Big Idea

The argument in one line.

The highest-leverage use of AI for YouTube is not script-writing — it is automating the research loop that surfaces outlier video formats before your competitors spot them manually.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

You run or manage YouTube channels for clients and want to see how an agency scales research and scripting with Claude.
You are a solo creator who currently does YouTube research manually and wants a replicable browser-automation approach.
You are building an AI-assisted editorial workflow and need a reference for how to chain research → brief → interview → script → thumbnail.
You want to see Claude's computer-use / browser-control capability applied to a concrete business use case.

SKIP IF…

You are looking for a deep technical implementation guide — the prompt and skill files are withheld behind a paid inner circle.
You already have a mature research automation system; the demo here is entry-level compared to what the presenter runs internally.
You expect production-quality AI editing advice — the editing section is brief and mostly tool name-drops.

TL;DR

The full version, fast.

The presenter runs 50+ YouTube channels and uses Claude's computer-use capability to automate outlier-video research: Claude navigates YouTube in a live browser, scores videos against the channel's average view count, and populates a Google Sheet with title formats and outlier scores — replacing a full-time researcher. Scripts are generated through a two-step process: first an AI interview extracts the creator's expertise conversationally, then a custom Claude skill writes the script in the creator's voice using that context. Thumbnails are prompted in Claude and rendered in Magnific/Nano Banana. Editing uses Gling for silence removal, and Claude for visual planning. The prompts and skills behind the live demos are gated behind a paid program.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →

Chapters

Where the time goes.

00:00 – 02:00

01 · Cold open & credentials

Fable 5 benchmark claim, presenter establishes authority via client results (Glencoco $1M in 6 months, Sam Plera $400K/month), promises end-to-end AI YouTube tutorial.

02:00 – 11:40

02 · AI-powered YouTube research

The core demo: a Claude skill + prompt that navigates YouTube in a live browser, scores videos by outlier ratio against channel averages, and populates a Google Sheet with title formats and outlier rankings. Presenter shows the Glencoco case study to justify why research — not production — is the primary leverage point.

11:40 – 20:00

03 · AI script briefs & client interview system

A Kanban-triggered workflow (via Victor AI + Slack) auto-generates script briefs. Each brief includes reference video transcript analysis, Signal vs Noise breakdown, TAM analysis, hook draft, and an AI interview that extracts the client's expertise conversationally. WhisperFlow enables voice input. Output feeds a custom Claude script-writing skill.

20:00 – 23:00

04 · Full script output & results

Shows a complete AI-generated script document: reference analysis, TAM fit check, hook, body, CTA placement. Demonstrates result on web design channel (245K views, multiple $10-20K client closes).

23:00 – 25:00

05 · Thumbnail creation with AI

Two use cases: (1) generating impossible-to-photograph shots using a Claude prompt fed into Magnific/Nano Banana; (2) building a self-updating thumbnail swipe-file skill via a cron job that checks a Google Sheet for new entries daily.

25:00 – 29:00

06 · AI in editing & visual planning

Gling for silence removal, AI for visual planning from transcript, brand guidelines generation with Claude, Remotion for animation, Artlist/EpidemicSounds for music. Honest caveat: sound design AI is not ready yet.

29:00 – 31:00

07 · Monetization: funnels & brand kits

Claude Design (claude.ai/design) for brand kit generation. Claude Code builds complete sales funnels from brand kit + offer context. Live demo of two complete funnels generated entirely by AI.

31:00 – 33:00

08 · Misc uses & CTA

AI comment-section responder, digital product creation, content repurposing. Closes with inner circle pitch.

Atomic Insights

Lines worth screenshotting.

Outlier video ideas almost never come from within your niche — they come from spotting a working format in a completely different niche and transplanting it.
Claude's computer-use capability can navigate YouTube, score every video on a channel against its own average, and log outliers to a spreadsheet without any human interaction.
An AI interview — not a blank script template — is the fastest way to extract high-quality expertise from a busy client; they answer questions, they don't write.
The most common scripting failure is writing for the wrong total addressable market; Claude's brief process includes an explicit TAM analysis step before writing a word.
You can run multiple Claude browser instances simultaneously, each researching a different niche and writing to the same database — effectively a team of YouTube researchers running 24/7.
Claude's 'Signal vs Noise' analysis of a reference video transcript tells you which elements of a successful video are actually causal vs coincidental before you try to copy them.
Thumbnail image generation with Magnific lets you create shots of a client that are physically impossible to photograph — at quality indistinguishable from real photos.
A custom Claude skill trained on your own voice, scripts, and course material will produce a first draft closer to publishable than any generic prompt — and improves with each feedback loop.
Setting a cron job on Claude to self-update a thumbnail swipe-file skill daily is a concrete example of a self-improving AI system with no ongoing manual maintenance.
Gling (silence removal) plus manual Premiere cleanup is the current state-of-the-art for AI editing; fully autonomous editing is not yet viable for long-form.
Claude Code can build a complete sales funnel — headline, copy, design, layout — from a brand kit plus offer context, with no designer required.
The expectation set in a YouTube title must be explicitly verified against the hook and body before publishing; failing this audit is the primary cause of poor retention on technically well-made videos.

Takeaway

Research is the product; everything else is execution.

WHAT TO LEARN

Automating the discovery of outlier video formats — not polishing scripts or thumbnails — is where the sustainable edge lives on YouTube.

01Cold open & credentials

Client results ($1M in 6 months from 12 videos) are used as proof of methodology, not luck — the implication is that the system is replicable, not the person.

02AI-powered YouTube research

Cross-niche research is the source of every high-leverage video idea — the best formats in your niche almost always came from somewhere else first.
Claude's computer-use capability can navigate YouTube, compute outlier scores against a channel's own average, and populate a structured database without any human interaction.
Running multiple Claude browser instances simultaneously — each researching a different niche, writing to one shared database — is a scalable replacement for a full-time research team.

03AI script briefs & client interview system

An AI interview extracts creator expertise more effectively than a blank script template because answering questions is cognitively easier than authoring from scratch.
A script brief that includes Signal vs Noise analysis and TAM fit-check before writing a single line produces better first drafts and fewer revision cycles.

04Full script output & results

Custom Claude skills trained on your own voice, reference scripts, and audience context compound over time — each feedback loop narrows the gap between first draft and publishable quality.

05Thumbnail creation with AI

Thumbnail generation via Claude prompt + Magnific enables shots that are physically impossible to photograph, removing a common creative constraint on concept quality.
A self-updating thumbnail swipe-file skill — triggered by a daily cron job that checks a Google Sheet — is a low-maintenance system that improves without ongoing manual curation.

06AI in editing & visual planning

Silence removal tools (Gling, Riverside) are a genuine time-saver as a first editorial pass, but full AI editing for long-form is not yet viable; editors still own the fine cut.

07Monetization: funnels & brand kits

Claude Code can assemble a complete sales funnel from a brand kit and offer context — removing the design bottleneck that delays most creators from launching new offers.

Glossary

Terms worth knowing.

Outlier score: A ratio comparing a video's view count to the channel's median view count. A video that scored 10x the channel median would have a high outlier score, signaling that its format or topic resonated beyond the average.
Claude skill: A persistent memory or instruction set saved inside Claude that carries forward context — like a voice map or scripting rules — across separate conversations, so the model doesn't have to be briefed from scratch each session.
Computer use / browser-use: Claude's ability to control a real web browser — clicking, scrolling, reading page content — as if it were a human operator, enabling it to do research on live websites rather than working from static data.
TAM (Total Addressable Market) analysis: In the video's scripting context, an audit of whether a video's hook and title will resonate with the full audience that might click — not just a narrow sub-segment — to maximize initial reach and retention.
Signal vs Noise: A step in the presenter's brief workflow where Claude categorizes elements of a reference video's transcript: 'Signal' = elements that actually drive the video's success; 'Noise' = elements that are present but don't matter and shouldn't be copied.
Magnific / Nano Banana: An AI image generation platform (formerly Freepik) used to render photorealistic thumbnail images from Claude-generated prompts, capable of producing shots that would be impractical to photograph.
Gling: An AI tool that automatically removes silences, bad takes, and filler from raw footage before an editor does the fine cut — a time-saving first pass, not a replacement for editorial judgment.
Remotion: A code-based video rendering library that Claude can use to animate still images programmatically, enabling AI-generated motion graphics without a dedicated motion designer.
WhisperFlow: A browser extension that enables voice-to-text input directly into web apps, used here to speak answers into Claude's interview interface rather than typing.

Resources

Things they pointed at.

05:36toolClaude (computer-use / browser automation) ↗

11:34toolVictor AI (Kanban brief generation via Slack)

13:40toolWhisperFlow (voice-to-text browser extension)

23:00toolMagnific / Freepik (AI image generation) ↗

25:44toolGling (AI silence removal for video) ↗

25:44toolRiverside (AI editing alternative) ↗

27:05toolRemotion (programmatic video rendering) ↗

27:17toolArtlist (music licensing) ↗

27:17toolEpidemicSounds (music licensing) ↗

29:06toolClaude Design / claude.ai/design (brand kit builder) ↗

29:06toolClaude Code (funnel / web app builder) ↗

18:20toolVidIQ (YouTube analytics — mentioned as insufficient) ↗

18:20toolViewStats (YouTube analytics — mentioned as insufficient) ↗

Quotables

Lines you could clip.

03:16

“The video ideas don't just fall from the sky. They come from seeing enough videos that are working across the platform so that we can tell what is working on YouTube right now and then bringing that over to our niche when nobody else has done it before.”

Concise thesis on cross-niche research — standalone and quotable→ IG reel cold open↗ Tweet quote

13:40

“It's getting the information for me. It's effortless because I'm the expert and all I have to do is answer its questions as if I'm on a podcast or getting interviewed.”

Captures the insight that AI interview beats blank-page scripting — universally applicable→ TikTok hook↗ Tweet quote

05:36

“Imagine that running on your computer twenty four seven. And if you take it to the extreme level like we're doing, imagine having multiple computers running that twenty four seven.”

Vivid scale image for AI automation — good contrast hook→ newsletter pull-quote↗ Tweet quote

17:44

“It analyzes the total addressable market of the video and makes sure the hook actually fits that total addressable market and captures and resonates with all of them so that we maximize how many viewers that we resonate with.”

TAM framing applied to YouTube hooks — rarely heard, actionable→ newsletter pull-quote↗ Tweet quote

The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

metaphoranalogystory

00:00Claude Fable five just dropped, and this AI model is on a whole other level from anything else we've ever seen before. This is not a normal new AI model release where it improves slightly. I mean, just look at the benchmarks.

00:11It's absolutely crazy. It's way faster, way more powerful, and it's especially effective when it comes to long complex thinking tasks. Now for YouTube, this is absolutely game changing if you know how to use it right.

00:23And I'm saying this as somebody who runs over 50 YouTube channels actively right now for brands and businesses, many of which we've grown from zero to over a $100,000 a month, like Glencoco, who went from 0 to $1,000,000 in the first six months from only 12 videos. The first video got over 280,000 views and made them over a 100,000 in the first thirty days, and they grew from zero to 10,000 subscribers in the first sixty days in the sales niche.

00:43Sam Plera, who went from 0 to $400,000 per month from YouTube. So we've had to learn how to use AI at the cutting edge level and stay ahead of the curve because in order for us to have an efficient operation at that scale that produces results like that, we need to know the best ways to use AI for YouTube. And I'm telling you that Claude Fable five has just took that to another level.

01:01So in this video, I'm gonna be showing you exactly how to take advantage of Claude Fable five to help grow your YouTube channel and be ahead of the curve when it comes to using AI for YouTube. I'm going to be covering how to use AI at a cutting edge level for every single part of the YouTube process from research strategy and ideation to scripting to editing to thumbnails, even some other stuff such as monetizing your channel effectively and making that super easy so you can make money from YouTube as a business or as a creator.

01:25Starting with research strategy and ideation because this is the most important part of growing your YouTube channel, the idea is everything. You can have the best editing in the world. You can have the nicest design thumbnails, but if the idea sucks, the video is not gonna perform well.

01:36To highlight the importance of this, I've got this channel here, which we took from 0 to a million dollars collected from YouTube in the first six months. No external traffic, no paid ads, no nothing. And it went from zero to 10,000 subscribers in the first sixty days in the sales niche, which is a very niche niche.

01:48If we go to their channel, you can see where we stepped in with them. This video here was the first video that we made. Beforehand, they just had a bunch of their own business asset videos, which they embedded on their website, but they weren't really succeeding on YouTube as you can tell.

02:01But this first video, 290,000 views, and this was the one that generated over a $100,000 in deals in the first thirty days of it going live. And then back to back after that, these videos also got over a 100,000 views.

02:09This one getting 200,000 views. And as I said, zero to 10 k subscribers in the first sixty days, 0 to $1,000,000 collected from YouTube in the first six months. Now the way we came up with this first video idea was because in a completely different niche, we saw this video of speedrunning an entire dropshipping store was working very well.

02:25Speedrunning entire Shopify store, breakfast, I show everything 239,000 views. And the only reason we saw that was because we were doing a lot of YouTube research, and this is why YouTube research is so important because without this, would have never been able to come up with that video idea.

02:39The video ideas don't just fall from the sky. They come from seeing enough videos that are working across the platform so that we can tell what is working on YouTube right now and then bringing that over to our niche when nobody else has done it before. And so that's how we came up with that video idea.

02:53And then for this one, same story. I made 500,000 cold calls and learned this. We did this video because we saw a video, which was I made 300 YouTube videos and learned this.

03:00That video got 480,000 views, was the best performing video for this channel. And so then we took it to this niche for this client and did I send 10,000,000 cool emails and learn this, and that got a 175,000 views.

03:09And then after that, we took it to Glen Coco's channel, which got over a 100,000 views. Last but not least, watch me book 10 sales calls in ten hours. We did that one because we did this video for our client instantly, which was watch me start and sell a service in ten hours, 440,000 views.

03:22You can see me there auto playing in that video. This video made multi 6 figures for them. People called it the best advertisement they'd ever seen in the comments.

03:28You can check the comments. This guy then copied it, got 360,000 views, but we did this video because we saw this one here from this channel, Dia, watch me make and sell a digital product in ten hours.

03:37So the pattern that you see there is that we never would have came up with those video ideas without having the visibility over what was working on YouTube across all these different niches. Some of these niches we don't even have channels in, and that came from research. Now the problem with this used to be that you had to do manual research in order to find these high performing videos and figure out what's working on YouTube or use a tool like one of 10 or ViewStats, but those platforms don't give you the full visibility.

04:01They only give you a fraction of the picture because they're not able to show you everything that's emerging and working right now in the niche the moment that it is. And so oftentimes, you're not able to spot these outlier videos and outlier formats early without doing that manual research. But now because of AI, we're able to automate that research process with the right prompt and with the right skill, and I'm gonna get onto that in a second.

04:23But you can see here what AI is producing by actually taking control of your browser and doing the research process as if it were you or as if it was a full time YouTube researcher working full time for you. You can see here it's filling out this Google Sheet with the video link, the views, the outlier score, the upload date, and it's even started to try and predict what the title format of the video is so that it can build this separate sheet within the sheet where it has all of these title formats stored within it, where it's basically grouped videos that are of a similar title type and that have similarities, distinct similarities between the titles that are also all outliers in order to find what are the patterns that are driving success in these videos.

05:08And then we also have a scoring criteria that we trained it on in order to be able to effectively order these so that we can sort them in order of strength. And we have our own definitions for this. I'm not gonna get too deeply into how we do that.

05:21But the point is that it's finding these outlier videos that are working extremely well across different niches, then it's also finding the patterns between them, and it's building out this database so that I don't even have to do YouTube research. And this is how we've been able to scale YouTube research at the scale that we're at.

05:36Now I can't give you the full custom AI YouTube system that we use because we keep that exclusive for our inner circle members and our clients. But what I can show you is what it can do even with a basic prompt and a Claude skill. I'm gonna give it this prompt here, which I'll leave in the description along with any other prompts and assets that I use in this video.

05:52Use the YouTube research skill, so I'll also include the skill, to research a YouTube niche live in my browser, and then I'm telling it to navigate visibly, so on and so forth. You can read this. The niche that we're gonna do, let's do YouTube growth.

06:04Right? And let's see what it comes up with. I'm gonna let this play out, and this may take a while.

06:08So you can see here that it's running on the side, and if we go to Chrome, like, I'm gonna literally sit like this and not touch a single thing. So it's basically going through these seed search terms relating to the niche.

06:21So based on the niche, it's generating core seed keywords and search terms to explore. It's an actually scraping the website data like the JSON, the JavaScript data to pull what is on the page so it can most efficiently grab the channels, the videos, so on and so forth.

06:38And it's just navigating YouTube. So like it's going to the channel here, it's found vidIQ has videos that have performed outstanding well. It's scrolling the channel.

06:47I'm literally not even touching the computer, and it's literally scraping all the videos, their views, when they were posted. It's calculating the average view count of the video and the median view counts, the mean and the median, so that we're actually able to pull an accurate average score. And then from there, it's finding the videos that have performed outstandingly well to a statistically significant level where it's actually performed really well compared to their average video, but also their decent performers.

07:14We want the ones that have performed really well, then it's gonna log those in a database as well. So you can see it's just going here. Since Fable five released, this web browsing ability is faster, and it's also able to handle the more long, complex tasks.

07:27So the whole idea is you can leave this running on your computer device, or you can build a form of computer devices where there are separate instances of this running all at the same time in different niches, each being like your own mini YouTube researcher, and it's building all in the same database and it's finding all these formats.

07:46It's finding these from the initial scrape that it did. So when it searched the search term, it scraped all the data on that and then it found videos that were above a certain view threshold indicating that those channels knew how to make good YouTube videos that would perform well. It then built the list of those channels, and now it's going through the list of channels that it built from that initial search, exploring them in their entirety, finding all the best performing videos on that channel.

08:12You can see some of the ones popping up there like that one with 1,000,000 views, for example, and it's building this database of it. As you run this over a long period of time and you have this running in the background and you're getting thousands of data points across different niches, you can imagine the sort of patterns that it's going to find between these videos.

08:29A lot of these videos aren't gonna have patterns, and they're just gonna be one off outlier videos. But a lot of these videos are gonna have patterns between them, like what I showed you here in this view where we have all these title formats where you have every paradox in eight minutes, every martial art type explained in twelve minutes.

08:45It's not previewing the thumbnail there, but this one here, every single fallout of vault explained in minutes. And there's many, many examples of that. Same with the give me minutes.

08:53Give me ten minutes, I'll delete your fat, I think this was. Give me nine minutes, and I'll give you ten years. Give me ten minutes, and I'll show you how to not something.

09:00Right? But it's finding these patterns between the outliers. And once you find those, you know what's working on YouTube because you can also see when these videos were posted as well, and you can figure out if this is something that's working right now, if this is a new and emerging format, or maybe you just see a one off outlier that's really interesting that you think can you can take to your niche.

09:19So this is where you start to get the ideas because you know what's working on the platform. And now because of AI, if you're leveraging it properly, like we are at a cutting edge level, then you're able to automate that process of finding out what's working. And so now what you can see is it's actually just opened up a new browser, which is the spreadsheet, and it's gonna start to fill out this spreadsheet.

09:38So I'm gonna let this play and probably speed ramp this up because when will it do its thing. I'll just switch to the clone tab, and I was just checking what it was doing, and then I've just switched back, and I can see it here. So it's gonna need to format this, by the way, so it looks pretty ugly right now, and that's totally fine.

09:54It's just because it hasn't formatted it. But it's gonna separate these into its separate fields. But the point is is that we can see the data working.

10:01This was one of the channels there. Look at this. It's just filled it all up there.

10:04And now if I go back to Cord, it's probably gonna say something soon about how it's filled it up. But those were the channels that we explored. VidIQ, Robert Benjamin, Make Money Map.

10:11Right? And if we go to some of these so this video was an outlier for VidIQ. It got over a 100,000 views, for example.

10:17In fact, it tells me the views here. So I can go at one here. Like, it found one of Paddy's, which was 2,200,000 views, and it found some of Make Money Mats, the future of YouTube do this before 2026.

10:28And if we actually go to his channel and we scroll down, we can see the future of YouTube. There you go. So this was a massive outlier.

10:35So you can see it's successfully found that outlier there and ignored all of the ones that aren't outliers. 292,000 views when all the other videos around it got significantly less views.

10:45Even this one, yes, it was an outlier, but it wasn't as much of an outlier as that one there. I've literally just ran that for a couple of minutes.

10:51Imagine that running on your computer twenty four seven. And if you take it to the extreme level like we're doing, imagine having multiple computers running that twenty four seven.

11:01And this is literally just the prompt that I gave Claude Kowicke. Bearing in mind, the system that we're running behind the scenes is on a whole other level. This was just to demo you a simple version that you can take in action.

11:10But the system that we're using for ourselves, our clients, the channels that we run, our inner circle members as well, which, by the way, we have, like, an inner circle where we help businesses and brands grow on YouTube, but also if you're somebody in the YouTube economy, like a YouTube strategist, then we're helping you scale and build these systems into your businesses so you can help your clients get better results.

11:28The systems that we're using behind the scenes for ourselves and also for our members and our clients is on a whole other level to this. Like, this is literally just a simple prompt as you saw I gave it with a Claude skill as well that we've built. So from this alone, this is one of the highest leverage implementations and use cases of AI specific to YouTube that you can start using right now because the video ideas are literally everything, but the video ideas come from the research.

11:52And you can see here it's gave a summary how it went, discovered the channels via three YouTube searches, settled on vidIQ, make money Matt, Robert Benjamin, so on and so forth. Like, this was just a little sample of what it can do because I wanted to demo this to you guys, but you would never stop this. Like, you would just let this continue to run.

12:07And as I said, if you're taking this to the biggest extreme, we'd have multiple instances running of this. So super powerful, and, yeah, if you wanna grab the prompt and the skill and all that stuff, I'll leave a link in the description to that.

12:17And if you want the full custom AI solution, join the inner circle or become a client of ours. So that is YouTube research and strategy, and now we're gonna move on to how we're using it for script writing. Now for script writing, I'm gonna show you the entire step by step system, even how it integrates with our project management system.

12:32And we have this Kanban view where it moves through the certain stages. And so what I can do here is I'm gonna go with this one right here and I'm just gonna say that I want this to be an interview to full script.

12:45And then using an AI tool called Victor, which I'll leave a link to in the description, and if you use that link, I think you get some free credits. This tool directly integrates with Slack, is commonly used for communication across teams and project management, and it is notifying me of a brief being written. So now you can see here, starting brief generation, client views, client's Jake Trinder, video title.

13:04It's even gave the record ID so that you can track it directly back to your project management system, and it says that it's working on it. Now instead of waiting around for this, I'm just gonna show you what the output of one of these briefs looks like. So for example, if I go to my client many charts Kanban view, we can scroll here and see that the AI auto populated this field, brief link, with the brief.

13:24And if we go to the brief, it's generated a full script brief based off of the skill that I have developed with Claude on how exactly I want these briefs written and what I want included. Everything from the title and the overall concept to also the video that we're taking inspiration from and breaking down the transcript of that video.

13:42It literally goes to YouTube, breaks down the transcript of it, figures out why that video has worked, reverse engineers it, and then says how we're going to apply it to our video. Then it gives an overall vision of the video. It gives the hook or suggestion of the hook.

13:54We would actually go in and copy chief this and adjust the hook at the final stage when it actually comes to recording the script. It gives suggestions for a mid roll placement to end CTA. So what video do we lead into at the end of the video because you wanna increase session time.

14:09You always wanna point and direct them to another one of your videos so that increases watch time, and then the YouTube algorithm pushes you more. So it does that. And then this is the really cool part, an interview process to extract the information from the client so that we can write a full script for them that isn't AI slop and that is actually hyper accurate to their voice and the value and experience that they would provide in that video.

14:32And we made this because we needed to solve this bottleneck where clients struggled to record videos because they weren't being given a full script and it took too much effort from them being a busy business owner and handling with their clients. But we didn't wanna write a full script for them. That was AI slop.

14:46And even though AI can write great scripts with minimal input, we're looking to make the best videos in the niche because all of our clients, the goal is to have them be the top channel in their niche. And so what we developed was this where it literally interviews the client. You can see, hey, Sarah.

15:00Good to have you here. Should take about fifteen to twenty minutes, and then it gives the concept of the video. Does that match how you teach this, or is there something about the framing or angle that you'd reshape before we go deeper?

15:10And then using something like Whisper Flow, which is an extension, I can leave a link to that in the description as well, I can toggle this and go, hey, Claude, I think that this is pretty spot on.

15:21The one thing that I would say is I wanna really focus on talking about Instagram story sequences because that's a really pivotal part of actually monetizing on Instagram. And then when I send that, I might actually do this with a different one because I don't want it to remove the current information that we did for this one.

15:37I'm gonna do it for this one here because this was mine. So yeah. So that sounds pretty spot on, but I wanna make sure that we absolutely nail the checklist.

15:44And so I'm assuming you're probably gonna ask me about that and ask me what would be part of my checklist. But apart from that, spot on. Straight away, responds.

15:52And it's this back and forth process that's super fast, and it's extracting all the information that it needs from me. So before we get into the items themselves, here's the hook. Okay.

16:01I like this. I need to update the social proof, though. I wouldn't say I run a YouTube agency.

16:04I would say I run over 50 YouTube channels, and I've taken these channels from zero to over a $100,000 a month, then I named dropped some specific examples of our clients like Sam, who went from 0 to 400 k a month from YouTube. Glencoco, that went from 0 to $1,000,000 in the first six months, so on and so forth. And I'm using WhisperFlow to do this, as I said, just to be super quick, I'm not gonna type it out.

16:23Correct. It's 50 plus channels now. That's updated.

16:26We're growing quite a bit as a business. It's actually 80 channels now, so it's not 50, but I'm using 50 as example. Now the checklist, if someone hands you a video idea, then what's the first thing you look at?

16:34The first thing I would look at is, is this actually proven to work across YouTube? You don't want it to be proven in your niche because that means it's already been done, but has it been done somewhere else on YouTube so that we know that it's something that resonates with the market right now? And that's the key part right now.

16:48So these are separate bullet points on the checklist of, you know, is it proven, but is it also proven right now? And it's great because when it gives you this prompt, sometimes you just go off on a tangent and you just give it so much information because it sets you off on a flow. When you sit there in front of a script and you're actually having to script a video, your mind goes blank when you've got a blank page in front of you.

17:08But here, it's getting the information for me. It's effortless because I'm the expert and all I have to do is answer its questions as if I'm on a podcast or getting interviewed. That's exactly what this is, but it's AI doing it.

17:18How do you actually check those two things? It's literally prompting me in the right direction of making this the best video because it's not just getting me to say what's on the checklist. It's how do you actually go ahead and check that, which might not have even been something that I would have thought of including in this video if I went and did it myself before going through this process.

17:34And so I would go through this loop. I would flesh it all out. And then from there, we have a massive context pack, which we can feed into the AI to then write the script.

17:43And then feeding it into that AI at that point as well, There's a custom Claude skill that we have to create the script in our exact voice in the exact way that we want it. And to give you an example of an actual output right at the end and the result of it using this entire process that I've just shown you, let me pull that up right now.

17:59So for example, this script here for this client, this is all AI, by the way. Yes. We checked it and we adjusted things, but the way this is formatted, the majority and bulk of everything that you see here was done with AI.

18:11So it's given the reference video here to check out. It's gave the analysis of the reference video and the structure, the flow. It's even determined what is signal from the video that we should consider.

18:22Signal being stuff that is actually something that you should take from that video and noise, stuff that you could if you weren't aware of, you could easily take that and apply it to your video, actually but it doesn't make a difference. It analyzes the total addressable market of the video and makes sure the hook actually fits that total addressable market and captures and resonates with all of them so that we maximize how many viewers that we resonate with and ultimately increase retention and how many people watch through the video, the hook, the body, so on and so forth.

18:49And it's format at all of this and it's even got this bit at the end where it's just done its own calculations to make sure that the expectation being set in the title has been met in the hook and the video to make sure it's a good video. Because if we set an expectation in the title and the thumbnail and then we don't deliver on it in the actual video, then it's not gonna perform well because the viewer will feel like it's been a bait and switch.

19:07They'll feel clickbaited, etcetera. Now let me show you the result of this.

19:10So if you go with this channel, we started with this channel here. This is the first video that we made for them. You can see the views that they were getting.

19:16200, 100, 400. First video, 50 k views. So you know that this process works.

19:21This channel is the fastest growing channel in the web design niche, and he signed a ton of high ticket clients. He charges, like, 20 k per website build in some cases. So you can see which videos we did and which videos we didn't.

19:32So this one here didn't do as well, but still massive outlier compared to usual. This one actually was the worst performer. Then this one, twenty thousand views.

19:39And then this one, two hundred and forty five thousand views using the exact process that I showed you. But the specific video that I showed the script for just now is this one here, brutally honest advice about web design in nine minutes, 200,000 views. And we knew we got deals from this video as well.

19:51Like, this video had a massive ROI. It probably cost him about 1 to $2,000 to produce this video with us, and he got multiple clients with it charging in the range of 10 to $20,000 for building a website.

20:03So very profitable from this video. So you can see the entire process from start to finish of idea to building the brief to getting the information from the client through an AI interview process to then taking that context and writing the full script with it with a custom Claude skill and then the result of it.

20:22That is how you can leverage AI. Now, obviously, I can't give you the custom skills that I'm using internally for, again, clients and my YouTube inner circle, but the way you can build the skill is super simple. All you have to do is go to Claude and be like, I wanna build a YouTube script writing skill.

20:34Here's a bunch of my favorite YouTube videos that have performed extremely well. I love the way that these YouTube videos flow. I love the sound of their voice.

20:41And then you wanna give it a bunch of context of you, your brand, your target market, your business, your course material if you have it, your other content from other platforms. You wanna feed it all this information of you so it can build a knowledge and voice map so that when it comes to writing a script, it can do it in your voice and with your actual expertise.

20:59And then once you've got a video idea, you wanna make sure you have the reference video that you're taking inspiration from if you are taking inspiration from a particular outlier video. And then you wanna get the transcript of that reference video and also paste that into Claude.

21:10You And wanna tell Claude, I wanna make a full script for this specific video idea. Here's the video we're taking inspiration from. Here's the transcript, and you also have my knowledge and voice map too.

21:20Please write this script. And then you can refine that script and give Claude your feedback and also tell it to remember that via a Claude skill for next time. And through this constant tight feedback loop on the first script, getting it to a point where it's perfect in your mind and ready to record, through going through that process, you're going to a Claude skill as well at the end of it that you can take forward to the next script, which means it's gonna nail it that time and get it very, very close, and you're just gonna have way less feedback to give it.

21:47And it's gonna build that skill really custom to how you wanna make your YouTube videos. It's already very intuitive with copywriting and knowing how we hook people in and so on and so forth, but you can refine it with that custom skill.

22:00Now after you've got the script complete, it's then time to record the video. Now there's not a ton of genius stuff that you can do with AI right now on this, but in terms of making sure your set up is good, you can take photos of yourself, take different angles of your room, give that all to Claude and tell it, please analyze my room, analyze my space, analyze my setup, and tell me what improvements to make.

22:18Tell me how to make best use of my space. You can also give it a screenshot from the actual recording so I could take a screenshot from this video and tell it what do I need to adjust and improve to make this shot better and make this video perform better and go through that feedback loop with it to where you can get yourself to a great point.

22:36But the next step where we use AI to have a massive advantage is in thumbnails. And the way we use it for thumbnails is if we have a really creative concept where we need a certain shot of a client that just isn't possible because it's a really difficult shot to get or it requires a really high quality or a certain back drop, so on and so forth.

22:51We use AI in order to do that at a hyper accurate level to where you literally can't tell it's AI. It looks exactly like the person. So what you do is you go to Claude and you use this prompt on screen, which I'll also leave in the link in the description with all of the other stuff.

23:04Basically, what it does is it gives you a fully exhausted prompt. You take that prompt and you bring it into a software called Magnific, which was previously called Freepik, and you use Nano Banana two is what we're currently using to generate this image. You just paste the prompt in because it's already generated it.

23:19In that prompt, describe the specific style of image or shot that you need, whatever you need, and it gives you that image back perfectly. I'll be able to pop up some examples right now of images we gave it and the thumbnails that it's actually allowed us to create as a result of that, which is crazy. It is directly contributors towards certain videos being massively successful, and I don't think those videos would have been as successful if we didn't have as banger of a thumbnail concept for them, and AI is what allowed us to do that.

23:46Now the other way you can use AI for thumbnails is by coming up with the concepts themselves. What does the thumbnail actually look like, and why is this certain thumbnail going to drive good results? Now we've obviously got our own custom system for this where it's trained off of all of our internal client data, and we know what works because we're running all these different channels and different niches, and we've got all this database and this research that we've used leveraging our custom AI research system, but also the internal research team that we have.

24:11And so we have this really dialed in system for being able to generate these thumbnail concepts for particular video ideas. But something that you can do really easily is by building a Google Sheet of just a list of a bunch of thumbnail inspiration.

24:24Basically, building a thumbnail inspiration swipe file of successful thumbnails and styles of thumbnails. And then you can feed that into Claude and tell it, grab the thumbnails of all of these, and then use this to build a Claude scale and understand what exactly makes a successful thumbnail.

24:37And then you can update that spreadsheet as time goes on, and the Claude skill will also update. You can actually set a cron job on Claude to check the spreadsheet for new thumbnail entries every day and self update and self improve the scale. And so you can imagine you can tie the research system that I showed earlier to this thumbnail system to as we get more and more research and find more and more successful thumbnails, the thumbnail assistant checks that spreadsheet, pulls the new thumbnails, figures out what's working about them, and then updates its skill.

25:09And so you basically have these AI employees, one being a researcher, one being a thumbnail strategist working together because you're able to actually automate the running and the instances of these different AI systems.

25:24Now after thumbnails, we have editing, which is a big part of the workflow that can be optimized with AI. And I'm just gonna rattle off a ton of different ways that you can use AI throughout this process because I've got a lot here.

25:34I have to make notes for this because there's so many ways that you can leverage AI in the editing process. Now the first part is using AI for cutting, basically removing the silences, bad takes, mistakes, repetitions, so on and so forth. Right now, we're using Gling.

25:45There are alternatives like Riverside does this. There's another one called TimeVault. You wanna figure out what works best for your specific types of videos, and you just wanna test a bunch of these different softwares because it can save you a ton of time rather than having to manually cut out all these gaps.

25:58And then you take it into something like Premiere and make sure that the cut is super clean. It doesn't do it 100% of the way there, but it can save you a hell of a lot of so definitely leverage that. Now another part where we use AI for the editing of videos is actually in the planning.

26:12Taking the transcript of the video and then planning the visuals that we want for that video from everything we want in the hook in terms of showing social proofs that viewers believe that we're an authority on the subject to visualizing certain things, the way through the video to visualizing things we're educating them on in order to break down complex concepts into simple visuals that overall improves the viewing experience and increases viewer satisfaction, increases retention, and therefore increases the performance of videos.

26:38Now also what you wanna do is you wanna have a set of brand guidelines that you can also generate with Claude and go back and forth with Claude on to build so that for your own channel or for your client's channels, it's going to actually be able to create and ideate these visuals whilst considering certain brand requirements and just certain decisions that it should make based off of your client's brand and niche.

27:01Now for animating these visuals, Claude can't do this directly right now, so it uses something called Remotion. And what I recommend doing for this is generating the still images first. So you've done the planning.

27:11You've got these ideated visuals, and then you generate the still shots before going into the actual animation cycle. Because then if you do that first and the visuals end up being terrible and wrong and not what you wanted, you've wasted a ton of credits and a ton of time. So generate the images first and then bring all that context into actually making the animations.

27:30Now for sound design and music, AI is not quite there right now. So as of right now, the time recording this, I wouldn't really use AI for that. For music, I just use something like Artlist or EpidemicSounds.

27:40Again, there's an iteration cycle that you have to go through here, a tight feedback loop with the AI to get it to the place that you want it. It's not gonna magically know the exact way that you want visuals in your video, and so it is gonna be an upfront time investment. But once you've nailed it, you can leverage this for every video going forward.

27:57And if you're running an agency with different clients or you're a YouTube strategist helping different clients, you can use that across all your clients, and it's gonna save you a ton of time and a ton of money as well because it means that the editors have to do less work and spend less time on the videos. So therefore, it costs less per video.

28:12You're able to have higher profit margins from the channels or from your clients, and you're actually able to do more and do it faster. And that's why it's worth the time figuring out how to use AI for editing. It's not gonna replace editors, it's gonna give editors a ton of leverage, which then gives the benefits that I've just described.

28:26And also for the editors as well, if you're an editor, it allows you to take on way more projects and make a lot more money too. So there's also that learn AI for editing. Now all of that's how you use AI to actually help grow your YouTube channel and have videos that are successful, that go viral, that get views from your target markets.

28:40But then you've got the part of actually converting them into paying customers, clients, and making money from the platform.

28:47Not monetizing from AdSense, by the way. I'm talking about monetizing off platform through your products and services. Like, if you sell a high ticket recurring service or if you have a low ticket software subscription or a low ticket community, whatever it may be.

28:58And these are some of the actual best use cases or the use cases that have been most useful for me, for example, beyond the ones that I've just gone through. Because, for example, if I go to one of my funnels, this is one of my funnels, and this was entirely generated by AI. The headline, the way everything is laid out, the design of it, these reviews, the copy throughout, literally everything.

29:18It put it all together. And now if I wanna update this, all I have to do is tell AI, I wanna update this thing. And then I leave it and it gets it done.

29:25Now if you wanna do this yourself and build your own funnel where you can use it for people booking calls for your service or lead magnets, whatever, go to claud.ai/design, and this is gonna take you to a recent implementation from Claude, which is the ability to create your own brand kit. And so you can see that I've got my own design system here, which I generated with Claude, which keeps things super on brand for my colors, fonts, so on and so forth.

29:52So you can see there's like this design system that it's designed here. I mean, this is just amazing. Look at all of this.

29:57And I just went back and forth with Claude to build this. And now anytime I wanna build something like a funnel, the script briefs that I showed earlier, an application, literally anything.

30:06It stays super consistent with that branding. And the way that you do this is you can just go to design systems and then create a design system and then go through the process. You drop it your assets, you talk with it, you refine it, and you get it to something that you like.

30:18And then from there, you can just tell Claude, here's everything about my offer. Here's all the context you need to know. Here's my target markets, everything.

30:26Here's my design system as well, which you can download, and then you can give it to Claude Code. You would use Claude Code for this, and it will just build you the funnel. And this is exactly what it built me.

30:36And just to show another example as well, this funnel here, it built this in its entirety too. This is a completely separate funnel. I had a different sort of feel that I wanted with this one.

30:44It's added the images of me as well, so I gave it that. It just makes it so easy to make a super effective funnel, which just reduces a lot of friction for launching new offers, launching lead magnets, etcetera. In fact, you go to the description and you grab the prompts or whatever I went through in this video, that will be on a website using ClorDesign, so you'll be able to see it in practice with this video.

31:05And on top of this, there's a bunch of miscellaneous ways that you can use AI creatively, such as, for example, you can build an AI that automatically responds to your comment section using your knowledge map and voice map to respond to those comments accurately as if it were you, which then helps cultivate a community. You can use AI to help you build digital products or build a new offer so that you can monetize your audience better.

31:26You can use AI to scrape your own YouTube channel and all of the transcripts and use that to repurpose content across other platforms. There's so much that you can do with this, but you have to get creative with the ways that you use it. But some of the use cases that I gave you in this video, can implement straight away, and this will have a massive impact on growing your channel.

31:41If you want access to all the AI systems that I'm actually using behind the scenes with my client channels and with my inner circle members, then you can join the inner circle. I'll leave a link to it in the description. You can check it out, and if it makes sense, we'll let you inside.

31:54To be clear, I don't just teach how to use AI at a cutting edge level for YouTube. It's also about how do you actually do YouTube research yourself in the first place so you know how to do it so that you know how to guide the AI properly and customize it to your specific situation. Learning YouTube strategy and ideations that you can come up with the best video ideas in your niche.

32:11Script writing as well, making the best videos through actually writing really good scripts that are retentive and ultimately provide the viewer with the best experience and the best value. Editing the videos with AI if you're editing the videos yourself or actually how to source an editor.

32:24Same for thumbnail design using AI for thumbnail design, but also how to source a designer and actually build that YouTube team. Literally everything end to end on YouTube, we're going through in this inner circle. You have direct one on one access to me in that group, and also we have group calls every other day.

32:36So it's super high touch point to make sure that you're actually growing your channel and monetizing YouTube effectively, same way my clients have that have grown from 0 to hundreds of thousands of dollars per month from YouTube. So check the link in the description if you're interested in that. But if not, watch this video here that I'll pop up on the screen now because this goes through a lot of the core pillars of growing a YouTube channel in a lot more detail than I did in this video.

32:55This video was just specific to AI. You need to watch that one after this one, so go watch that now, and I'll see you in the next video.

The Hook

The bait, then the rug-pull.

The title is pure hype — 'changed YouTube forever' — but the 33 minutes underneath it are something more useful: a working agency's actual Claude stack, demonstrated live, warts and inner-circle-gating included. The presenter runs over 50 YouTube channels and opens with a benchmark drop before getting to the part that matters: a browser-use research loop that finds outlier video formats while the operator sleeps.

Frameworks

Named ideas worth stealing.

05:36model

Outlier Research Loop

Seed search terms from target niche
Scrape channels from search results above view threshold
Calculate mean + median view count per channel
Flag videos that significantly outperform channel median
Log to Google Sheet: link, views, outlier score, upload date, title format
Group by title format to find repeating patterns

Claude navigates YouTube via computer-use, scores each video as an outlier relative to its own channel, and builds a live database of working title formats and formats across niches.

Steal forAutomating YouTube research for any channel or client

11:40model

Script Brief System

Kanban trigger fires brief generation (via Victor AI)
Claude fetches and transcribes reference video from YouTube
Signal vs Noise analysis of reference transcript
TAM (total addressable market) analysis of hook
AI interview to extract client expertise
Context pack compiled → custom script-writing skill writes full draft
Feedback loop refines skill for future scripts

A multi-step pipeline from video idea to publishable script, with the client's expertise extracted via conversational AI interview rather than a blank-page writing session.

Steal forAny creator or agency that needs to script videos for non-writers

23:00model

Self-Improving Thumbnail Skill

Build Google Sheet of winning thumbnail examples
Feed into Claude to build initial thumbnail assistant skill
Set cron job: Claude checks sheet daily for new entries
Claude auto-updates skill with newly discovered patterns
Tie to research system so new outlier thumbnails feed in automatically

A thumbnail strategy AI that continuously improves itself by ingesting new high-performing thumbnails without human intervention.

Steal forAgencies or prolific creators who want thumbnail intelligence that compounds over time

CTA Breakdown

How they asked for the click.

VERBAL ASK

31:00product

“If you want access to all the AI systems that I'm actually using behind the scenes with my client channels and with my inner circle members, then you can join the inner circle.”

Soft close after the full tutorial — inner circle positioned as the unlock for the full custom skill/prompt stack withheld throughout the video. A second CTA points to a next video for organic retention.

MENTIONED ON CAMERA