I'm describing!).

The version of content that works for entrepreneurs is teaching.

You HELP people.

You teach what you know. You share what you're learning. You give opinions. People find you. They trust you before they ever speak to you. By the time they're in your DMs, they've watched 40 hours of you. They know your voice. They know your taste. They're warm.

Most People Never Start

The reason most people don't do this is the same reason most people don't go to the gym. Every guru tells you you need to write a newsletter, shoot eight videos a day, write LinkedIn posts, do livestreams, build an email list, run paid ads, post carousels, comment on other people's posts, run a podcast, blah blah blah.

Total overwhelm. It feels like a full-time job on top of your actual job. Because it is. Content creation can easily be full time work for a team of people.

So people freeze. They never start. Or they post twice, get no likes, and quit. Entirely understandable.

The reason I've been able to do this without becoming a full-time content guy is because I built a system that takes one hour of actual human input and turns it into twenty pieces. AI does the heavy lifting around the edges. The middle bit - the bit with my face and my opinions and my voice - stays human. This is key.

One Person + One Session + System = Content Machine

Let’s look at this high level.

Every morning I get sent a brief by my AI. “Here are some topics you could cover. Some trending tweets. Some viral posts”. I read through them and decide (over coffee) what I want to cover.

I do one livestream in the morning. Roughly 25-45 minutes. Camera on, slides up, talking. That's the human input. Everything else is built on top of this very human creation.

The livestream gets transcribed. The transcript goes into a content-processing agent (Claude Cowork on my end, Hermes also fine, OpenClaw also fine, doesn't really matter). That agent has skill files for every output. Newsletter. Scripts. Carousels. Giveaway PDF. Reel covers. Thumbnails. It produces a dashboard for me with all the assets.

I run some human checks and passes. Tidy up the newsletter. Shoot some short form videos.

And I’m done.

One hour of human. Twenty plus pieces of content.

With AI doing all the heavy lifting in the middle.

The Full Loop

Alright this is going to initially look overwhelming. But I’ll walk you through it. Ready?

If your face melted like that guy in Indiana Jones I don’t blame you.

We’ll break it down.

Eight stages. Half agent. Half human. Don't gloss over which is which.

1. Topic selection (agent). My agent scans X, Smol AI News, Raindrop saves, and a few other feeds every morning. Drops a briefing on my phone before I'm even up. Mix of evergreen and timely. Timely gets clicks. Evergreen builds catalogue. You want both.

   

2. Prep (agent + human decision). Agent drafts an outline. I record a voice note with my hot take, my angle, my opinions. Agent revises. Agent generates the slide deck (currently using ChatGPT Images 2.0 - was NotebookLM six months ago, will be something else by Christmas). The step stays. The tool changes.

   

3. The Live (human). This is your job. Get on camera. Give opinions. Answer questions. The AI is not doing this. The AI cannot bring the personality. The AI did not have the breakfast conversation that prompted the new angle. This is the bit you cannot outsource.

   

4. Post-live (human). Send the raw video to my editor Harms via WhatsApp. Drop the file into Descript. (I do this manually because it takes 5 seconds and the API isn't worth wiring up. Don't over-engine

   er.)

   

5. Transcript (agent). Out comes the text. Goes into the brain.

   

6. Content processing (agent). Skill files run in sequence. Newsletter draft. Scripts. Carousel. Giveaway PDF outline. Reel covers. Thumbnails. Dashboard.

   

7. Long form + short form streams. (agent+human) Long form: newsletter, LinkedIn header, carousel, giveaway PDF. Short form: 12 scripts (I manually choose and shoot 5-8 of them), reel covers, descriptions, hashtags all by AI. human-edit the newsletter.

   

8. Schedule and feedback (agent). Schedule across platforms. Once a week the agent compares the drafts it produced against the final versions I published. Learns what I cut, what I rewrote, what I left in. Drafts get sharper. The loop tightens.

Then we go again. Daily. Forever!

The Three Places To Keep The Human In

I’m not a fan of 100% AI content.

It has that ChatGPT stank. And a lot of it (not all!) is slop.

I understand the appeal. Build a system. Let it rip. Infinite content. Infinite marketing. Infinite money.

Yeah but no.

Lazy people will build the machine and then turn it on full automatic and wonder why the engagement craters.

Because humans are smart. And don’t want to read your slop!

So what I do is make sure that humanity is used in the right places.

For me personally it’s three core places:

1. The voice note at prep. Before any output is generated, you record your raw, unedited, ramble-y, often contradictory voice note about the topic. That voice note carries your tone, your taste, your opinions, your angle. The agent uses it as fuel for everything downstream. If you skip this, everything else sounds like AI. For me this becomes the basis of the slide deck and livestream I do. But it starts from a voice note.

2. The livestream. You get on camera. Or, if you genuinely cannot do that yet, you have the AI interview you in voice mode. The AI asks. You answer in your real voice. Either way, the human is talking and being recorded for the system to use.

3. The human edit on the newsletter and the recording of the short videos. Don't ship AI-written newsletters. Don't ship AI-voiced videos. Write. Record. Stay in the frame. Do the work.

For your content you need to decide where you add the human. We’ll all be slightly different. The main thing is not to automate 100% just because you can.

The MVP - Start Here

The above is a lot.

Here's the MVP, minimum viability product. Do this. Build the rest later.

Pick one topic. Just one. Today.

Open Claude or ChatGPT in voice mode. Say: "Interview me on the topic of X. Ask me one question at a time. Don't generate anything until I've answered six questions."

Answer in your normal voice. Plain English (or you native tongue!). Ramble. Don't edit.

At the end, ask it to write: (a) a newsletter draft/blog post, (b) five short-form video scripts.

That's it. No need for a full livestream. No editor. No fancy AI agents. No HTML dashboards. Just ChatGPT or Claude on your phone and you yapping.

That’ll spit out drafts for you - don’t just publish them. Instead edit the newsletter/blog post. Record the video scripts on your phone. Post them. Done. Nothing fancy.

Build the rest as you grow. Build agents. Add skill files for repetitive tasks. Add the livestream when you're ready. Add the dashboard when you outgrow the chat window. Add the carousels when you're sick of being just text-and-video. But do all of these step by step not on day 1.

Don't do what I do. Do what I started doing. Tiny first. Then layer in.

Kyle