His EXACT System For Writing Scripts With AI — Former Ali Abdaal Scriptwriter, George Blackman

You sort of learn all these little ways in which it may appear to be doing something well, and then the output that comes out, it actually reveals, oh, hang on, it wasn't doing these like 10 things that it claimed very confidently that it was.

I think this is a place that everyone is now in with pretty much any industry, whether it's to do with writing specifically, whether it's coding, whether it's like anything that could feasibly be affected by AI. there is this sense of uncertainty and like, where should I put my energy right now? Should I be putting my energy into getting really good at ideation because the script can just be written for me? Or is ideation already off the cards because AI can just do that for me? Do I even need to bother learning to code anymore because AI can just do that for me? And I think scriptwriting is such a broad church. There are so many aspects to it that even with that, it's like, okay, sure, AI could write my whole script, but should I still be the one looking at my hook? Because I know the hook's super important. Or are these key moments where I call out to a product that I'm trying to sell or something in the script, are they so important that I should still be using my brain, or is AI better than me? And so more broadly, we're going to talk about the sort of process that I use these days to write a script using AI. But I think the broader meta skill and the thing to try and figure out, and that I'm trying to figure out, and I think we're all trying to figure out, is where are the parts where it's the most useful? Where is it worth putting my energy into learning to use AI well? And where should I deliberately take a stand and be like, no, I'm using my brain here, not through some moral crusade, but because it's actually better to do so.

First step, and this is where the use case for AI is much, much stronger. I think even just to refer back to your previous question, it's like, where are people needed? It's almost easier to list the few places where AI is definitely a no-brainer. And the first step of my process these days is very much where it is a no-brainer, which is just explaining my thoughts about the script that I want to write verbally, using either like a voice note thing or just directly into Claude. Just that ability to spiel off in natural language what I think about a topic, even if it just allows me the chance to clarify my thoughts on something that I've never said out loud. Kind of the same idea as journaling or something. It allows you to just get more out than if you're just vaguely thinking about something in your head.

Yes, this is assuming you've already thought of the packaging, or I guess, yes, ideally the packaging. Not everyone does it. There's no shame if you don't come up with your packaging first, although obviously try to. But yes, the idea is already decided at this point. I guess we could talk about ideas as well, but that's less where I'm involved. This is, I know what I'm gonna talk about. Maybe I don't know exactly how I want to frame it. An example from my channel specifically was like, I wanted to do a video about my sort of journey in my business, but I knew I needed to tie it back to YouTube scriptwriting in some way.

And so I spoke out loud to an AI about this and I was like, here are the sort of points in my business when I was a scriptwriter, then I was like a consultant and now I run this community and blah, blah, blah, blah. But what's the sort of angle that makes this interesting for my audience who cares maybe a bit about business, but more about scriptwriting? And what it allowed me to do was get those thoughts out and then have some time going back and forth on the angle that we should approach the subject. And ultimately, we kind of landed on this approach of, at every stage of my business, how have my thoughts about scriptwriting changed? And like, what's the key lesson about scriptwriting I learned at each revenue level or whatever, which kind of, yeah, tied it back to my audience's interests.

So first thing, I currently use Claude. I used to use ChatGPT. I switched to Claude because I wasn't loving ChatGPT's outputs. Since I've switched to Claude, I think it seems like ChatGPT has had a kind of resurgence at the time of recording. I think it's on 5.5 now. Is it, or is it something else? Anyway, sidetrack. I use Claude.

Either one is fine. Yeah, exactly.

Yeah, exactly that. But in terms of giving the AI enough context, I think there's really like an easy route and a hard route. So the easy mode of this, I would say, is you need a style guide and you need an audience avatar, and arguably then some project instructions, like telling the project what it is for. That's really the simplest thing. How do you make those? Again, it's very simple. This is that meta skill again, where it's like you explain to it, this is what this project is. I need a style guide, I need an audience avatar. What do you need from me so you can create those for me? If that makes sense.

How many scripts of mine or examples of writing Or if you don't have any examples of writing, you would say, like, can I just speak to you for 5 minutes about a random topic and you get my voice from that? But those would be the 2 things that you can just ask your LLM for help making. So that's the easy mode. I think the advanced mode, shall we call it, and this is how I start my projects these days, is same thing, starting a new project with WhisperFlow or just anything that you can leave a voice note to and just explaining. This is the task that I have before me. So I did this when I was upgrading my course. I was like, I need to upgrade my course. My course is 36,000 words long. Since releasing it, I have uploaded 2 years' worth of newsletters.

I want to integrate those newsletters into my course. I have XYZ other concerns. 36,000 words is quite long. Will you be able to handle that? Just giving it all the context that I thought it might need, and then basically saying to it, like, what is my first step here? What do you need from me? And I think people underestimate this approach a lot because it kind of feels like, uh, I guess I need to know how to use this thing before I start using it. But the joy of it is, and why it feels so kind of weirdly addictive, is you can just say like, I don't know what I'm doing. Like, what do I, how do I, how do I proceed?

Yes. Yeah, 100%. And like recently building this app, I was suddenly aware, like, hang on, there are probably some legal ramifications of selling a subscription thing that relates to AI. And so I just said to it, like, what are the legal ramifications? And then this enormous block of text came out and it was like, we have these 25 tasks to do. You need to contact this lawyer, you need to blah, blah, blah, blah, blah. And suddenly it's like, oh yeah, okay. I mean, I didn't know how to do that, but now I do.

The next kind of thing that I like to have is something that trains it on what a YouTube script looks like in relation to other types of writing. You can totally build these yourself. You could even just take a bunch of transcripts from your own channel, from channels in your niche that are doing well, not to say steal how they do stuff, but structurally, what What do you detect about the structure of a YouTube script, and how can we use that to train you on what a script should look like? I have a free thing that has my version of what YouTube structure looks like.

That's kind of the next thing. And because I think then it saves you the hassle of having to be like, to say, here is my brain dump of thoughts, now structure it like a YouTuber would. It kind of has this preloaded context of what a YouTube script would look like. So that's really the next step, is once you've brain dumped your stuff, you want to either through the preloaded training or saying to it outright, let's first of all figure out what the overarching structure of this would look like. And it's going to come back to you with a suggestion and you see if it fits and go from there.

3 main blocks, shall we call them, that I would think of for a YouTube script. The hook is the obvious first block. It's quite distinct how you would go about writing a hook. There are different frameworks that different people have for this. For me, for educational hooks, I always think target, transformation, stakes. So that is, who is my target? Who's my audience? What is the transformation this video offers? And what is at stake if they don't achieve that transformation? The second main block is a segment. So these are your subtopics within the video. For every segment of my video, I think of it in terms of setup, tension, and payoff.

So at the start of each segment, we set up why it's important for the viewer to know this thing before we can move on to the next thing. The tension is where we teach the lesson, essentially. We kind of reveal the information. And then the payoff at the end of that segment is where we ultimately I suppose reveal the answer, but the understanding of that segment is now complete and we can move on to the next one. And then the third one is a call to action. So whether that is a call to action at the end to promote your next video or a product or service, or a mid-video call to action promoting whatever it might be, a free lead magnet or something like that.

You said it was target transformation stakes.

Target is always very easy for me 'cause I've got a 5-page audience avatar document that I've made in the way that we kind of described earlier. So I don't need to think much about that. I know who I'm targeting. It's YouTubers who are interested in scriptwriting, scriptwriters for hire who wanna learn to get better at scriptwriting so they can get more clients and blah, blah, blah. Transformation, if I take an example of one of my most recent videos and very apt for the conversation we're having now was how to set up a Claude project to turn Claude into a reliable scriptwriting partner, as opposed to a sort of irritating assistant who forgets everything. The transformation within that is like, you will go from not knowing how to do that to knowing how to do that. What's at stake is this idea, and it's something that I think a lot of my audience have felt, and it's something I certainly felt when I was getting on board with AI, is the sense that you will actually end up spending more time using AI to write your scripts than if you just use your brain. So determined are you that this is the new thing and, hey, I should be more productive with this, that you end up just getting frustrated with it.

It doesn't quite do what you want and you end up wasting more time. And so for me, that's what's at stake if they don't achieve the transformation, is you will remain in that state where this sort of perceived productivity of AI slows you down, results in worse scripts, means your channel doesn't grow kind of thing. So that's what I think before going into physically trying to write a script. And the joy of AI these days is that you can say, look, here's my target, here's my transformation, here's what's at stake. Here is a hook framework that has performed really well on my channel in the past. Please express these factors in terms of this structure. And it's pretty good at very finite structural tasks like that a lot of the time.

That's it, dude. And I'm out of a job. See you later. Yes. So the joy of all of this is that it— this allows me to go from blank page to V1 bullet point scripts very quickly, much more quickly than before.

Exactly that. 100%. That's, that's a really good point. Um, so not only does it save you that time in the moment, sat down. It may well be that you don't even have to be sat down for it to happen. But I think also it saves you so much mental strain having to restructure these thoughts on a page. That perfectionism creeps in. It feels like even if you're just typing notes, like, no, I should arrange these in the correct order first and blah, blah, blah.

All of that thinking power is saved by just being able to talk to it like a person. So we have our structure saying we've agreed with it. Yes, maybe with a bit of back and forth. I am now happy with the 6 segments that we have for this video about Jay becoming a dad for the second time. Now I want to move it forward into something more word for word. This is kind of a bit of an inflection point. Maybe that's the wrong word for it. It's a decision point on the tree as to whether you want to say, look, this is as much help as I needed.

Now I sincerely want to be able to just do the rest of this by hand, as it were, by brain. I don't want to then have an AI physically try and write me a full word-for-word script, because either that's not going to make me feel creatively fulfilled, or B, I just don't trust that it's going to do what I want it to do yet. So there is a point at which you might just decide, this is where I finish the AI thing, and that's completely fine.

Let's say at this point we have our structure. I'm going to assume at this point we've mentioned bullet point scripts a few times. It may be that you had your initial, like, here are the one sentence headline things for each of our segments. So like 6, literally 6 sentences. This is the skeleton. We're assuming we've now expanded it into, say, 2 or 3 bullet points per segment. And so really the choice here is if you are someone who really enjoys riffing, if your niche is the sort of niche that will allow you to sit down and riff, then you can very much test that. You can think, okay, I'm literally gonna go with what we have, sit down, start talking, see how it feels in the moment, see how much pain is deferred.

to the edit because of how much I may have rambled if I was just going off of bullet points. And then adapt and adjust each time you sit down. You think, ah, actually I felt completely out of control when I was sat down last time. This time I wanna go from 2 to 3 bullet points per segment to 5 to 6. I'm gonna turn that dial up. I think of it like a dial, just turning it up a little bit, testing it again next time, see how it feels. Did I ramble? Did that feel better? Was it now too prescriptive? Now I actually wanna dial it back to— 4 bullet points or something like that. So that's a really— that's, I suppose, a personal thing.

And that's if you choose the riffing side of things. If you now want to go fully word for word, and this is something I massively prefer, the amount of pain that is saved on editing, which is thankfully not even my job, but even just on filming day, the amount of false starts and thinking like, oh, did I screw this up? Wait, did I actually say that thing 5 minutes ago that sets up the thing I need to say now. For me, that's just not worth it. I much prefer to put an extra 30 minutes, 60 minutes into writing a full word-for-word script to save me that pain later.

Then the focus, and this is what I always focus on, is obviously the hook is very important, but the setups and the payoffs are the most important thing that I like to write word for word almost regardless. So your setup line at the start of each segment is like another hook, basically. It's the, the reminder of not only why I care about this overall topic, but why this particular subtopic is so essential in me moving forward. So if we take the example of my Claude project video, or yet, you know, how to set up a Claude project, the first step of that was literally you need to set up a project because a lot of people don't do that. But the way we set that up compellingly is not step 1, set up a project on your Claude chat. It's But before we get into the 5 documents you're going to need, we first have to fix one of the most common mistakes I still see dozens of my students making that turns AI into a useless assistant or something like that.

Yeah, exactly that. Exactly that. So focusing on those open loops, those setups at the start of each segment, and just reminding the viewer in a curiosity-inducing way why this next step is important. That's what, those are those moments in your retention graph where people will go. It's when you finish one subject and you don't immediately make it clear why the next subject is gonna be important to them specifically in the context of what they want and what the video title promised.

Can I do it on my app where I wrote it?

This actually is that very video that we were just talking about. This is how to set up a CLO project for scriptwriting. So you can see this, this is the software that I've built. And so I break it up into these setup, tension, and payoff things that I've spoken about a lot.

So in terms of the hook, have you ever spent hours fighting with Claude? This already, even though it's not explicitly saying, if you are a YouTuber who struggles with, but it's implicitly showing people, I understand you, this is the struggle that you have. So if you've ever spent hours fighting with Claude only to check the clock and realize this YouTube script took 2 times longer to write than if you just use your brain, then listen up. I'm using quite casual language in moments like this. With this video, we were experimenting with some like kind of comical intro B-roll. So this is the idea that it's like, I show that I understand you, I know who my target is. The problem with AI to help write scripts isn't that it's not good enough yet, it's because you're not giving it enough context. I guess this is kind of implying the stakes. It's a little bit tricky to always distill these into the exact 3 things, but if I was listing these things out, I would be thinking, when I list out my stakes, I'd be saying AI continuously takes longer.

People don't understand anything about giving it context. They don't understand what is even necessary to improve the AI that they're using. And so for me, this is how the stakes kind of filter through. But in any case, it's like, this is a moment I often find comes up in my hooks, is there's this kind of turning point moment. It's like a, here's your current situation, but the truth is X. Well, the real problem is this.

Yes. Yeah, for sure. Having written the last 8 videos with the help of AI, again, this is, we're kind of talking about that future state, but I'm not explicitly being yet like, if you want to write your next script using AI, It's sort of both implying my own credibility, but also it's presenting an image of their future state. So again, I think of that as the transformation expressed through credibility, I suppose. And spent the last 3 months in the trenches training my own scriptwriting app. I think I got rid of in the trenches in the end because when I spoke, it just sounded a bit weird. I'll show you the exact 5 documents I think it became in the end. You need to turn Claude from a forgetful assistant into a reliable scriptwriting partner.

So that's fully the transformation that we're ending on. So that's how I think about the hooks. And then now we come down to the first segment. So we're not saying in this first segment, yes, it's about making a new project, but we're not saying the first step is making a project. So that's how I set it up. Tension is a really interesting one. We could talk for quite a long time about tension, but the way I like to think of this is if I back up for a second, the problem people have with payoffs is they feel like they have to defer giving any value at all until the payoff. Like, it feels like it has to be this big explosive moment at the end of a segment.

Exactly, exactly that. So yeah, lots of payoffs throughout, but even within that, I still see people thinking that, like, sure, I get the idea I have to have a payoff at the end of my segment, but that means I have to literally say nothing of any value until I hit it, and the rest is all kind of preamble. What I like to do is think of it in this slightly different way, where it's like I start by showing people kind of— I suppose we've done a little bit of this in the hook, but we're now going into detail on the user's present state. So it's once again showing at the start of this, we've got their curiosity with the setup. Now we're showing here is the thing you probably do currently. So in this case, it's, I was surprised to find a bunch of my students were still using Claude in an extremely siloed off way, starting a new chat every time, explaining who they were and what they wanted AI to help them with. This is very much what we've been discussing in this podcast anyway. And then we kind of move through into here's what you actually need to do, create a project.

Here's what a project is. And then go and create it. This is a slightly atypical one because this is so step-by-step, go and do a specific thing. It's normally a bit more kind of nebulous. But for anyone who does have a channel where you're showing people step-by-step things, I think this will be the key takeaway here is make sure that you, rather than just saying step 1, do this, it's explaining in a curious— opening a loop about why this is important, then showing them what they currently do and why it's bad, why it doesn't work. And then you can move through into here's what you should do instead. So then the payoff line, and this is— I've kind of shifted the way I think about these over time, but you'll notice that it's not really like we're giving the final, like, here's what you have to do in that payoff line. It's more bringing people to a level playing field and saying, so here is the state we are now in as a result of what you have just learned.

Here is the new status quo. You have now done that. You are set up, and we are ready to do whatever comes next.

To a degree, I think maybe this is a slightly atypical example because we still want the next setup to be curiosity-inducing and open a loop, which is why you can see here, it's like I say here, the first document— I'm not going to say the first document is this. It's the first document is going to help you fix thing that you're curious about. But if I look at any other examples in here that explain this well, I think this is probably a good example that shows it off. I wouldn't say we're necessarily signposting what comes next yet. I do think that is reserved for the following setup. But again, it's like clarifying with the information you didn't have that you now have, that you learned in this segment. Let's just take a pause to reflect on what is it that we have now unlocked the ability to do? Where is it that we now stand in the overall context of this topic? And I think this is a good example of this, where after we've uploaded, and this is, I think, the 3rd document that I show people, We're now just clarifying. Okay, take a pause.

We've been saying all this stuff about these documents you have to upload, but practically speaking, let's zoom out. What can your project now do so that you understand why we're doing all of this?

Yeah, this is a really kind of tricky semantic dichotomy. Sounds like the name of a band. Yes, I think I agree with you in principle, but I think we want to try and avoid— it's treading the line between that and not just making it a summary. So, yes, tell them what you told them, but I think I want to do it in a slightly different way. So I'm not just repeating So you just added this document. It's more like you add it. So having added this document, new information, this is now what your project is capable of thanks to what you just did, rather than just simply repeating it.

Does that distinction make sense?

Yes, exactly. I think that's a really, a really good distinction. Make people feel good, catch them up, as it were. Like if anyone's kind of gotten lost or they're wondering, wait, have I, have I grasped this? Just taking that moment to say, yes, you've got this, and doesn't it feel great? As opposed to saying, here's what you just learned in much the same way.

I always come back to the wise man Gandalf when I'm asked this question, which is that a wizard arrives precisely when he means to. Uh, this is a tricky one. There are kind of no rules to this. I think a lot of it comes from just the soft skill of doing this over and over again and figuring out, okay, when I talk about X in my videos, my retention always drops off. On reflection, this period where my retention was kind of diving continuously, I realized that I was sort of talking about the same topic but in slightly different words twice in a row or something like that. I think really it just comes down to honest reflection and looking back at what you've written. Um, I, I know one thing that can certainly help with this is just, just showing it to someone else, whether it's in your niche or even better, sometimes someone who's not in your niche, someone who can look at it objectively with no, um, feeling that, or no investment in the topic, no kind of preconceptions of what it's about. And just saying like, is this interesting to, to read? Did, did, did, does, does it fall off at any point for you? When did you get bored? So outsourcing it is a, is a good one.

But I think overall, if we're just talking, you sat at your desk, I think having no preconceptions of how many segments a video should have, uh, is key. If you're trying to target a certain length of your video, I know people like to try and have longer videos for the watch time, whatever. But if you've got an overly inflated video that doesn't have any substance to it, then it's not going to be interesting. So I would say a wizard arrives precisely when he means to. How long is a piece of string? Like, make it the number of segments that it has to be. Be objective and look at, am I Including this segment because it contributes to the understanding of the subject, or am I doing it just to bloat the watch time?

One just thought that came to me as you were saying that there is, I think, a way of making sure that you're still putting some of that thought and time into pre-production, even if you're short of time, especially with AI now, is even if you feel like you don't know what you want your video to be and you don't know how you want to structure it and you feel overwhelmed by the idea of it, spieling to an AI and having it present a structure back to you will immediately set off red flags in your mind of things that it definitely got wrong. So if you look at a structure that it has come back with, I guarantee, and maybe it'll nail it, but most of the time it will have come up with some idea that you're like, no, that's totally not it. And then you realize, oh, actually, I do have some thoughts on this and I do have my opinions on this. But having it presented to me in like slightly the wrong order is much easier then for me to think, okay, no, I would shift segment 1 to segment 2. I would maybe add an extra segment about this. I would connect these 2 ideas like this. And then suddenly you've got your structure out and it feels like you have invested that time in pre-production, but you're sort of spurred into it by seeing what was definitely not the right structure, as opposed to just trying to do it from a blank page.

Yes. Okay, so in this video, I didn't have any product CTAs. I think because I mentioned my AI scriptwriting tool in the hook, I just put the first link or the second link as a link to the waitlist for this tool. That's actually been a really effective call to action for me, is not mentioning it at all, but just showing the thing that I'm talking about the whole video and then putting a link in there in the description. people intuitively just wanna go and learn more about it.

No, so this is my absolute every single time without fail thing that I do is every video has to link to another one in a compelling way. But at the end of every single video, I'm thinking about a link. So that is, back to something in this video. So we're trying to make it not apparent that the video is about to end. We're basically saying we've already spoken about XYZ thing, or now you know how to do blah, blah, blah, blah, blah. If you couldn't see the timeline at the bottom, you might not know that the video is about to end. Then we want to add some curiosity. So about something in the next video.

So in this case here, if we go through these one by one, link back to something in this video, Now, this project setup is a great starting point. Done. It's like simple as that. That's all you need. Um, that shows that we are talking about, we're referring, fucking hell, sorry. Grammarly is really getting on my nerves.

Yeah, dude.

It's okay.

I'm just removing you from Chrome Grammarly. You are, you are so gone. Um, no, it's still here. This is ridiculous. Let's see.

Okay, so if we take a look here, this is clearly the link. This is where we say, now, this project setup is a great starting point. It's just referencing something, well, basically the entirety of the topic of this video. Then we raise the curiosity here where I say, but what if I told you we can ask Claude to gradually improve itself over time so each new script you write gets closer and closer to your style every time? I think in the moment I must have riffed something else here because I say time twice in quick succession. We want to avoid repetition. And then finally, we just have the promise, which is usually spelled like this. There we are. CTA, which is where we give the explicit promise.

And this is where the facade is kind of broken. And we're like, well, yeah, we're promoting another video, go watch it. But ultimately, it's an offer, right? We're giving the viewer something else for free. Well, watch this video next where I'll walk you through how to do this using a simple prompt that you already have inside your Claude product project. So it's like, we're promising something that is exciting for you and feels low lift.

I think that's really cool. I love that. I think Like anything, it's kind of balanced and you're trying to figure out what am I optimizing for mostly here? The advantage of this strategy and always ending with something else is that you can create these kind of infinite loops of people just watching more and more and more stuff. And it doesn't matter which video they land on first, there's always going to be kind of infinite trailer videos they could keep binging if they wanted to. I suppose the only risk with this strategy is, hey, if someone— if it's like the first video that someone's landed on is that one and they get pitched something at the end, what is the effect that has on them? Do they think, oh, I've been sort of sold to here? It was kind of an ask, not an offer. But ultimately, like, I think that sounds like a really, a really good tactic and one that I might steal. And clearly you would think he would have been testing something like that, right? And the pros would outweigh the cons.

Yeah, I think, again, that's something you can really easily start using AI to help you with, where it's for something like this that has a formula that I try and use. Anything that's formula-based, if you've got something simple, we spoke about it with the hook as well. It's very easy then to plug in the actual content that you want and to ask it to express that content in terms of that formula. So in your case, I suppose using a formula like that, but saying, like, hey, actually, we don't want to promote another video. We want to promote X or Y product. We want to promote the lab. We want to talk about whatever. We want to promote a VSL that talks about the lab or something.

I think it would be fairly easy then for it to come up with a relatively subtle plug in the style that we've just seen here, but for a VSL.

Okay. Yeah.

That's such a good question. I think to aid my case here, at least in terms of the objectivity, I'm in the fortunate position where I've managed to sort of work my way out of having to write scripts for people. And so I think I'm now able to grow my personal brand or whatever you wanna call it as someone who helps people learn to do scriptwriting in whatever context that is. And now that context is AI related. So I think I'm able to divorce myself a little bit from the desire to like have scriptwriters around because it's what I used to do a lot and it's what I still do for my own channel. One thing that I've learned a lot about since starting to really use AI a lot and using it to code my own scriptwriting app, for example, is I've spent a lot of time watching how AI actually does what it does behind the scenes. Like, in Claude, for example, you can sort of see how it's thinking as it's thinking, and you can read its decision-making. You sort of learn all these little ways in which it may appear to be doing something well, and then the output that comes out, it actually reveals, oh, hang on, it wasn't doing these 10 things that it claimed very confidently that it was.

So I think when it comes to scriptwriting, the final say on when have I done enough revisions for this script, When is this actually record ready? That ultimate decision comes down to a person's brain, because AI will always have something else that it's like, oh, we could change this, and, oh, you actually didn't do this correctly, and, oh, this could be stronger so we could make it land better, or whatever. And it will just keep doing that kind of infinitely. And I think that moment where you have to say to yourself, no, actually, I think this is ready. I know my audience as well-rounded as I possibly could. I'm the one who says that we've now done enough revisions on this, and I think it's going to hit really well. But I think there are those moments that come throughout the process. So it might be when structuring out the script in the first place, AI will be like, hey, here's a brilliant structure based on your audience avatar that I think they're going to love. And you can read that and be like, I like 70% of these ideas, but ultimately I want to rearrange X and Y because I know this will be how I would actually explain it or how they would receive it better.

And so again, the point of deciding structuring is now done, that kind of needs to be me, because AI will keep on coming up with new things that you should add or change or improve.

Ooh, that's a really good question. I think with a lot of anything creative, there's an element of brutality that needs to come, that comes down to it in the end. There have been some scripts that I have written which have seemed like a great idea, have been maybe like 5 segments or kind of subsections initially. And then as we've kind of workshopped, either me personally or me in collaboration with an AI, have workshopped it, it's kind of trimmed down and down to maybe the point where there's like 3 segments. And it reaches a point where I'm like, I think this could just be a newsletter or something that's a bit more short form. And so I think the first thing would be to say, accepting the reality that it may be that there just isn't enough to the topic. But I think this is one of the real meta skills to learn, and it's something I'm sure you're aware of anyways, like just the ability to talk to AI to extract from it exactly what you want. So talking to it in natural language and say, look, here's the predicament I'm in.

I have this exact topic I want to talk about. Here's the context on my last video about having a kid. This one I want to do now is about how I wasn't prepared properly. But I'm worried there isn't going to be enough to talk about it. Knowing what you know about my audience, if you don't have an audience avatar, you'd be like, here are 10 transcripts of some videos we did recently. Here are some recent comments from them. Knowing what you know about them, what blind spots do I have that could relate to this topic that we could put into this video? And kind of using it then to challenge you and to almost make you answer some questions you might not have even thought to ask yourself that can then contribute to making a video that feels like it warrants existence.

It's an interesting moment, isn't it? Because it's quite— it becomes harder and harder to extricate yourself from it and to reason with yourself and think as you would've thought before what's now possible were possible. My business didn't rely on me moving with the times like this. Would I be going as all in on it as I am? I don't know. I'm certainly loving building an app. This is something that I had spoken to different companies about in the last few years that I, as sort of being brought on as like a as the sort of the brains, as it were, as the scriptwriter behind a tool that would then be coded by a bunch of very intelligent people. The fact that those very intelligent people are now not strictly needed, although I'm sure anyway, there's a whole discussion there, but I love that I've been able to build my own thing. This is the tool that I've dreamt of making, and it does all the things that I've always wanted. That's very cool.

Do I feel the most creatively fulfilled that I have? Uh, in the 4 or 5 years I've been running this business? In some ways, no. And as time goes on, the more I kind of want to shut it all down, go live in the woods, and make hour-long YouTube videos just chopping wood.

Yeah.

Yeah, I think I, I'm completely with you. It's hard. And I don't know if you feel this, especially from a business point of view, where I'm trying to build an AI-related business that doesn't completely remove the human element. I don't want to build a scriptwriting tool that writes every single word for you, and you go away and you record something that you just didn't— you had nothing to do with. Which is why the sort of core feature of it is this discursive element that we've discussed, where it's like there's a chat, you chat to it, you leave your voice notes, And you kind of work with them as if they were a scriptwriter that you'd hired. But at the same time, I'm very conscious that this tool also would completely serve people who want to start a YouTube channel as a cash cow, faceless, no shade on faceless in general, but like one of those channels where it's like, hey, I don't have to do anything. I can AI the voiceover. This guy's got this tool out and it can just AI the script and I can AI the thumbnail.

And so being conscious that you're kind of, you're serving both the more conscious people who want to keep the human element involved, but also the people who want to just outsource it all. And yeah, that is kind of conflicting. It's a weird moment.