His EXACT System For Writing GREAT YouTube Scripts With AI
- URL: https://www.youtube.com/watch?v=So77eF3rCao
- Channel: Jay Clouse
- Fetched: 2026-10-18T23:31:07+00:00
- Language: English (auto-generated)
Transcript
[00:00] Where is it worth putting my energy into
[00:01] learning to use AI well? And where
[00:04] should I deliberately take a stand and
[00:06] be like, “No, I’m using my brain here
[00:07] because it’s actually better to do so.”
[00:09] Meet George Blackman. He started as a
[00:12] full-time script writer for Ali Abdal.
[00:14] He’s written scripts for creators from
[00:16] 30,000 to over 10 million subscribers,
[00:18] and he now teaches creators how to write
[00:20] their own scripts that keep people
[00:22] watching and turn them into paying
[00:24] customers. In this episode, you’ll learn
[00:26] his exact system for writing scripts
[00:27] with AI, starting with the framework
[00:30] behind every hook he writes, for
[00:32] educational hooks. I always think target
[00:34] transformation stakes. He’s also
[00:36] brutally honest about the ways AI lies
[00:38] to you in the process. You sort of learn
[00:40] all these little ways in which it may
[00:42] appear to be doing something well and
[00:44] then the output that comes out, it
[00:45] actually reveals, oh, hang on, it wasn’t
[00:47] doing these like 10 things that it
[00:48] claimed very confidently that it was.
[00:49] and stick around to the end where George
[00:51] shows me the tool he built for writing
[00:53] killer scripts using AI that saves you
[00:55] time without sacrificing your unique
[00:57] voice. Thank you to one of 10 for
[00:59] sponsoring this video. As just a a frame
[01:02] for the overall conversation, some may
[01:05] think that okay, because we have AI,
[01:09] is that doing all of the work of script
[01:11] writing now? What say you? I think this
[01:13] is a place that everyone is now in with
[01:16] pretty much any industry whether it’s uh
[01:18] to do with writing specifically whether
[01:20] it’s coding whether uh it’s like
[01:22] anything that could feasely feasibly be
[01:24] affected by AI there is this sense of
[01:26] uncertainty and like where should I put
[01:29] my energy right now? Should I be putting
[01:30] my energy into uh like getting really
[01:33] good at ideiation because the script can
[01:35] just be written for me or is ideation
[01:37] already off the cards cuz AI can just do
[01:39] that for me? um do I even need to bother
[01:41] learning to code anymore because AI can
[01:43] just do that for me? And I think script
[01:45] writing is such a broad church. There
[01:47] are so many aspects to it that even with
[01:50] that it’s like okay sure AI could write
[01:52] my whole script but should I still be
[01:54] the one looking at my hook because I
[01:55] know the hook’s super important or are
[01:57] these like key moments where I call out
[01:59] to a product that I’m trying to to sell
[02:00] or something in the script. Are they so
[02:02] important that I should still be using
[02:03] my brain or is AI better than me? And so
[02:05] more broadly like we’re gonna we’re
[02:06] going to talk about the sort of process
[02:08] that I use these days to write a script
[02:10] using AI. But I think the broader meta
[02:13] skill and the thing to try and figure
[02:15] out and that I’m trying to figure out
[02:16] and I think we’re all trying to figure
[02:17] out is like where are the parts where
[02:18] it’s the most useful? Where is it worth
[02:20] putting my energy into learning to use
[02:22] AI well and where should I deliberately
[02:25] take a stand and be like no I’m using my
[02:27] brain here not through some like moral
[02:30] crusade but because it’s actually better
[02:31] to do so. Can you walk us through, let’s
[02:34] imagine you are making a video for the
[02:37] George Blackman channel. How are you
[02:39] going about script writing today? The
[02:41] first step and this is where the the use
[02:44] case for AI is is much much stronger. I
[02:47] think even just to re refer back to your
[02:49] previous question, it’s like where are
[02:50] people needed? It’s almost easier to to
[02:52] list the few places where AI is
[02:54] definitely a no-brainer. And the first
[02:56] step of my process these days is very
[02:58] much where it is a no-brainer, which is
[03:00] just explaining my thoughts about the
[03:02] script that I want to write verbally
[03:04] using um either like a voice note thing
[03:06] or or just directly into Claude. Just
[03:08] that ability to spiel off in natural
[03:10] language what I think about a topic,
[03:12] even if it just allows me the chance to
[03:16] uh clarify my thoughts on something that
[03:17] I’ve never said out loud. Kind of the
[03:19] same idea as as journaling or something.
[03:20] It allows you to to just get more out
[03:22] than if you’re just vaguely thinking
[03:23] about something in your head. And is
[03:25] this assuming that you’ve already
[03:26] decided on the video idea if not
[03:29] package? Yes, this is assuming you’ve
[03:31] already you’ve already thought of the
[03:32] packaging or I guess yes, ideally the
[03:34] packaging. Not everyone does it. There’s
[03:35] no shame if you don’t come up with your
[03:36] packaging first, although obviously try
[03:37] to, but yes, the idea is already decided
[03:40] at this point. I guess we could talk
[03:41] about ideas as well, but that’s less
[03:43] where I’m involved. Um, this is I know
[03:45] what I’m going to talk about. Maybe I
[03:47] don’t know exactly how I want to frame
[03:49] it. Um, an example from my channel
[03:51] specifically was like I wanted to do a
[03:53] video about my sort of journey in my
[03:57] business, but I knew I needed to tie it
[03:59] back to YouTube script writing in some
[04:01] way. And so I spoke out loud to an AI
[04:04] about this and I was like, here are the
[04:05] the sort of points in my business when I
[04:07] was a script writer, then I was like a
[04:08] consultant, now I run this community and
[04:09] blah blah blah blah blah, but what’s the
[04:12] sort of angle that makes this
[04:13] interesting for my audience who cares
[04:15] maybe a bit about business but more
[04:16] about script writing? And what it
[04:19] allowed me to do was get those thoughts
[04:20] out and then have some time going back
[04:22] and forth on the angle that we we should
[04:25] approach the subject. And ultimately we
[04:27] kind of landed on this approach of at
[04:29] every stage of my business, how have my
[04:32] thoughts about script writing changed
[04:33] and like what’s the key lesson about
[04:34] script writing I learned at each revenue
[04:37] level or or whatever which kind of um
[04:39] yeah tied it back to my audience’s
[04:40] interests.
[04:41] >> Yeah. Let’s let’s imagine that somebody
[04:42] watching this doesn’t have any type of
[04:44] complex projects set up. Maybe they’ve
[04:48] never used Claude or Chat GPT to do
[04:52] anything like this. And so what should
[04:55] be the starting point? Is it just like
[04:57] dumping this in or do you think they
[04:59] should take some time and say like,
[05:00] “Okay, here’s my audience. Here’s this.
[05:02] Here’s that.” Like what context does
[05:05] this thing need before you start talking
[05:07] about a particular video, if it needs
[05:09] context at all? So, first thing I
[05:11] currently use Claude. I used to use
[05:13] ChatgBT. I switched to Claude because I
[05:15] wasn’t loving ChatgBT’s outputs. Since
[05:17] I’ve switched to Claude, I think uh it
[05:20] seems like Chat GBT is then has had a
[05:22] kind of resurgence at the time of
[05:23] recording. I think it’s on 5.5 now, is
[05:25] it? Or is it something else? Anyway,
[05:27] sidetrack. I’ll use Claude. Either one
[05:30] is fine is what we’re hearing. Isn’t it?
[05:31] >> Either one is fine. Yeah,
[05:32] >> you’ll probably have workable good help
[05:35] using either one of these things.
[05:37] >> Yeah, exactly that. Um, but in terms of
[05:39] giving the AI enough context, I think
[05:41] there’s really like an easy route and a
[05:42] hard route. So the easy mode of this I
[05:45] would say is you need a style guide and
[05:48] you need an audience avatar and arguably
[05:50] then some project instructions like
[05:52] telling the project what it is for.
[05:54] That’s really the the simplest thing.
[05:56] How do you make those? Again, it’s very
[05:58] simple. This is that meta skill again
[05:59] where it’s like you you explain to it
[06:01] this is what this project is. I need a
[06:04] style guide. I need an audience avatar.
[06:06] What do you need from me? So you can
[06:08] create those for me if that makes sense.
[06:11] How many scripts of mine or like
[06:12] examples of writing or if you don’t have
[06:14] any examples of writing you would say
[06:15] like can I just speak to you for 5
[06:18] minutes about a random topic and you get
[06:19] my voice from that. But th those would
[06:21] be the two things that um you can just
[06:23] ask your LLM for help making. So that’s
[06:25] the easy mode. I think the advanced mode
[06:28] shall we call it and this is how I I
[06:30] start my projects these days is same
[06:32] thing starting a new project with um
[06:35] whisper flow or just you know anything
[06:36] that you can you can leave a voice note
[06:38] to and just explaining this is the the
[06:42] task that I have before me. So I did
[06:43] this when I was upgrading my uh my
[06:46] course. I was like I need to upgrade my
[06:48] course. My course is 36,000 words long.
[06:50] Since releasing it I have uploaded 2
[06:53] years worth of newsletters. I want to
[06:54] integrate those newsletters into my
[06:56] course. I have XY Z other concerns.
[06:59] 36,000 words is quite long. Will you be
[07:01] able to handle that? Just giving it all
[07:03] the context that I thought it might
[07:04] need. And then basically saying to it
[07:06] like what is my first step here? What do
[07:07] you need from me? And I think people
[07:10] underestimate this approach a lot
[07:12] because it kind of feels like ah I guess
[07:13] I need to know how to use this thing
[07:16] before I start using it. But the joy of
[07:18] it is and why it feels so kind of
[07:20] weirdly addictive is you can just say
[07:21] like I don’t know what I’m doing. like
[07:22] what do I how do I how do I proceed?
[07:24] This is legitimately like what everybody
[07:27] is selling in the world of teaching you
[07:30] how to claude. By the way, it’s like I
[07:31] did this thing. Pay me $99 and I’ll show
[07:34] you how I did it. And it inevitably
[07:35] comes back to and then you ask it how to
[07:38] do it and it will tell you like that’s
[07:40] that’s the entire secret of working with
[07:42] AI is like just start with I’m trying to
[07:45] do this thing. What do you need from me?
[07:47] Literally, if you counted how many times
[07:48] I asked my my open claw, “What do you
[07:51] need from me?” It would be astronomical.
[07:54] Yes. Yeah. 100%. And and like recently
[07:57] building this app, I was suddenly aware
[07:58] like, “Hang on, there are probably some
[07:59] legal ramifications of selling a
[08:02] subscription thing that relates to AI.”
[08:05] And so I just said to it like, “What
[08:07] what are the legal ramifications?” And
[08:08] then this enormous block of text came
[08:10] out and it was like, “We have these 25
[08:12] tasks to do. You need to contact this
[08:13] lawyer. You need to blah blah blah
[08:15] blah.” And suddenly it’s like, oh yeah,
[08:17] okay. I mean, I didn’t know how to do
[08:18] that, but but now I do. Either mode or
[08:20] hard mode, you get up, you set up the
[08:22] context for your voice guide, your
[08:24] audience avatar, and now you start
[08:26] saying, “Okay, now that I’ve zeroed in
[08:28] on a video idea, I’m going to just brain
[08:31] dump voice dictation or type whatever
[08:34] pleases you, what I’m thinking about
[08:36] this video.” How do you give it the
[08:38] instruction for the next step? The next
[08:41] kind of thing that I like to have is
[08:43] something that trains it on what a
[08:45] YouTube script looks like in relation to
[08:47] other types of writing. You can totally
[08:49] build these yourself. You could even
[08:51] just take a bunch of transcripts from
[08:53] your own channel, from channels in your
[08:54] niche that are doing well. Not to say
[08:57] steal how they do stuff, but like
[08:58] structurally, what what what do you
[09:00] detect about the structure of a YouTube
[09:01] script? And how can we use that to train
[09:04] you on what a script should look like? I
[09:06] have a free thing that has my version of
[09:08] what YouTube structure looks like. It’s
[09:10] in the description right now. That’s
[09:11] kind of the next thing. And because I
[09:13] think then it saves you the hassle of
[09:14] having to be like to to say here is my
[09:17] brain dump of thoughts now structure it
[09:19] like a YouTuber would. It kind of has
[09:22] this pre-loaded context of what a
[09:23] YouTube script would look like. So that
[09:25] that’s really the next step is once
[09:27] you’ve brain dumped your stuff, you want
[09:31] to either through the pre-loaded
[09:32] training or saying to it outright, let’s
[09:34] first of all figure out what the
[09:35] overarching structure of this would look
[09:37] like. and it’s going to come back to you
[09:39] with a suggestion and you you see if it
[09:41] it fits and go from there. For people
[09:43] who are stubborn and haven’t downloaded
[09:44] this free thing yet, what are like the
[09:46] highlevel bits of this in there that
[09:50] people should know is like part of this
[09:52] structure. Three main blocks, shall we
[09:56] call them, um that I would think of for
[09:57] a YouTube script. The hook is is the
[09:59] obvious first block. Uh it’s quite
[10:02] distinct how you would go about writing
[10:03] a hook. There are different frameworks
[10:05] that different people have for this. For
[10:07] me, for educational hooks, I always
[10:09] think target transformation stakes. So
[10:11] that is who is my target, who is my
[10:12] audience. What is the transformation
[10:14] this video offers and what is at stake
[10:16] if they don’t achieve that
[10:17] transformation? The second main block is
[10:19] is a segment. So these are your
[10:21] subtopics within the video. For every
[10:23] segment of my video, I think of it in
[10:25] terms of setup, tension, and payoff. So
[10:27] at the start of each segment, we set up
[10:29] why it’s important for the viewer to
[10:31] know this thing before we can move on to
[10:34] the next thing. The tension is where we
[10:37] teach the lesson essentially. We kind of
[10:38] reveal the information and then the
[10:40] payoff at the end of that um that
[10:42] segment is where we ultimately
[10:45] I suppose reveal the answer but like the
[10:47] the understanding of that segment is now
[10:49] complete and we can move on to the next
[10:50] one. And then the third one is a is a
[10:52] call to action. So whether that is a
[10:53] call to action at the end to promote
[10:56] your next video or product or service or
[10:58] a midvideo call to action um promoting
[11:01] whatever it might be a free lead magnet
[11:02] or or something like that. Can you give
[11:04] me an example hook on a video you’ve
[11:06] done for yourself recently? I like this.
[11:07] You said it was target transformation
[11:09] stakes. Target transformation stakes. I
[11:12] would love to hear like an example of
[11:14] that. Target is always very easy to me
[11:16] cuz I’ve got a five-page audience avatar
[11:18] document that I’ve made in in the way
[11:20] that we kind of described earlier. So, I
[11:22] don’t need to think much about that. I
[11:23] know who I’m targeting. It’s YouTubers
[11:25] who are interested in script writing. Uh
[11:27] script writers for hire who want to
[11:29] learn to get better at script writing so
[11:30] they can get more clients and blah blah
[11:31] blah. Transformation. If I take an
[11:33] example of one of my most recent videos
[11:35] and very uh apt for the conversation
[11:37] we’re having now was how to set up a
[11:39] Claude project um for to to turn Claude
[11:42] into a reliable script writing partner
[11:45] uh as opposed to a sort of irritating um
[11:47] you know assistant who forgets
[11:48] everything. The transformation within
[11:49] that is is like you will go from not
[11:53] knowing how to do that to knowing how to
[11:55] do that. What’s at stake is this idea
[11:57] and it’s something that I think a lot of
[11:59] my audience have felt. Something I
[12:00] certainly felt when I was getting on
[12:01] board with AI is the sense that you will
[12:04] actually end up spending more time using
[12:06] AI to write your scripts than if you
[12:08] just use your brain. So so determined
[12:10] are you that this is the new thing and
[12:12] hey I should be more productive with
[12:14] this that you end up just getting
[12:16] frustrated with it. It doesn’t quite do
[12:17] what you want and you end up wasting
[12:18] more time. And so for me that that’s
[12:20] what’s at stake if they don’t achieve
[12:22] the transformation is you will remain in
[12:24] that state where this sort of perceived
[12:26] productivity of AI slows you down
[12:28] results in worse scripts uh means your
[12:30] channel doesn’t grow kind of thing. So
[12:32] that’s what I think before going into
[12:36] physically trying to write a script. And
[12:37] the joy of AI these days is that like
[12:39] you can say look here’s my target here’s
[12:40] my transformation here’s what’s at stake
[12:42] here is a hook framework that has
[12:45] performed really well on my channel in
[12:46] the past please express these factors in
[12:50] terms of this structure um and it’s it’s
[12:53] pretty good at very finite structural
[12:56] tasks like that a lot of the time and
[12:58] then out of that comes a perfect script
[12:59] and I’m good to go right that’s it dude
[13:01] and I’m out of a job see you later
[13:05] yes so the joy way of all of this is
[13:08] that it this allows me to go from blank
[13:10] page to V1 bulletoint scripts very
[13:14] quickly, much more quickly than before.
[13:17] I’m hearing this and it’s like I could
[13:18] do this while I’m out on a walk with a
[13:20] stroller. Like if I had this
[13:21] infrastructure set up um and I had an
[13:24] idea, I could literally do this while
[13:26] I’m out on a walk and it could create
[13:28] the artifact waiting for me when I get
[13:30] back to my desk. And that is a
[13:32] meaningful source of friction to
[13:34] eliminate that would make this feel much
[13:36] more real. And I would go from like I
[13:38] should make that video someday to like I
[13:39] bet I can record that tomorrow with a
[13:41] little bit of extra work here.
[13:42] >> Exactly that 100%. That’s that’s a
[13:45] really good point. Um, so not only does
[13:47] it save you that time in the moment sat
[13:50] down, it may well be that you don’t even
[13:51] have to be sat down for it to for it to
[13:52] happen. But I think also it saves you so
[13:55] much mental strain having to restructure
[13:59] these thoughts on a page that
[14:01] perfectionism creeps in. It feels like
[14:03] even if you’re just typing notes like
[14:04] no, I should arrange these in the
[14:05] correct order first and blah blah blah.
[14:07] All of that thinking uh power is is
[14:10] saved by just being able to talk to it
[14:11] like a like a person. So we have our
[14:13] structure say and we’ve agreed with it.
[14:14] Yes, maybe with a bit of back and forth,
[14:17] I am now happy with the six segments
[14:19] that we have for this video about Jay
[14:21] becoming a dad for the second time. Um,
[14:24] now I want to like move it forward into
[14:25] something more more word for word. This
[14:27] is kind of a bit of an inflection point.
[14:29] Maybe that’s the wrong word for it. It’s
[14:31] it’s a a decision point on the the tree
[14:34] as to whether you want to say, look,
[14:36] this is as much help as I needed. Now, I
[14:39] sincerely want to be able to just do the
[14:41] rest of this by by hand, as it were, by
[14:44] brain. I don’t want to then have an AI
[14:46] physically try and write me a full word
[14:48] for word script because I either that’s
[14:50] not going to uh make me feel creatively
[14:52] fulfilled, or b I just don’t trust that
[14:53] it’s going to do what I want it to do
[14:55] yet. So, there is a point at which you
[14:57] might just decide this is where I finish
[14:59] the AI thing, and that’s that’s
[15:00] completely fine. If I were to say, I
[15:02] don’t want AI to go any further, then I
[15:04] suppose you would then teach us how to
[15:05] think even more like a script writer and
[15:07] work through this, which is probably
[15:09] best to do anyway because the more
[15:11] knowledge we have, the the more
[15:13] collaborative we can be with bossing
[15:16] this machine around if that’s what we
[15:17] choose to do. So, let’s let’s say AI
[15:20] turned around a V1 of the script and
[15:22] we’re looking at it and we want to take
[15:24] this to from from this V1 to something
[15:27] that I’m ready to record. Let’s say at
[15:29] this point we have our structure. I’m
[15:31] going to assume at this point we’ve
[15:32] mentioned bulletoint scripts a few
[15:34] times. It may be that you had your
[15:36] initial like here are the one sentence
[15:38] headline things for each of our
[15:40] segments. So like six literally six
[15:42] sentences. This is the skeleton. We’re
[15:44] assuming we’ve now expanded it into uh
[15:47] say two or three bullet points per per
[15:49] segment. And so really the choice here
[15:51] is if you are someone who really enjoys
[15:54] uh riffing, if your niche is the sort of
[15:56] niche that will allow you to sit down
[15:57] and riff, then you can very much test
[15:59] that. You can think, okay, I’m literally
[16:00] going to go with what we have. Sit down,
[16:02] start talking, see how it feels in the
[16:03] moment. See how much pain is deferred to
[16:06] the edit because of how much I may have
[16:07] rambled if I was just going off of
[16:09] bullet points. And then adapt uh and
[16:12] adjust each time you sit down. You
[16:13] think, “Ah, actually, I felt completely
[16:15] out of control when I was sat down last
[16:17] time. This time I want to go from two to
[16:19] three bullet points per segment to five
[16:22] to six. I’m going to turn that dial up.
[16:23] I think of it like a dial. Just just
[16:24] turning it up a little bit. Testing it
[16:26] again next time. See how it feels. Did I
[16:28] ramble? Did that feel better? Was it now
[16:30] too prescriptive? Now I actually want to
[16:31] dial it back to four bullet points or
[16:34] something like that. So that’s a really
[16:35] that’s I suppose a personal thing. And
[16:38] that’s if you choose the the riffing
[16:39] side of things. If you now want to go
[16:43] fully word for word, and this is
[16:44] something I massively prefer, the the
[16:46] amount of pain that is saved on editing,
[16:49] which is thankfully not even my job, but
[16:51] even just on filming day, the amount of
[16:53] false starts and thinking like, “Oh, did
[16:54] I screw this up? Wait, did I actually
[16:56] say that thing 5 minutes ago that sets
[16:58] up the thing I need to say now?” For me,
[17:00] that’s just not worth it. I much prefer
[17:02] to put an extra 30 minutes, 60 minutes
[17:04] into writing a full word for word script
[17:06] to save me that pain later. If you
[17:08] prefer to spend less time scripting and
[17:10] more time recording, the OneUp of 10
[17:12] team just released a new feature that
[17:13] will help you edit your talking head
[17:15] videos fast. It’s called auto cut and
[17:17] it’s part of their new video editor. I
[17:20] have a great video setup here in my
[17:22] studio, but I don’t actually like to
[17:24] script full videos. That means my
[17:26] recordings have all kinds of extra takes
[17:28] and dead space. And it’s not that it’s
[17:30] hard to edit those out, but it’s
[17:32] timeconuming and I dread doing it.
[17:34] AutoCut takes your raw talking head
[17:36] footage, watches it frame by frame,
[17:38] removes silences, rambling, and retakes,
[17:41] enhances your audio, and gives you a
[17:43] clean cut that’s ready to post in just
[17:46] minutes. It doesn’t just put those extra
[17:47] takes in the trash, either. It saves
[17:50] them so you can instantly switch to
[17:51] whatever take you prefer. The One10 team
[17:54] is known for their deep knowledge of
[17:55] YouTube, and this tool is already better
[17:57] and faster than any other tool I’ve
[17:59] tried for cutting your raw footage
[18:01] quickly. visit the link in the
[18:02] description and you can try auto cut
[18:04] yourself for free. Then the focus and
[18:08] this is what I always focus on is
[18:10] obviously the hook is very important but
[18:12] the setups and the payoffs are the most
[18:14] important thing that I like to uh write
[18:18] word for word almost regardless. So your
[18:20] setup line at the start of each segment
[18:22] is is like another hook basically. It’s
[18:24] the the the reminder of not only why I
[18:27] care about this overall topic, but why
[18:29] this particular subtopic is so essential
[18:31] in me moving forward. So if we take the
[18:34] example of my claude project video or
[18:37] yeah, you know how to how to set up a
[18:38] claw project. The first step of that was
[18:42] literally you need to set up a project
[18:43] because a lot of people don’t do that.
[18:45] But the way we set that up compellingly
[18:47] is not step one set up a project on your
[18:50] Clawude uh your Claude chat.
[18:52] But before we get into the five
[18:54] documents you’re going to need, we first
[18:56] have to fix one of the most common
[18:59] mistakes I still see dozens of my
[19:01] students making that turns AI into a a
[19:04] useless assistant or something like
[19:06] that.
[19:06] >> I see. I’m hearing that as an open loop.
[19:08] >> Yeah, exactly that. Exactly that. So,
[19:10] focusing on those open loops, those
[19:12] setups at the start of each segment and
[19:13] and just reminding the viewer in a
[19:15] curiosityinducing way why this next step
[19:17] is important. That’s what those are
[19:19] those moments in your retention graph
[19:21] where people will go. It’s when you
[19:23] finish one subject and you don’t
[19:25] immediately make it clear why the next
[19:27] subject is going to be important to them
[19:29] specifically in the context of what they
[19:31] want and what the what the video title
[19:33] promised. If you had a literal script
[19:35] you’ve worked on in the recent past,
[19:37] especially if it’s for your channel so
[19:38] you know that you have the permission to
[19:39] share it where you could be like, “Let
[19:40] me walk you through what the hook looked
[19:43] like in this video and then I’m going to
[19:45] show you one of these setups and payoffs
[19:47] in one of these segments.” I think that
[19:48] would be enormously helpful.
[19:49] >> Can I do it on my app where I wrote it?
[19:52] >> Sure.
[19:52] >> This actually is that very video uh that
[19:54] we were just talking about. This is how
[19:56] to set up a claw project for script
[19:58] writing. So you can see this this is the
[19:59] software that I’ve built and so I break
[20:01] it up into these setup tension and
[20:03] payoff things that I’ve spoken about a
[20:04] lot. Very visually pleasing. This does
[20:06] make a lot of sense. It’s modular.
[20:08] You’re talking about the structure in a
[20:09] modular way. You’re showing this in a
[20:10] modular way. And for what it’s worth,
[20:12] anybody watching this, this is not a
[20:14] paid endorsement or partnership with
[20:16] George on this, but this is a great way
[20:18] to visualize it. And so, here we are.
[20:20] So, in terms of the hook, have you ever
[20:21] spent hours fighting with Claude? This
[20:23] already, even though it’s not explicitly
[20:24] saying if you’re a YouTuber who
[20:26] struggles with, but it’s implicitly
[20:28] showing people, I understand you. This
[20:30] is the struggle that you have. So, if
[20:32] you’ve ever spent hours fighting with
[20:33] Claude only to check the clock and
[20:34] realize this YouTube script took two two
[20:36] times longer to write than if you just
[20:37] used your brain, then listen up. I’m
[20:39] using quite casual language in moments
[20:41] like this. With this video, we were
[20:43] experimenting with some like kind of
[20:44] comical intro B-roll. So, this is the
[20:47] idea that it’s like I show that I
[20:48] understand you. I know who my target is.
[20:50] The problem with AI help scripts isn’t
[20:52] that it’s not good enough yet. It’s
[20:53] because you’re not giving it enough
[20:54] context. I guess this is kind of
[20:56] implying the stakes. It’s a little bit
[20:58] tricky to always distill these into the
[21:00] exact three things, but like if I was
[21:01] listing these things out, I would be
[21:04] thinking when I list out my stakes, I’d
[21:05] be saying AI continuously takes longer.
[21:08] People don’t understand anything about
[21:10] giving it context. They don’t understand
[21:12] what is even necessary to improve the AI
[21:14] that they’re using. And so for me, like
[21:16] this is how the stakes kind of filter
[21:18] through. But in any case, it’s like this
[21:20] is a moment I often find comes up in my
[21:22] hooks is there’s this kind of turning
[21:24] point moment. It’s like a here’s your
[21:26] current situation, but the truth is X.
[21:29] Well, the real problem is this. I’ve
[21:30] heard of a framework. It’s P A I P S and
[21:33] the I is intrigue. And this is reminding
[21:36] me of intrigue. Like PA is pain,
[21:39] agitate. We’ve we’ve kind of talked
[21:40] about that already, but intrigue is like
[21:43] sort of a point of surprise before
[21:45] introducing a potential future as that
[21:47] next P. So, this is this line makes me
[21:50] feel like ah it’s a it’s a head turner.
[21:53] It’s like you know something that I
[21:54] don’t know.
[21:55] >> Yes. Yeah. Yeah. For sure. Having
[21:57] written the last eight videos with the
[21:58] help of AI. Again, this is we’re kind of
[22:00] talking about that future state, but I’m
[22:01] not explicitly being yet like if you
[22:04] want to write your your next uh your
[22:06] next script using AI. It’s sort of both
[22:08] implying my own credibility, but also
[22:11] it’s uh it’s presenting an image of
[22:13] their future state. So again, I think of
[22:14] that as the transformation expressed
[22:16] through credibility, I suppose. And
[22:18] spent the last 3 months in the trenches
[22:20] training my own script writing app. I
[22:21] think I got rid of in the trenches in
[22:22] the end cuz when I spoke it just sounded
[22:24] a bit weird. I’ll show you the exact
[22:26] five documents I think it became in the
[22:28] end.
[22:30] Uh you need to turn Claude from a
[22:31] forgetful assistant into a reliable
[22:33] script writing partner. Uh so that’s
[22:34] fully the transformation that we’re
[22:35] ending on. So that’s how I think about
[22:37] the hooks. And then now we come down to
[22:39] the first segment. So we’re not saying
[22:41] in this first segment yes that it’s it’s
[22:43] about making a new project but we’re not
[22:44] saying the first step is making a
[22:46] project. So that’s how I set it up.
[22:48] Tension is a really interesting one. We
[22:50] could talk for quite a long time about
[22:52] uh about tension, but the way I like to
[22:54] think of this is, if I back up for a
[22:57] second, the problem people have with
[22:59] payoffs is they feel like they have to
[23:01] defer giving any value at all until the
[23:03] payoff. Like it feels like it has to be
[23:05] this big explosive moment at the end of
[23:06] a segment,
[23:07] >> right? Like one payoff per video. I’m
[23:09] thinking about this like these are like
[23:11] you’re you’re uh nestling multiple
[23:13] payoffs into this. The whole video
[23:15] itself is a payoff. There are five
[23:17] payoffs that add up to the payoff is
[23:20] what I’m seeing. Exactly. Exactly that.
[23:23] Um, so yeah, lots lots of payoffs
[23:25] throughout. But even within that, I
[23:27] still see people thinking that like,
[23:29] sure, I get the idea I have to have a
[23:30] payoff at the end of my segment, but
[23:33] that means I have to literally say
[23:34] nothing of any value until I hit it, and
[23:36] the rest is all kind of preamble. What I
[23:38] like to do is think of it in this
[23:39] slightly different way where it’s like I
[23:42] start by showing people kind of
[23:45] I suppose we’ve done a little bit of
[23:46] this in the hook, but we’re now going
[23:47] into detail on the user’s present state.
[23:51] So it’s once again showing at the start
[23:52] of this that we we’ve got their
[23:53] curiosity with the setup. Now we’re
[23:55] showing here is the thing you probably
[23:57] do currently. So, in this case, it’s I
[23:59] was surprised to find a bunch of my
[24:00] students were still using Claude in an
[24:01] extremely siloed off way, starting a new
[24:03] chat every time, explaining who they
[24:04] were and what uh what they wanted AI to
[24:06] help them with. This is very much what
[24:08] we’ve been discussing in this podcast
[24:09] anyway. And then we kind of move through
[24:11] into here’s what you actually need to
[24:12] do, create a project, here’s what a
[24:14] project is, um, and then go and create
[24:18] it. This is a slightly atypical one
[24:20] because this is so step by step. Go and
[24:22] do a specific thing. It’s normally a bit
[24:24] more kind of nebulous, but for anyone
[24:25] who does have a channel where you’re
[24:26] showing people step-by-step things, I
[24:28] think this would be the key takeaway
[24:30] here is make sure that you rather than
[24:31] just saying step one, do this. It’s
[24:34] explaining in a cur opening a loop about
[24:37] why this is important, then showing them
[24:39] what they currently do and why it’s bad,
[24:41] why it doesn’t work, and then you can
[24:43] move through into here’s what you should
[24:44] do instead. So then the payoff line and
[24:46] this is I’ve kind of shifted the way I
[24:47] think about these over time but you’ll
[24:49] notice that it’s not really like
[24:52] we’re giving the final like here’s what
[24:54] you have to do in that payoff line. It’s
[24:56] more bringing people to a level playing
[24:58] field and saying so here is the state we
[25:00] are now in as a result of what you have
[25:02] just learned. Here is the new status
[25:04] quo. You have now done that you are set
[25:06] up and we are ready to do whatever comes
[25:08] next. It it almost reads like it’s a
[25:10] setup for the next setup which makes me
[25:12] think about our first conversation where
[25:13] we talked about signposting. Is this how
[25:16] you see signposting today? To a degree I
[25:19] think maybe this is a slightly atypical
[25:20] example because we still want the next
[25:23] setup to be curiosity inducing and open
[25:26] a loop which is why you can see here
[25:27] it’s like I say here the first document
[25:29] I’m not going to say the first document
[25:31] is this. It’s the first document is
[25:33] going to help you fix thing that you’re
[25:35] curious about. But if I look at
[25:40] any other examples in here that that
[25:41] explain this? Well, I think this is
[25:43] probably a good example that that shows
[25:45] it off. I I wouldn’t say we’re
[25:48] necessarily signposting what comes next
[25:50] yet. Um I do think that is reserved for
[25:53] the following setup, but again, it’s
[25:56] like clarifying with the information you
[25:59] didn’t have that you now have that you
[26:00] learned in this segment.
[26:02] Let’s just take a pause to reflect on
[26:04] like what what is it that we have now
[26:07] unlocked the ability to do? Where is it
[26:09] that we now stand in the overall context
[26:11] of this topic? And I think this is a
[26:13] good example of this where after we’ve
[26:15] uploaded, this is I think the third
[26:16] document that I show people. We’re now
[26:18] just clarifying. Okay, take a pause.
[26:20] We’ve been saying all this stuff about
[26:21] these documents you have to upload, but
[26:23] practically speaking, let’s zoom out.
[26:25] What can your project now do so that you
[26:27] understand why we’re doing all of this?
[26:29] Yeah, tell me if this resonates with
[26:30] you. I think I’m following what you’re
[26:31] saying, but the other thing that I’m
[26:34] latching on to is some speaking advice
[26:36] I’ve gotten, which is tell them what
[26:37] you’re going to tell them. Tell them
[26:39] tell them what you told them. And this
[26:41] almost mirrors that a little bit in some
[26:43] of these cases, but maybe maybe the
[26:45] payoff isn’t like totally saying, “Let
[26:47] me close the loop on what I just told
[26:49] you, but it seems like maybe it could
[26:50] be.”
[26:51] >> Yeah. I This is a really
[26:54] kind of tricky semantic dichotomy.
[26:58] Sounds like the name of a band.
[27:02] Yes, I think I agree with you in
[27:03] principle, but I think we want to try
[27:04] and avoid it’s treading the line between
[27:07] that and not just making it a summary.
[27:08] So, yes, tell them what you told them,
[27:10] but I think I want to do it in a
[27:12] slightly different way. So, I’m not just
[27:14] repeating, so you just added this
[27:16] document. It’s more like um you added so
[27:19] having added this document, new
[27:21] information, this is now what your
[27:23] project is capable of thanks to what you
[27:25] just did. Um rather than just simply
[27:28] repeating it.
[27:29] >> Yeah. Does that distinction make sense?
[27:30] It’s kind of a tricky one.
[27:31] >> And that’s kind of what I’m trying to do
[27:32] with like these comparisons cuz when you
[27:34] tell me like yes, but it it clarifies
[27:36] more like what is a payoff then? What is
[27:38] a good payoff? It’s not a summary. It’s
[27:41] not a setup for the setup. Uh it’s
[27:43] putting a period on the end of the
[27:44] sentence. Yes. But it’s also it’s a
[27:47] breath in one way. It also seems like
[27:49] it’s making the person feel good and
[27:52] excited and accomplished in another way.
[27:55] Uh which obviously summaries don’t do.
[27:58] Yes. Exact. I think that’s a really a
[27:59] really good distinction. Um, make people
[28:01] feel good. Catch them up as it were.
[28:04] Like if anyone’s kind of gotten lost or
[28:05] they’re wondering, wait, have I have I
[28:07] grasped this? Just taking that moment to
[28:09] say, yes, you’ve got this and doesn’t it
[28:10] feel great? Um, as opposed to saying,
[28:13] here’s what you just learned. In much
[28:15] the same way, the AI in V1 is going to
[28:19] give us like, here’s how many sections I
[28:20] think we should have. Is the goal in
[28:22] your mind to reduce the number of
[28:25] sections
[28:27] necessary? How do I think about the
[28:29] number of sections that are in a video?
[28:31] I always come back to the the wise man
[28:34] Gandalf uh when I’m asked this question,
[28:36] which is that a wizard arrives precisely
[28:38] when he means to. Uh this is a tricky
[28:40] one. There are kind of no rules to this.
[28:42] I think a lot of it comes from just the
[28:43] soft skill of doing this over and over
[28:46] again and figuring out, okay, when I
[28:50] talk about X in my videos, my retention
[28:53] always drops off. On reflection, this
[28:55] period where my retention was kind of
[28:56] diving continuously, I realized that I
[28:59] was sort of talking about the same topic
[29:01] but in slightly different words twice in
[29:03] a row or something like that. I think
[29:05] really it just comes down to honest
[29:06] reflection and looking back at what
[29:07] you’ve written. Um, I I know one thing
[29:10] that can certainly help with this is
[29:11] just ju just showing it to someone else,
[29:13] whether it’s in your niche or even
[29:15] better sometimes someone who’s not in
[29:16] your niche. Someone who can look at it
[29:17] objectively with no um feeling that or
[29:21] no investment in the topic, no kind of
[29:23] preconceptions of what it’s about and
[29:25] just saying like is this interesting to
[29:26] to read? Did did it fall off at any
[29:29] point for you? When did you get bored?
[29:30] So outsourcing it is a is a good one.
[29:32] But I think overall, if we’re just
[29:35] talking you sat at your desk, I think
[29:37] having no preconceptions of how many
[29:40] segments a video should have uh is key.
[29:43] If you’re trying to target a certain
[29:45] length of your video, I know people like
[29:48] to try and have longer videos for the
[29:49] watch time, whatever. But if you’ve got
[29:51] an uh overly inflated video that doesn’t
[29:53] have any substance to it, then it’s it’s
[29:55] not going to be interesting. So, I would
[29:56] say a wizard arrives precisely when he
[29:58] means to. How long is a piece of string?
[29:59] Like, make it the number of segments
[30:00] that it has to be. be objective and look
[30:02] at am I including this segment because
[30:05] it contributes to the understanding of
[30:07] the subject or am I doing it just to
[30:09] bloat the watch time. Something that
[30:11] I’ve been trying to push myself to do
[30:12] more of is just slow down and feel
[30:16] better about every step of the process
[30:19] because that’s the input I can control
[30:21] that will probably lead to better
[30:22] outputs. But I get to this point where
[30:23] it’s like I’ll get really excited when I
[30:25] feel like I have the idea that I want
[30:27] and it’s like I’m just going to go rip
[30:28] that off and record that. and
[30:31] underinvesting in pre-production I think
[30:34] is like the the biggest tax I’m paying
[30:38] right now on the channel is is I’m just
[30:41] underinvesting in pre-production all the
[30:43] ways and you see that cost in the
[30:47] performance of a video. If a video
[30:49] doesn’t do well and you you think well
[30:51] like well I was proud of it and I did
[30:54] everything I can. that’s different than
[30:57] I should have taken more time on that
[30:58] because then you’re probably not going
[31:00] to go and just remake that same video
[31:01] the next week. It’s it’s just worth it’s
[31:05] worth feeling better in pre-production.
[31:07] One one just thought that came to me as
[31:09] you were saying that there is I think a
[31:11] a way of making sure that you’re still
[31:13] putting some of that thought and time
[31:14] into pre-production even if you’re short
[31:16] of time especially with AI now is even
[31:19] if you feel like you don’t know what you
[31:22] want your video to be and you don’t know
[31:23] how you want to structure it and you
[31:24] feel overwhelmed by the idea of it
[31:27] spieling to an AI and having it present
[31:28] a structure back to you will immediately
[31:32] set off red flags in your mind of things
[31:34] that it definitely got wrong. So if you
[31:36] look at a structure that that it has
[31:38] come back with I I I guarantee and maybe
[31:42] it’ll nail it but most of the time it
[31:44] will have come up with some idea that
[31:45] you’re like no no that’s totally not it
[31:46] and then you realize oh actually I do
[31:48] have some thoughts on this and I do have
[31:49] my opinions on this but having it
[31:51] presented to me in like slightly the
[31:52] wrong order is much easier then for me
[31:55] to think okay no I would shift segment
[31:56] one to segment two I would maybe add an
[31:58] extra segment about this I would connect
[31:59] these two ideas like this and then
[32:01] suddenly you’ve got your your structure
[32:03] out and it feels like you have invested
[32:04] that time in pre-production.
[32:06] Um, but you’re sort of spurred into it
[32:08] by seeing what was definitely not the
[32:10] right structure. Um, as opposed to just
[32:13] trying to do it from a blank page.
[32:14] >> Yeah. Can you go back to sharing your
[32:16] screen? I want to see you talk about
[32:17] your CTA now.
[32:18] >> Yes. Okay. So, um, in this video I
[32:21] didn’t have any product CTAs. I I think
[32:26] because I mentioned my AI script writing
[32:28] tool in the hook, I just put the first
[32:30] link or the second link as a link to to
[32:33] the weight list for this tool. That’s
[32:35] actually been a really effective call to
[32:36] action for me is is not mentioning it at
[32:38] all, but just showing the thing that I’m
[32:40] talking about the whole video and then
[32:42] putting a link in there um in the
[32:43] description. People intuitively just
[32:45] want to go and and learn more about it.
[32:46] >> And in that case then, is there no end
[32:50] besides like maybe an end card like you
[32:52] talked about in our first conversation
[32:54] of just like tying it to a next video or
[32:56] do you just let the payoff of the final
[32:58] segment be the the end of the video?
[33:00] >> No. So this is my absolute every single
[33:03] time without fail thing that I do is
[33:05] every video has to link to another one
[33:07] in a in a compelling way. But at the end
[33:10] of every single video I’m thinking about
[33:11] a link. So that is uh back to something
[33:15] in this video. So we’re trying to make
[33:18] it not apparent that the video is about
[33:19] to end. We’re basically saying we’ve
[33:21] already spoken about XY Z thing or now
[33:23] you know how to do blah blah blah blah
[33:24] blah. If you couldn’t see the timeline
[33:25] at the bottom, you might not know that
[33:27] the video is about to end. Then we want
[33:29] to add some uh curiosity. So um about
[33:34] something in the next video.
[33:37] Um so in this case here, if we go
[33:39] through these one by one, link back to
[33:40] something in this video. Now this
[33:41] project setup is a great starting point.
[33:43] Done. It’s like simple as that. That’s
[33:44] all you need. Um that shows that we are
[33:48] talking about we’re referring
[33:49] hell. Sorry. Grammarly on my
[33:52] >> I will leave when I decide.
[33:54] >> Yeah, dude.
[33:55] >> It’s okay. It’s okay. We can read it.
[33:57] It’s fine. I’m just removing you from
[33:59] Chrome. Grammarly, you are you are so
[34:00] gone. Um, no, it’s still here.
[34:07] This is ridiculous. Let’s see. I laughed
[34:09] so hard that I unplugged my own
[34:10] headphones.
[34:13] Okay, so
[34:16] if we take a look here, this is clearly
[34:18] the link. This is where we say, now this
[34:20] project setup is a great starting point.
[34:21] It’s just referencing something uh well,
[34:24] basically the entirety of the topic of
[34:26] this video.
[34:27] Then we raise the curiosity here where I
[34:29] say, “But what if I told you we can ask
[34:30] Claude to gradually improve itself over
[34:32] time?” So each new script you write gets
[34:34] closer and closer to your style every
[34:36] time. I think in the moment I must have
[34:37] rifted something else here because I say
[34:39] time twice in quick succession. Uh we
[34:41] want to avoid repetition. And then
[34:43] finally, we just have the promise uh
[34:46] which is usually spelled like this.
[34:47] There we are. uh CTA
[34:52] which uh is where we give the explicit
[34:54] promise and this is where the facade is
[34:56] kind of broken and we’re like we yeah
[34:57] we’re promoting another video go watch
[34:59] it but ultimately it’s an offer right
[35:01] we’re giving the viewer something else
[35:02] for free but watch this video next where
[35:04] I’ll walk you through how to do this
[35:05] using a simple prompt that you already
[35:07] have inside your claw product uh project
[35:09] so it’s like we’re promising something
[35:11] that is exciting for you and feels low
[35:13] lift so this is like your non-negotiable
[35:15] this ends every video I have a question
[35:16] on this point because I saw an
[35:19] interesting strategy and I’d love to get
[35:20] your take on this cuz I think there’s a
[35:22] cost to the strategy that you’re going
[35:24] to say as well, but I wonder if the
[35:25] benefit is is worth it. So, I saw a guy
[35:28] who had um an unlisted video on his
[35:33] channel. This is talking more. He had an
[35:35] unlisted video that’s essentially a
[35:37] video sales letter. And what he does is
[35:40] he ends some of his videos, not every
[35:42] video, but a video that like clearly
[35:44] showed some credibility. and somebody
[35:47] who watched the end of that video is
[35:48] probably a qualified lead for his
[35:51] business and his end card and his end of
[35:55] the video is linking to his unlisted
[35:58] essentially VSSL. What do you think
[36:00] about that strategy? I think that’s
[36:01] really cool. I I love that. I think like
[36:04] anything, it’s kind of balance and
[36:05] you’re trying to figure out what am I
[36:06] optimizing for mostly here. Um, the
[36:09] advantage of of this strategy and always
[36:10] ending with something else is that you
[36:12] can create these kind of infinite loops
[36:13] of people just watching more and more
[36:14] and more stuff. And it doesn’t matter
[36:16] which video they land on first, there’s
[36:18] always going to be kind of infinite
[36:20] trailer videos they could keep binging
[36:21] if they wanted to. I suppose the only
[36:23] risk with this strategy is, hey, if
[36:25] someone if it’s like the first video
[36:27] that someone’s landed on is that one and
[36:28] they get pitched something at the end,
[36:30] what is the effect that has on them? Do
[36:31] they think, oh, I’ve been sort of sold
[36:33] to here. Um, it was kind of an ask, not
[36:35] an offer. But ultimately, like I I think
[36:38] that sounds like a really a really good
[36:39] tactic and one that I might steal. Um
[36:41] and clearly you would think he would
[36:43] have been testing something like that,
[36:44] right? And the the pros would outweigh
[36:45] the cons.
[36:46] >> Yeah. And I imagine that same video
[36:47] could probably be like a good channel
[36:48] trailer. I know Ed from Film Booth has
[36:50] tried stuff like this where like his
[36:51] channel trailer is essentially a VSL. I
[36:54] think it’s really smart. I mean the
[36:55] biggest pain point we have on the
[36:56] channel and candidly why we’ve had to
[36:57] pull back some resources on the channel
[36:59] was we were spending far more than we
[37:01] were clearly earning from the channel
[37:04] because we never invested in connecting
[37:07] the channel to the core business and
[37:10] this is like a really consistent way to
[37:13] know that this is working if you have a
[37:15] video that you know when people watch
[37:17] this video they’re interested in
[37:18] learning more and going deeper with me
[37:19] and if that is true then I think this
[37:21] actually makes the YouTube as a
[37:23] discovery engine for the business a much
[37:25] shorter less uh luckfilled
[37:30] like uh process for getting people into
[37:33] the business. Yeah, I think um again
[37:35] that’s something you can really easily
[37:36] start using AI to help you with where
[37:38] it’s for something like this that has a
[37:40] formula that I try and use anything
[37:42] that’s formula based. If you’ve got
[37:43] something simple, we spoke about it with
[37:44] the hook as well. It’s very easy then to
[37:46] plug in the
[37:49] actual content that you want and to ask
[37:50] it to express that content in terms of
[37:52] that formula. So in your case, I suppose
[37:54] using a formula like that, but saying
[37:55] like, “Hey, actually, we don’t want to
[37:57] promote another video. We want to
[37:58] promote X or Y product. We want to
[37:59] promote the lab. We want to talk about
[38:01] whatever. We want to promote a VSSL that
[38:03] talks about the lab or something.” I
[38:04] think it would be fairly easy then for
[38:05] it to come up with a relatively subtle
[38:08] plug in the the style that we’ve just
[38:10] seen here, but for a VSSL. And it’s a
[38:13] spectrum because it doesn’t have to be
[38:14] like a straight up like sales video. It
[38:17] could be a video that you just know when
[38:19] people watch this video, a lot of them
[38:21] turn into customers. So, how can I feed
[38:23] more people into that video even beyond
[38:25] its own packaging and and CTR? I think
[38:27] is a a smart strategy. Intentionally
[38:30] antagonistic question. Okay.
[38:34] >> Yeah. Because you are a script writer
[38:37] and probably a lot of people who have
[38:38] historically been like, I could be a
[38:40] freelance script writer. I’m going to
[38:42] get into this. there is a certain amount
[38:44] of like incentive
[38:47] if not bias towards saying like we
[38:51] should keep human script writers. So if
[38:53] you were to take as much of an objective
[38:56] stance away from yourself as you can
[38:59] where would you say the human input the
[39:03] human in the loop in this process is
[39:05] most essential right now? That’s such a
[39:07] good question. I think to aid my case
[39:10] here, at least in terms of the
[39:11] objectivity, I’m in the fortunate
[39:13] position where I’ve managed to sort of
[39:15] work my way out of having to write
[39:16] scripts for people. And so I think I’m
[39:19] now able to grow my personal brand or
[39:22] whatever you want to call it as someone
[39:24] who helps people learn to do script
[39:25] writing in whatever context that is. And
[39:27] now that context is AI related. So, I
[39:29] think I’m able to divorce myself a
[39:32] little bit from the desire to like have
[39:33] script writers around because it’s what
[39:35] I used to do a lot and it’s um it’s what
[39:36] I still do for my own channel. One thing
[39:38] that I’ve learned a lot about since
[39:39] starting to really use AI a lot and
[39:41] using it to code my own script writing
[39:43] app, for example, is I’ve spent a lot of
[39:45] time watching how AI actually does what
[39:49] it does behind the scenes. Like you know
[39:51] in Claude for example you can sort of
[39:52] see how it’s thinking as it’s thinking
[39:53] and you can read its decision-m uh you
[39:56] sort of learn all these little ways in
[39:57] which it may appear to be doing
[39:59] something well and then the output that
[40:01] comes out it actually reveals oh hang on
[40:03] it wasn’t doing these like 10 things
[40:04] that it claimed very confidently that it
[40:05] was. So I think when it comes to script
[40:09] writing the final say on like when have
[40:11] I done enough revisions for this script
[40:13] when is this actually record ready that
[40:16] ultimate decision comes down to a
[40:18] person’s brain because AI will will
[40:20] always have something else that it’s
[40:21] like oh we could change this and oh you
[40:22] actually didn’t do this correctly and oh
[40:24] this could be stronger so we could make
[40:26] it land better or whatever and it will
[40:28] just keep doing that kind of infinitely
[40:30] and I think that moment where you have
[40:33] to say to yourself no actually I think
[40:34] this is ready I know my audience
[40:36] as wellroundedly as I possibly could.
[40:39] I’m the one who says that we’ve now done
[40:41] enough revisions on this and I think
[40:42] it’s going to it’s going to hit really
[40:43] well. But I think there are those
[40:45] moments that come throughout the
[40:46] process. So it might be when structuring
[40:47] out the script in the first place. AI
[40:49] will be like, hey, here’s a brilliant
[40:51] structure based on your audience avatar
[40:52] that I think they’re going to love. And
[40:54] you can read that and be like, I like
[40:56] 70% of these ideas, but ultimately I
[40:58] want to rearrange X and Y because I know
[41:00] this will be how I would actually
[41:01] explain it or how they would receive it
[41:03] better. And so again, the point of
[41:05] deciding structuring is now done. That
[41:07] kind of needs to be me because AI will
[41:09] keep on coming up with new things that
[41:11] you should add or change or improve.
[41:13] there is a certain courage to
[41:16] cut the AI off, you know, and I think
[41:19] that’s that’s kind of a meta-kill
[41:21] through all AI usage that people need is
[41:25] courage and trust in your own
[41:28] intuition, your own taste, your own gut
[41:32] to know when
[41:34] you’re getting into like the recursive
[41:35] improvement for improvement sake that is
[41:38] no longer actually improvement. Maybe
[41:41] this is related, maybe it’s not, but
[41:42] it’s an aside I want to take. I I have a
[41:45] similar video that I want to make, but
[41:47] it’s it’s kind of talking about like,
[41:48] hey, I’m about to have a second kid, and
[41:51] the first time I did this, I didn’t
[41:53] prepare well, and so I’m preparing
[41:54] differently this time. And where I kind
[41:56] of get hung up is I don’t know if
[42:00] there’s enough there. There when you’re
[42:02] kind of laying this out through I’m
[42:04] imagine you’re using like whisper flow
[42:06] or some sort of dictation. When you’re
[42:08] laying this out, when do you feel
[42:10] confident that you’ve got a video that’s
[42:13] long enough to warrant existing? Oo,
[42:16] that’s a really good question. I think
[42:18] with a lot of anything creative, there’s
[42:21] an element of brutality that needs to
[42:23] come that comes down to it in the end.
[42:26] There have been some scripts that I have
[42:27] written which have seemed like a great
[42:29] idea have been maybe like five segments
[42:32] or kind of subsections initially and
[42:34] then as we’ve kind of workshopped either
[42:36] me personally or me with in
[42:37] collaboration with an AI have
[42:39] workshopped it it’s kind of trimmed down
[42:40] and down to maybe the point where
[42:42] there’s like three segments and it
[42:43] reaches a point where I’m like I think
[42:45] this could just be a newsletter or
[42:46] something you know that’s a bit more
[42:47] short form and so I think the first
[42:49] thing would be to say accepting the
[42:51] reality that it may be that there just
[42:54] isn’t enough to the topic but I I think
[42:56] this is one of the real meta skills to
[42:59] to to learn and it’s something I’m sure
[43:01] you’re you’re aware of anyway is like
[43:03] just the ability to talk to AI to
[43:06] extract from it exactly what you want.
[43:08] So talking to it in natural language and
[43:09] say look here’s the predicament I’m in.
[43:11] I have this exact topic I want to talk
[43:13] about. Here’s the context on my last
[43:14] video about having a kid. Uh this one I
[43:16] want to do now is about how I wasn’t
[43:17] prepared properly but I’m worried there
[43:19] isn’t going to be enough to talk about
[43:21] it. Knowing what you know about my
[43:23] audience, if you don’t have an audience
[43:25] avatar, like here are 10 transcripts of
[43:27] some videos we did recently. Here are
[43:29] some recent comments from them. Knowing
[43:30] what you know about them, what blind
[43:33] spots do I have that could relate to
[43:36] this topic that we could put into this
[43:38] video? And kind of using it then to
[43:39] challenge you and and to almost make you
[43:41] answer some questions you might not have
[43:43] even thought to ask yourself that can
[43:44] then contribute to uh making a video
[43:47] that feels like it warrants existence.
[43:49] Uh to kind of close us out here, what’s
[43:51] your relationship to AI right now?
[43:53] Obviously, you’re embracing it for
[43:54] making a tool and you’re using it in the
[43:55] script writing process. Are you just
[43:57] like full in like, “Yep, toothpaste out
[43:59] of the tube. I’m using it. I love it.”
[44:01] How do you feel about AI? It It’s an
[44:03] interesting moment, isn’t it? Because
[44:05] it’s quite It becomes harder and harder
[44:07] to extricate yourself from it and to
[44:10] reason with yourself and think as you
[44:13] would have thought before what’s now
[44:15] possible
[44:16] were possible. my business didn’t rely
[44:19] on me moving with the times like this,
[44:22] would I be going as all in on it as I
[44:24] am? I don’t know. Uh I’m certainly
[44:27] loving building an app. This this is
[44:29] something that I had spoken to different
[44:31] companies about in the last few years
[44:33] that I sort of being brought on as like
[44:35] a
[44:38] as the sort of the the brains as it were
[44:41] as the script writer behind a tool that
[44:43] would then be coded by a bunch of very
[44:45] intelligent people. the fact that those
[44:47] very intelligent people are now not
[44:49] strictly needed. Although I’m sure
[44:50] anyway there’s a whole discussion there,
[44:51] but it’s I I love that I’ve been able to
[44:53] build my own thing. This is the tool
[44:55] that I’ve dreamt of making and it does
[44:56] all the things that I’ve always wanted.
[44:57] That’s very cool.
[45:00] Do I feel the most creatively fulfilled
[45:02] that I have uh in the four or five years
[45:06] I’ve been running this business?
[45:08] In some ways, no. And as time goes on,
[45:12] the more I kind of want to shut it all
[45:14] down, go live in the woods, and makeong
[45:16] YouTube videos just chopping wood. This
[45:18] is exactly how I feel. I feel like the
[45:20] more I embrace AI, the less I enjoy my
[45:22] work. And I like I myself start to feel
[45:25] like a factory rather than any type of
[45:28] artisan. like anything that I’ve used AI
[45:32] to do to get to the point of done and
[45:34] skipped the doing
[45:37] >> I just have no affinity or pride in that
[45:42] work. Uh not to say it’s bad but to say
[45:44] like I just don’t feel pride in it
[45:47] because I didn’t do it.
[45:48] >> This is my biggest thing is like
[45:50] >> okay this can help me do a lot of things
[45:52] that I have been doing but
[45:55] to what end? I’ve got to find the things
[45:59] that are absolutely non-negotiable that
[46:01] I enjoy taking pride in that I want to
[46:02] do. And for me, writing like the essays
[46:06] has been part of that. Writing the book
[46:08] is part of that. YouTube scripts, like
[46:10] we we had a video we did and I just I
[46:13] don’t get excited about writing the
[46:15] intros for our podcast episodes. And so
[46:17] I was like, here my AI agent, here are
[46:20] my top 10 performing videos. Learn what
[46:23] you can about the structure. Here’s the
[46:25] transcript of the rest of the edit. Help
[46:28] me make an intro for this. And it did as
[46:32] good of a job as I would have given my
[46:35] current skill set and the time I would
[46:37] put into it. Did I then show that off
[46:39] and say, “Look at what I made uh for
[46:41] that intro?” Absolutely not. Um so it’s
[46:45] I’m I’m deeply conflicted. Yeah, I think
[46:47] I I’m completely with you. It’s hard and
[46:50] I don’t know if you feel this especially
[46:51] from a business point of view where I’m
[46:54] I’m trying to build an AI related
[46:57] business that is
[46:59] that that is um that doesn’t completely
[47:02] remove the human element. I don’t want
[47:04] to build a script writing tool that
[47:05] writes every single word for you and you
[47:08] go away and you record something that
[47:09] you just didn’t you had no nothing to do
[47:11] with. which is why the sort of core
[47:14] feature of it is this discursive element
[47:15] that we’ve discussed where it’s like
[47:17] there’s a there’s a chat you chat to it
[47:19] you leave your voice notes and you kind
[47:21] of work with them as if they were a
[47:23] script writing a script writer that
[47:24] you’d hired but at the same time I’m
[47:26] very conscious that this tool also would
[47:29] completely serve people who want to
[47:31] start a YouTube channel as a cash cow
[47:35] faceless no shade on faceless in general
[47:37] but like one of those channels where
[47:38] it’s like hey I don’t have to do
[47:39] anything I can AI the voiceover this
[47:41] guy’s got this tool out and it can just
[47:43] AI the script and I can AI the
[47:46] thumbnail. And so being being conscious
[47:47] that you’re kind of you’re serving both
[47:49] the more conscious people who want to
[47:52] keep the human element involved but also
[47:54] the people who want to just outsource it
[47:56] all. And yeah, that is kind of
[47:58] conflicting. It’s a weird a weird
[48:00] moment.
[48:00] >> Yeah, it’s tough. I think um I think
[48:03] people want to maximize their outcome
[48:07] and minimize their effort and you follow
[48:10] that all the way down and people want
[48:12] everything just like fully out of the
[48:13] box done. When you actually watch their
[48:15] behavior, that’s what they want most of
[48:16] the time, most people. So, it’s yeah,
[48:19] it’s it’s challenging and it’s I
[48:22] restrain myself from the temptation to
[48:24] become like an AI educator because
[48:27] there’s so much tailwind to it and it’s
[48:29] what people want, but it’s just not what
[48:31] I want to do. It’s an interesting time,
[48:34] but I appreciate you uh sharing with us.
[48:36] We’ll put um a link to the the document
[48:39] you recommended and this tool that you
[48:42] showed us in the description. George,
[48:44] thanks for coming back on the