Copy This System, It’ll Blow Up Your Social Media — Transcript

Full transcript

1. 0:00Today we're talking about content
2. 0:01systems. If you want to learn how to
3. 0:03make great content faster [music]
4. 0:05without sacrificing originality, this is
5. 0:07going to be one of the best workflows
6. 0:08you ever learn. Now, I know this works
7. 0:10because this is the exact same system
8. 0:12I'm personally running every single day.
9. 0:14I have a million followers. I've done
10. 0:15billions of views. And this month alone,
11. 0:17I'll post over 50 unique short form
12. 0:20videos on Instagram alone. And these
13. 0:22aren't clips or AI slop. These are
14. 0:24native premium short form videos that'll
15. 0:26get tens of millions of views and drive
16. 0:28thousands of leads. The only way that
17. 0:30much premium volume is possible is by
18. 0:32building this type of content system.
19. 0:34Now, I'll be honest, the truth is, as
20. 0:35weird as it sounds, I've never really
21. 0:37liked making content. And the reason why
22. 0:39is because there's just so much
23. 0:41repetitive work that goes into every
24. 0:43single rep. It's just draining to do
25. 0:45every day. And so, what I needed was a
26. 0:46system that allowed me to use AI to
27. 0:49automate the boring parts so that I
28. 0:51could free up more of my time to be
29. 0:53creative on the fun parts. And that's
30. 0:55exactly what I built. So, in this video,
31. 0:56I'm going to take you through my entire
32. 0:58AI powered content system. First, we'll
33. 1:00go through one rep end to end across my
34. 1:03six-step workflow, and then at the end
35. 1:05of the video, I'll take you through my
36. 1:0610X badge system for how you grow as
37. 1:08fast as possible on any platform. Now,
38. 1:11the first part I want to focus on is my
39. 1:12endto-end workflow. And I think this
40. 1:14will be the most valuable 10 to 15
41. 1:16minutes you watch on content for the
42. 1:18rest of this year. To make one video, I
43. 1:20go through the same six stages every
44. 1:22single time. topic, format, substance,
45. 1:25hook, script, and [music] edit. Now, in
46. 1:28every one of those stages, I'm able to
47. 1:29use AI to speed up the process without
48. 1:32sacrificing creativity or originality.
49. 1:35So, here's how the process works in
50. 1:36detail. Step one is always topic. The
51. 1:39very first thing I do is figure out what
52. 1:40topic I'm going to make the video about,
53. 1:42cuz the truth is nothing else really
54. 1:43matters if you pick a losing topic. Now,
55. 1:45for topic, this first step, I lean
56. 1:47completely on data and AI. My creativity
57. 1:49in this workflow comes a little bit
58. 1:51later in substance, hook, and scripting
59. 1:53and sometimes the edit, but I don't need
60. 1:54to reinvent the wheel on the topic
61. 1:56front. You can just use data to get the
62. 1:58best topics. And the reason why is
63. 2:00because of the psychology of the viewer
64. 2:02and buyer. In every niche, there are
65. 2:04certain topics that always outperform
66. 2:06others, and those map to whatever the
67. 2:08categories or topics that those viewers
68. 2:10think are more important. Now, to use
69. 2:12data and build a workflow for solving
70. 2:14topics, I use Sandcastles.ai plus
71. 2:16Claude. That's the money combo and it
72. 2:18crushes everything else. The first thing
73. 2:20I do is I go to Sandcastles and I build
74. 2:22a watch list of all the top performing
75. 2:24channels in my niche. Now, you may
76. 2:26already know who these people are, but
77. 2:27even if you don't, you can just click
78. 2:29describe and write in the type of
79. 2:31content you make and Sand Castles will
80. 2:33suggest all the best performing channels
81. 2:35for you to track. I like to get anywhere
82. 2:36from 20 to 30 channels in my watch list
83. 2:39before I move to the next step. Now, for
84. 2:40me, the sweet spot when I'm adding those
85. 2:42channels is the micro, small, and medium
86. 2:44account size range between 10,000
87. 2:46followers and a million. You don't want
88. 2:47to go any bigger than that. By the way,
89. 2:49the reason I'm building out this channel
90. 2:50list instead of just scrolling
91. 2:51mindlessly is because I'm trying to
92. 2:53speed this up. I don't want any noise.
93. 2:55When you scroll, you get so much noise.
94. 2:57Sometimes you find good videos, but a
95. 2:59lot of noise. This reduces all the
96. 3:00noise, so it's only signal. Okay, once I
97. 3:02have that channel list made, there's two
98. 3:03more things I'm going to do in Sand
99. 3:04Castles before I go to Claude. The first
100. 3:06thing is I'm going to go to the videos
101. 3:07tab. I'm going to go sort and I'm going
102. 3:09to sort by outlier score. Then I'm going
103. 3:11to set the posted date to either three
104. 3:13or six months depending on how far you
105. 3:15want to look back. And I'm going to set
106. 3:16the minimum bound on engagement rate to
107. 3:182%. This will give you the best videos
108. 3:21from that channel set in the relative
109. 3:23range you want to analyze. Now, this may
110. 3:25reveal only 15 videos. It might reveal
111. 3:27500 videos. It depends on how many
112. 3:29channels you have in your list and how
113. 3:30many bangers those channels have made.
114. 3:32The next step is to deep analyze every
115. 3:35single one of these videos. And you can
116. 3:36do it by hovering over any of them and
117. 3:38pressing analyze. Or you can bulk
118. 3:40analyze by pressing the bulk button and
119. 3:42then specifying the number of videos you
120. 3:44want to do. Personally, I like doing a
121. 3:45hundred at a time. You can do less than
122. 3:47100, but the more you do, the better
123. 3:49those trends will be when you're
124. 3:51analyzing it in aggregate. So, the more
125. 3:53the better. Now, when you deep analyze
126. 3:55and you click in, you'll realize you get
127. 3:56everything you'd ever want on the video.
128. 3:58The transcript information about the
129. 4:00topic, the idea, the hook format, the
130. 4:02exact hook, the storytelling format,
131. 4:04literally everything. And the best part
132. 4:06of Sandcastle is now you can export all
133. 4:08that information across all the top 100
134. 4:10videos out into a CSV. And we're going
135. 4:12to upload that CSV into Claude in a
136. 4:14minute. Now, before we go to Claude,
137. 4:16there's one other thing I like to do in
138. 4:17Sand Castles to set myself up for the
139. 4:19next round of doing this in the future.
140. 4:21I go to the automations tab and I write
141. 4:23an automation rule. All I do is put five
142. 4:26in the outlier score and then 2% in
143. 4:28engagement. So, what that means is
144. 4:29anytime any of my channels on my watch
145. 4:31list post a video above a 5x outlier in
146. 4:34the future, this automation will
147. 4:36automatically process it for me. So,
148. 4:38when I come back into Sandcastles, it's
149. 4:40just ready to go and I don't need to go
150. 4:41through that bulk analysis process. It
151. 4:43just basically autodetects outliers for
152. 4:45you. So, I like to set that and then
153. 4:47move forward. All right, back to the
154. 4:48main flow. I've got my CSV of the top
155. 4:50100 videos. I'm going to open Claude and
156. 4:52I'm going to drag that CSV into Claude.
157. 4:55And when I use Claude, I use the desktop
158. 4:57app and I'm using Co-work. You can use
159. 4:59chat, co-work, or claude code. In the
160. 5:01browser, which is what most people do,
161. 5:02they just type in claude.ai. They can
162. 5:04only use chat. So co-work is way more
163. 5:06powerful. You should be using co-work or
164. 5:08claude code for this. Now, I'm going to
165. 5:10go through the exact prompt that I feed
166. 5:12into Claude to do this topic research
167. 5:13right now. But by the way, all these
168. 5:15prompts I mentioned in this video,
169. 5:16there'll be like four or five prompts
170. 5:17throughout, I've got a doc below. It's
171. 5:19completely free where I've written down
172. 5:21all those prompts. You can just copy
173. 5:23paste. So you don't have to worry about
174. 5:24like scrambling and copying what I'm
175. 5:25saying word for word. But I'm going to
176. 5:26say it word for word on this first one
177. 5:28so that you can get a sense of what
178. 5:29we're doing. All right. So this is the
179. 5:30prompt that I'm going to run in Claude
180. 5:31right away when I first get the CSV in.
181. 5:34I'm uploading a report of the top 100
182. 5:35videos in my niche over the last X
183. 5:37months. Whatever you put in for post
184. 5:39date, 3 months, 6 months, whatever it
185. 5:40is. These were posted by the top
186. 5:41creators in my niche. Now, if you
187. 5:43haven't set up the background, you're
188. 5:44going to use this background part as
189. 5:45well. For background, I make content in
190. 5:47Xnich. For me, it's like social media
191. 5:48marketing, social media growth with the
192. 5:50goal of helping build an audience of Y
193. 5:52avatar. For me, it's business owners
194. 5:54that are trying to grow on social media.
195. 5:56Specifically, my target authority
196. 5:57statement is I help X people with Y
197. 5:59thing to accomplish Z goal or to get Z
198. 6:02outcome. So, for me, it's I help
199. 6:03business owners grow faster on social
200. 6:05media so they can build their business.
201. 6:07First, I want you to go through the list
202. 6:08and screen out any videos that aren't on
203. 6:11the topic that I'd be covering. So, if
204. 6:13there's any random trend or personal
205. 6:15story from the creator that isn't on
206. 6:17topic, get rid of those. With the videos
207. 6:18you have left, run two analyses for me.
208. 6:21First, I want you to list all the videos
209. 6:23in order by outlier score. I want you to
210. 6:25give me the topic, the link to the
211. 6:27original video, and a one to twoline
212. 6:29sentence pitching me on why you think
213. 6:31this would be a good video to make given
214. 6:32my target authority statement. The topic
215. 6:34should not be at the category level.
216. 6:36They should be a three to five word
217. 6:38framing that summarizes what the video
218. 6:40is about specifically. Some of these may
219. 6:42overlap, but I'd expect dozens and
220. 6:44dozens of unique topics. Now, second, I
221. 6:46want you to run another analysis
222. 6:47clustering the videos by topic category.
223. 6:50There's no specific set number of topic
224. 6:51buckets I'm going for, but I want clear
225. 6:53delineation and break points between.
226. 6:55I'd expect somewhere between 8 to 15
227. 6:57different categories. I want you to rank
228. 6:59these buckets in order based on highest
229. 7:02average outlier score from the videos
230. 7:04within them. Within each bucket, I also
231. 7:06want you to list the individual videos.
232. 7:07I want you to give me the topic. I want
233. 7:09you to give me the link to the original
234. 7:10and then again the one to tweline
235. 7:12sentence explaining why this would be a
236. 7:13good video for me to make. Now, once you
237. 7:15have that, just run it. It will take
238. 7:16about 4 to 5 minutes to finish. And when
239. 7:18it finishes, this will give you the best
240. 7:20datadriven list of topics for your niche
241. 7:23without any scrolling required. Now, my
242. 7:25process is if I don't have an original
243. 7:27idea I want to make, I'm always starting
244. 7:30with this list. And it doesn't mean I
245. 7:31make every single one of those, but I
246. 7:33start with the top one and I ask myself
247. 7:35when I get to the substance, do I have
248. 7:37something unique to say? If I don't, I
249. 7:39pass. But I have the right to make any
250. 7:41of them if I want. This is how you use
251. 7:43data to get a curated list of topics
252. 7:45that you already know worked. And the
253. 7:46reason this looks so easy is because of
254. 7:48that Sandcastle's data. Having the
255. 7:49transcript and all that detailed
256. 7:51information about every video, you'll
257. 7:52see how powerful it gets with that one
258. 7:54report. We can basically run this full
259. 7:56flow. And if you're a social media
260. 7:57manager or you do this for many clients,
261. 7:59you just save yourself literally dozens
262. 8:01of hours a week because of how many
263. 8:02times you have to do this. Okay, so
264. 8:03that's step one, validated topic curated
265. 8:05by data. Now, step two in the flow is
266. 8:08the format. And this step is extremely
267. 8:09quick, but I will not move to step
268. 8:11three, the substance, until I've decided
269. 8:13what format I'm making. Now, the format
270. 8:15is usually a combination of the
271. 8:16structure of the video, the way the
272. 8:18points are laid out, and sometimes the
273. 8:20visual execution of how it's edited. All
274. 8:22three of those things together kind of
275. 8:23combine to make the format. So, for
276. 8:25example, a breakdown format is actually
277. 8:27very flexible. It doesn't lock you into
278. 8:29a certain narrative or story type. But
279. 8:31an S tier format, on the other hand, one
280. 8:33where you're like ranking items across
281. 8:34the S, a B, C, D, F tier, that kind of
282. 8:38dictates how you have to flow the story.
283. 8:40Now, with formats, there are 20 or so
284. 8:42buckets that typically dominate on
285. 8:44social media. Most creators end up with
286. 8:46like two to three hero formats that they
287. 8:48alternate between and then sometimes
288. 8:50they experiment on the edges trying to
289. 8:52see if one of the new ones can beat out
290. 8:53their core three formats. When you're
291. 8:55starting out, people typically
292. 8:56experiment till they find their core
293. 8:58ones. Now, I've built a visual database
294. 9:00of the top 20 formats with a bunch of
295. 9:02examples of each one. You can see that
296. 9:03here. I'll link it below for free if you
297. 9:05want to download and access it. But
298. 9:07here's how I approach formats in this
299. 9:08system. The first way is to just look at
300. 9:10that list of 20 and just pick one that
301. 9:12you think looks interesting and take a
302. 9:14shot, right? So, not really using data,
303. 9:16just using intuition. The second path is
304. 9:18to let the data tell you what formats
305. 9:20are also working best for your niche.
306. 9:23And of course, the Sandcastle's data CSV
307. 9:25that also has a column with all the
308. 9:27formats tagged. So, you can just again
309. 9:28go to Claude and ask it to break out the
310. 9:31best performing formats for you. And so,
311. 9:33this is exactly what I'd write in
312. 9:34Claude. Give me a breakout of all the
313. 9:36videos by storytelling format. rank them
314. 9:39based on the highest outlier score and
315. 9:40then list within each bucket all of the
316. 9:43individual links and the names of those
317. 9:45formats. Now, typically this analysis
318. 9:47very quickly will reveal the top
319. 9:49performing formats in your niche. And if
320. 9:50you're unsure of which format to try,
321. 9:52you really don't care about creativity.
322. 9:53You just want the best performing one.
323. 9:55Pick the top one on this list. For
324. 9:56example, the rating ranking and tier
325. 9:58list format that we have bucketed. You
326. 10:00could do ratings, you could do rankings,
327. 10:01you could do tier list, you could do
328. 10:02brackets. There's all these different
329. 10:03flavors within that bucket. So, what I
330. 10:06like to do is take that datadriven
331. 10:08analysis that Claude did and it has all
332. 10:09the unique URLs and then I'll just go
333. 10:11and open a bunch of the top ones within
334. 10:13a bucket I'm interested in and I'll just
335. 10:15see how the people actually executed it
336. 10:17visually. Once I find a unique flavor
337. 10:19that I like and think I can replicate,
338. 10:21well, then I'll take that as the model.
339. 10:23But easily that uses data to guide me to
340. 10:25the top performing ones and then my eyes
341. 10:27to kind of lock in on the flavor that I
342. 10:29want. Also, if you like scrolling to
343. 10:31discover new formats, you can always
344. 10:33scroll on Instagram and Tik Tok and just
345. 10:34copy the link to a video. And you can
346. 10:36actually paste it in the videos tab in
347. 10:38Sandcastles by pressing add video URL.
348. 10:41When you do that, it'll import that
349. 10:42one-off video in. And then when you
350. 10:44export your CSV, that will be included
351. 10:46in this analysis. Like I said, I won't
352. 10:48move forward to the next step until I
353. 10:49have my topic and format locked in based
354. 10:52on the Sand Castle's data because this
355. 10:54makes sure that I'm rowing in the right
356. 10:55direction. By doing it this way, I've
357. 10:57now saved myself a ton of time. I didn't
358. 10:58have to doom scroll and I've d-risked
359. 11:01because I'm using the topics and formats
360. 11:03that have already been proven to work.
361. 11:04Now, obviously, if your strategy or
362. 11:06niche requires more realtime stuff,
363. 11:08you're reacting to the news, so you need
364. 11:10like the timeliest stuff, then just
365. 11:11change the post date in Sandcastles from
366. 11:133 months down to like 2, 3, or 7 days.
367. 11:16This will give you more real time.
368. 11:18There'll be less videos in there, but
369. 11:19these are like the newest bangers that
370. 11:21are coming out. The truth is, most
371. 11:22niches are evergreen. You could take a
372. 11:24video from four months ago and just
373. 11:25remix it to be relevant today, and it
374. 11:27will still crush. So, most people don't
375. 11:29need to do that, but if you're doing
376. 11:30this news reaction stuff, then yeah,
377. 11:31change the post date. By the way,
378. 11:32depending on when you're watching this,
379. 11:33on the Sand Castle side, we have
380. 11:35something brand new coming out that's
381. 11:36not in this video that will make what I
382. 11:39just walked through like a hundred times
383. 11:40easier. So, stay on the lookout for that
384. 11:42if you like this AIdriven workflow.
385. 11:44Essentially, if you're using sand
386. 11:45castles to make content, you're going to
387. 11:46be able to outrun anyone in your niche
388. 11:48that isn't. We're just making it
389. 11:49automatic. All right, the third step in
390. 11:51my process is the substance. And this
391. 11:53really is the sauce layer. This is where
392. 11:55all the creativity lives. This is the
393. 11:57part where you don't really use the data
394. 11:58for. You kind of go authentic, original,
395. 12:01whatever word you want to describe to
396. 12:03encompass your uniqueness. It's in the
397. 12:05substance. The substance really boils
398. 12:07down to two things. The first is your
399. 12:09contrarian take or non-obvious
400. 12:11perspective on the topic. And the second
401. 12:13is the evidence you give to support that
402. 12:16contrarian take. Things like facts or
403. 12:18case studies or stories or metaphors or
404. 12:20examples, things like that. Now, here's
405. 12:21how I run the substance part of the
406. 12:23process. First, I look at the topic and
407. 12:25I make sure it's specific enough to
408. 12:27react to, right? So, if you're looking
409. 12:28at the topic category, the broadest
410. 12:30thing like hooks or storytelling or Nike
411. 12:33or Apple, that's way too broad. So, you
412. 12:35need the actual topic. That would be the
413. 12:36category. You want the subtopic or
414. 12:38topic. And if you use the prompt that I
415. 12:40gave you and you do the sand castle to
416. 12:41claim, you're just going to be given the
417. 12:42topic, so you'll be all good. But if you
418. 12:44do it some other way, just make sure
419. 12:45you're looking at the topic level unique
420. 12:47enough that you could actually react to
421. 12:49it. So, first I start with the
422. 12:50contrarian take or non-obvious takeaway.
423. 12:52And this really is what makes great
424. 12:53videos. This piece right here, the
425. 12:55contrarian take is basically something
426. 12:57you believe to be true about the topic
427. 12:59that most people do not. But then once
428. 13:01you reveal it, they could quickly get on
429. 13:03board to think that it's true as well.
430. 13:05So for example, let's just say the topic
431. 13:07category that surfaced for me was hooks.
432. 13:09And then the specific topic I was
433. 13:10working on was how to write better
434. 13:12hooks. That was like validated by the
435. 13:14data and that's what I'm working on. If
436. 13:15I'm trying to come up with a contrarian
437. 13:17take, I would then ask myself, what do I
438. 13:19believe about how to write better hooks
439. 13:20that I strongly think is true that most
440. 13:23people would be shocked to hear or not
441. 13:25believe today? So, I would write down
442. 13:26all the things that I know to be true
443. 13:28about how to write better hooks. And
444. 13:29that might include things like the first
445. 13:31sentence has to give context and have
446. 13:33zero fluff. So, like no delay. The first
447. 13:35sentence needs to be super punchy and
448. 13:37less than 12 words. The first sentence
449. 13:39needs to be written in fifth grade
450. 13:40language with really simple vocabulary.
451. 13:42But then maybe I think, wait a second,
452. 13:44actually the most valuable thing I could
453. 13:46tell them about writing a killer hook
454. 13:47actually has nothing to do with the
455. 13:48spoken hook. It's actually that the text
456. 13:50hook is 10 times more valuable and that
457. 13:52the text on screen is where your eyes go
458. 13:54first and then bingo. That epiphany
459. 13:56moment right there, that realization by
460. 13:58me is like actually the interesting part
461. 13:59is something over here that they weren't
462. 14:01thinking about. That's a contrarian
463. 14:02take. That's a non-obvious insight.
464. 14:04That's where the gold comes from in
465. 14:06content. So, the genius of being good at
466. 14:08content is really how easily you can
467. 14:09come up with these non-obvious
468. 14:11contrarian insights about whatever topic
469. 14:13you're making. And most of those come
470. 14:15really easy when you actually have legit
471. 14:17experience or expertise in an area.
472. 14:20That's why if you ever follow someone
473. 14:21and watch their videos, but the things
474. 14:22they say are always just replications
475. 14:24from someone else that you've already
476. 14:26heard, well, don't ever pay that person
477. 14:27for services because they don't actually
478. 14:29know what they're talking about, right?
479. 14:30They're not coming up with non-obvious
480. 14:31things, meaning they don't have
481. 14:33expertise outside the box in this area.
482. 14:35Now, once you figure out what the
483. 14:37contrarian take is in this substance
484. 14:39bucket, the rest of this process for
485. 14:41substance actually becomes a lot more
486. 14:43straightforward. The other part of the
487. 14:44substance step is the evidence. And when
488. 14:46I say evidence, what I mean is all the
489. 14:48other stuff you're going to say in the
490. 14:50video other than the contrarian take.
491. 14:52Now, this evidence, just like with a
492. 14:54lawyer, it's meant to support or prove
493. 14:56the contrarian take to be true. Because
494. 14:58once you say something that people don't
495. 15:00believe, you only have a limited amount
496. 15:01of time to convince them that they
497. 15:03should give you a chance and believe it.
498. 15:05So this evidence could be facts,
499. 15:07metaphors, stories, case studies,
500. 15:09examples, could even be opinions,
501. 15:11something to prove your take to be true.
502. 15:13So in the case of my hooks example where
503. 15:15I have this like non-obvious take around
504. 15:17the title text, I would then ask myself,
505. 15:19okay, what evidence could I show to
506. 15:21prove that my claim of title text
507. 15:23matters way more than spoken hook
508. 15:25actually is right? Well, one thing I
509. 15:27could show is proof. I could just show a
510. 15:28bunch of videos having title text with
511. 15:30no spoken hook with a bunch of views.
512. 15:32That could work. Another thing I could
513. 15:33do is show A versus B. one video
514. 15:35flopping and one crushing in the exact
515. 15:36same hook, but one has title text and
516. 15:38one doesn't. That could work. I could
517. 15:40also explain the psychology of why the
518. 15:42brain sees words faster than it hears.
519. 15:44That's more of like a theoretical
520. 15:45scientific explanation, but that could
521. 15:47work. Or I could do a combination of all
522. 15:48those things, right? So, I'm going
523. 15:50through this process of being like, what
524. 15:51evidence could I manufacture or have to
525. 15:54support my take? And that's the process
526. 15:55of putting together a winning video.
527. 15:57Now, the question is, where can you use
528. 15:58AI in this process to speed that flow
529. 16:01up? I'll say this, coming up with a
530. 16:02contrarian take using AI in like one
531. 16:04shot off one prompt. Coming up with it
532. 16:06is pretty tough. You could ask AI to
533. 16:08come up with contrarian takes on the
534. 16:09topic, but I find AI is good at
535. 16:11regurgitating what it's been trained on
536. 16:13and what it's heard. It's not great at
537. 16:14coming up with novel things outside the
538. 16:16box. However, once you do come up with a
539. 16:18contrarian take, you could use AI to be
540. 16:20like, "Hey, here's my topic. Here's my
541. 16:22format. Here's my contrarian take. What
542. 16:24evidence could make sense to use to
543. 16:26start supporting this take to be true?"
544. 16:28That it's good at. And so this is kind
545. 16:30of like the research on the take that
546. 16:32you might be able to do. Sometimes if
547. 16:33I'm feeling lazy and I don't want to
548. 16:35think for 30 minutes trying to come up
549. 16:36with the research, I'll just throw it
550. 16:38into Claude and be like, "Give me some
551. 16:39ideas for what I could use." That will
552. 16:41start spinning the wheels and then the
553. 16:42script will come out tighter. So if I'm
554. 16:44doing that, I would say something in
555. 16:45Claude like this. Hey, I'm making a
556. 16:46video about this topic. My contrarian
557. 16:48take is this. Can you help me think
558. 16:49through what the evidence or research
559. 16:52would need to be for this claim to make
560. 16:54sense? So in that case, I'm kind of
561. 16:56using Claude like a brainstorming
562. 16:57research partner, per se. Now, no matter
563. 16:59which way you do it, this substance
564. 17:01piece, the contrarian take and the
565. 17:02evidence supporting it, that really is
566. 17:04the foundational block for what great
567. 17:06content comes from. This is what I call
568. 17:08the last mile of great ideas. Now,
569. 17:10there's one more point I want to mention
570. 17:11on this substance piece, and that has to
571. 17:12do with the format you chose. If you
572. 17:15have a fairly restrictive format, that
573. 17:16can kind of dictate the flow of how the
574. 17:19evidence can be presented and what that
575. 17:20contrarian take could even be in the
576. 17:22first place. For example, if I'm making
577. 17:23a video about hooks, but I chose S tier
578. 17:26rank format, then I can't make the
579. 17:28contrarian take around the text hook
580. 17:30because I have to just rank hook types
581. 17:32in the S A B C DF visual. So, it
582. 17:35completely changes what I can use to fit
583. 17:38the format. So, in that case, my
584. 17:39contrarian take might be a certain type
585. 17:41of hook that everyone thinks is good,
586. 17:42but I rank bad or vice versa. But in
587. 17:45that case, the format helped put rails
588. 17:46on how I came up with the contrarian
589. 17:49take. So if you need help coming up with
590. 17:50these contrarian takes and you struggle
591. 17:52to do it, picking a more restrictive
592. 17:53format, putting guard rails on you can
593. 17:55actually be kind of helpful in the
594. 17:57development process. Constraints breed
595. 17:59creativity is like a common frame. So if
596. 18:00you add more constraints, you might be
597. 18:02able to be more creative. For example,
598. 18:04formats like the clone format that you
599. 18:06always see Ava doing, the A versus B
600. 18:07vers, or something like rating, ranking,
601. 18:10or tier list. These are more restrictive
602. 18:12formats that give you a set narrative
603. 18:14that you can play within. If you're like
604. 18:16me and you want to freestyle more
605. 18:17because you have like wild ideas, you
606. 18:18want to explain them. Well, then you
607. 18:19want a looser, more flexible format like
608. 18:21breakdown or case study or scenario. All
609. 18:23right. Now, at this point in the video,
610. 18:25we're halfway through the six-step
611. 18:26process. We've done topic, we've done
612. 18:28format, and we've done substance. And
613. 18:30those three things really combine
614. 18:31together to create the idea. When you
615. 18:33hear people talk about ideas is all that
616. 18:35matters, it's really a combo of those
617. 18:36three things. And this is where most of
618. 18:38the sauce is. When you build an AI
619. 18:40content system, what you're doing is
620. 18:41freeing up all your time so that you can
621. 18:43focus on what that non-obvious
622. 18:45contrarian take is and developing it.
623. 18:47That's what all the pros in content are
624. 18:49doing. But as you can see, I'm using AI
625. 18:51in every step of the way. I'm using AI
626. 18:53fully for topics. I'm using AI mostly
627. 18:55for formats almost always. And then I'm
628. 18:57coming up with the contrarian tech on my
629. 18:58own. And I'm using AI as a thought
630. 19:00partner for the evidence development.
631. 19:01All right, the fourth step in this flow
632. 19:03is the hook. Everybody talks about the
633. 19:04hook. Everybody cares about the hook.
634. 19:06Once you have the topic, the format, and
635. 19:08the substance, then it's time to move to
636. 19:09the hook. And we all know the hook is
637. 19:11made up of three things. The visual
638. 19:12hook, which is what's shown, the text
639. 19:14hook, which is the title text on screen,
640. 19:16and the spoken hook, which is what is
641. 19:17said and what the captions usually say.
642. 19:19Now, here's how I run my hook
643. 19:20development process. I do the same thing
644. 19:21every time. First, I start with the
645. 19:23spoken hook, and I go back to the data
646. 19:25that I pulled from Sand Castles. I go to
647. 19:27Claude and I say this. I want you to run
648. 19:29an analysis on all the hooks that we've
649. 19:30pulled from the top 100 videos. First, I
650. 19:33want you to bucket the videos by
651. 19:34storytelling format. Rank the
652. 19:37performance of the videos within each
653. 19:38format based on outlier score. Extract
654. 19:42out the hooks that were used for each
655. 19:43one and extract out the hook Mad Lib
656. 19:46format that's plug-andplay. I want you
657. 19:48to list out the original topic, the
658. 19:50video URL, the hook that was used, and
659. 19:52the Mad Lib hook format. So that when I
660. 19:54give you a new topic and I tell you what
661. 19:56storytelling format I'm using, I want
662. 19:58you to pull all the hooks from within
663. 20:00that bucket and generate all new hooks
664. 20:03using those Mad Lib hook formats. No
665. 20:05matter what topic I give you, depending
666. 20:06on how many hooks this nets me, maybe
667. 20:08I'll ask you to run this hook conversion
668. 20:10process again for another storytelling
669. 20:12format that is adjacent. I want you to
670. 20:14build this into a hookw writing skill
671. 20:16and let me know when you're ready for my
672. 20:18first topic. So by doing it this way,
673. 20:20what you're essentially doing is
674. 20:21bucketing all the videos by format.
675. 20:23you're getting all the hooks within that
676. 20:24bucket and ranking them. And then when
677. 20:26you pick a format, you can say just use
678. 20:28the other hooks that worked for that
679. 20:30format for mine. Essentially, it's a
680. 20:31data driven hook machine, but it's
681. 20:32matching up the hooks by format. If
682. 20:34you're not getting enough that way, you
683. 20:35can expand to other formats. And I like
684. 20:37matching the hooks by format. If you're
685. 20:38doing an S tier video, the best way to
686. 20:40find a winning hook is to look at all
687. 20:41other S tier video hook formats and be
688. 20:43like, which one crushed? And then use
689. 20:45that. Sometimes hooks don't transport
690. 20:46well across formats. Sometimes they do,
691. 20:48sometimes they don't. So that's why I
692. 20:50like keeping the format consistent at
693. 20:51first. Now, after I get the spoken hook,
694. 20:54the written hook down, locked in. I'll
695. 20:56then go watch the individual videos from
696. 20:58within that format category. That's why
697. 21:00I asked for the link. I'll open it and
698. 21:02I'll just watch every single one in a
699. 21:03new tab. And what I'm doing here is I'm
700. 21:05studying the visual hook execution and
701. 21:07the text hook. I can do this faster with
702. 21:09my eyes than I could from a visual AI
703. 21:11telling me like, oh, put the block up
704. 21:13there. I can just do it way faster with
705. 21:14my eyes. So, I like opening them up and
706. 21:16looking at it. Once I find one that
707. 21:18actually hooked me, then I will mimic
708. 21:20that visual execution. If you're doing
709. 21:22this and you see a text hook that you
710. 21:24really like that your eyes went to and
711. 21:25you got it right away, write that down.
712. 21:27That could be some version of the copy
713. 21:28that you write on your text hook. All
714. 21:30right, step five in the flow is the
715. 21:32script. So, at this point, we now have a
716. 21:34topic that is validated by data. We have
717. 21:36a format that's validated by data. We
718. 21:37have a non-obvious contrarian take that
719. 21:39we feel good about. We have evidence to
720. 21:41support that. And we have a hook written
721. 21:43in the style of our topic and format.
722. 21:45Now, I found that AI writing platforms,
723. 21:47none of them are great at writing
724. 21:49scripts out of the box. You'll never
725. 21:50find one without tuning it that's good
726. 21:52enough. So, this is what I do. I make my
727. 21:54own script profile based on my
728. 21:56transcripts, or you could do it off a
729. 21:58transcript of another creator that you
730. 21:59like. So, you're going to go back to
731. 22:01Sand Castles. This time you're going to
732. 22:02make sure either your profile or
733. 22:04whatever the profile of the creator you
734. 22:06want to train a voice model on. You want
735. 22:08to make sure they're in your watch list.
736. 22:09Go to the videos tab. Click the channel
737. 22:11filter. Click just them so it pulls up
738. 22:14only their videos. Again, sort by
739. 22:16outlier score, 2% for the minimum bound
740. 22:18of engagement, and go back like last 6
741. 22:20months. So, this is that person's best
742. 22:22performing videos. Now, what you want to
743. 22:23do is analyze like 10 to 20 of them.
744. 22:26Make sure to skip any ones where the
745. 22:28format is weird. It doesn't match what
746. 22:29you want. Like if they did a about me
747. 22:31video that's kind of like more
748. 22:32storytelling, narrative driven, and
749. 22:34that's not what you want, we'll skip
750. 22:35that one. Just get 10 to 20 videos that
751. 22:37are around the same style, the same
752. 22:39format, top performers, analyze all of
753. 22:41those. Now, you're going to export that
754. 22:43CSV with just those and you're going to
755. 22:45import that into Claude. And again, if
756. 22:46you already have winners in your own
757. 22:48voice, just do it with your own videos.
758. 22:49Don't copy some other creator if you
759. 22:51already have your own voice dialed. But
760. 22:52you don't have a voice dialed that you
761. 22:54want, go ahead and pick one other
762. 22:55creator that you want to mimic the tone
763. 22:57and rhythm and use them. I don't
764. 22:59recommend combining multiple creators
765. 23:00here because oftentimes, you know,
766. 23:02speaking patterns are like fingerprints.
767. 23:04I do it one way, they do it one way, she
768. 23:05does it one way. If you combine examples
769. 23:07from all three, it'll confuse the writer
770. 23:09and make it generic. So, you want to
771. 23:10pick one focused person and have 10 to
772. 23:1220 of the top performing transcripts
773. 23:14from that person. Export those out. Now,
774. 23:16you have the CSV. When I go back to
775. 23:18Claude, I open a new chat because this
776. 23:19is kind of like a new thread. Drop the
777. 23:21CSV in. And then this is what I prompt
778. 23:23for the scripting profile. I'm looking
779. 23:25to build a short form scripting profile
780. 23:27based on the transcripts that I've just
781. 23:29uploaded. First, I want you to take all
782. 23:31the transcripts and fix if there are any
783. 23:33small typos or weird things. Rewrite and
784. 23:35fix them so they're clean. Then, I want
785. 23:37you to take all those clean transcripts,
786. 23:39feed them in, and build a short form
787. 23:40script writing profile model based on
788. 23:43these. I want to be able to feed you a
789. 23:45topic, a format, a contrarian take, and
790. 23:47a hook. And I want you to be able to
791. 23:48write a script in the tone, rhythm, and
792. 23:51style of the script model you've just
793. 23:52made. I want you to save this as a
794. 23:54script writing skill. So, anytime I make
795. 23:56a new video, I can open this and let it
796. 23:58rip. Now, once you have this written,
797. 24:00it'll shock you how good this is. It'll
798. 24:01get you probably 90 to 95% of the way
799. 24:03there, especially if you already do the
800. 24:05work on the evidence. You already have
801. 24:07the hook. You have a lot of the bones
802. 24:08already there. You're not relying on it
803. 24:10to come up with new stuff. When most
804. 24:12people say AI writing is trash, what
805. 24:14they mean is AI coming up with certain
806. 24:16facts that are net new and writing it
807. 24:18together sounds kind of mid. But if you
808. 24:20actually guide it with a lot of the
809. 24:22datadriven research and you already have
810. 24:23a script model cloned based on the voice
811. 24:25and rhythm you like, it's going to get
812. 24:27you pretty close. From here, of course,
813. 24:29you may still need to go the last mile.
814. 24:30You may still need to tweak 5%. There
815. 24:32might be one or two sentences or one or
816. 24:34two things you want to tweak, but that's
817. 24:35about as close as you can get right now.
818. 24:37At this point, go ahead and take that
819. 24:38script and record. All right, the last
820. 24:40step in the flow, which is step six, is
821. 24:42the editing. At this point, every single
822. 24:44step has had a lot of AI driving it, but
823. 24:46we still preserve the creativity,
824. 24:47originality, which comes from the
825. 24:49contrarian take. Now, for those people
826. 24:50that say editing doesn't matter, there
827. 24:52are some formats that are lower frills
828. 24:54where editing is not really required. I
829. 24:56still think editing matters a lot from a
830. 24:58visual storytelling perspective. There's
831. 24:59so much emotion you can drive with the
832. 25:02right pacing and the right visuals, but
833. 25:04there's a wide spectrum in how you could
834. 25:05approach this. You really have four ways
835. 25:07to tackle editing. So, it's kind of
836. 25:09dealer's choice and how you want to do
837. 25:10it. The first way is to intentionally
838. 25:12pick a format that doesn't really
839. 25:13require a ton of editing. This would be
840. 25:14like yapping or the clone style or, you
841. 25:17know, just like more raw personal
842. 25:19storytelling without a ton of editing.
843. 25:21This would be the first pass. You can
844. 25:22bypass editing with this. Do this if you
845. 25:24don't have the time. It's going to be
846. 25:25harder to cut through with these. You
847. 25:26really need to dial in everything else
848. 25:28if you're going to go this way. The
849. 25:29second way would be to brute force the
850. 25:30editing yourself. You pick whatever
851. 25:32format you think is best based on the
852. 25:33data. And no matter what it takes, you
853. 25:35brute force it. This is what I did. You
854. 25:37could do it in Cap Cut. You could do it
855. 25:38in Reals. You could do it in Premiere
856. 25:39Pro. I taught myself how to do it until
857. 25:41I could outsource it. The third way
858. 25:42that's starting to emerge is people are
859. 25:44using claude code and a remotion skill
860. 25:47to actually automate some of the editing
861. 25:49with AI. I don't think this is very
862. 25:50premium yet. It's still like 2 out of
863. 25:5210, three out of 10 in terms of the
864. 25:53complexity, but this is starting to
865. 25:55work. If you're non-technical and you
866. 25:57don't want to spend all the time
867. 25:58learning this, I wouldn't put all the
868. 25:59eggs in this basket. However, this is
869. 26:01emerging. I'll continue to comment on it
870. 26:03as it gets better. The fourth way to
871. 26:04tackle editing is to hire an editor and
872. 26:06outsource it. The reason I don't talk
873. 26:08about editing in detail on this channel
874. 26:10the way I do about all the other stuff
875. 26:11is because if you're a business owner or
876. 26:12you're trying to build an actual revenue
877. 26:14engine from content, you should not be
878. 26:16editing. You cannot be editing. It's too
879. 26:17low leverage of a task for you as the
880. 26:19creator to be doing it. So, I recommend
881. 26:21outsourcing this. I've got a bunch of
882. 26:22editors on every single one of my teams.
883. 26:24If you want to work with the same people
884. 26:25I work with for these YouTube videos,
885. 26:27for all my shorts on tech, all my shorts
886. 26:28on marketing, I've got a link below. You
887. 26:30can fill that out and I will connect
888. 26:31you. All right, so that is it really.
889. 26:33That's the full endto-end sauce in terms
890. 26:35of one rep for the workflow. Its topic,
891. 26:38its format, its substance, its hooks,
892. 26:40scripts, and edits. We talked about in
893. 26:42every single one of those pieces, you
894. 26:44can use AI to automate a lot of it other
895. 26:46than the edit, but you could use
896. 26:47Remotion if you wanted to do that as
897. 26:49well. Now, my kind of macro meta
898. 26:51perspective is a thing called AI enabled
899. 26:53creativity. This is kind of what I'm
900. 26:54going all in on. What I think AI is
901. 26:56great for is taking away all the boring
902. 26:59stuff so that you can spend all your
903. 27:01time creating. It's not that AI is going
904. 27:03to remove human creativity. It's that it
905. 27:04frees up human creativity. That's how
906. 27:06I'm looking at it. So, for me, me coming
907. 27:08up with these crazy metaphors and wild
908. 27:10concepts like dopamine ladders or all
909. 27:12the stuff I do in the videos, that takes
910. 27:13time to cook that up. I don't have as
911. 27:15much time when I have to do all this
912. 27:174-hour editing or 3-hour research
913. 27:19sprints. But if I can use AI to
914. 27:20outsource all of that, I can spend all
915. 27:22my time cooking up the cool stuff. And
916. 27:24that's what I think the power really is
917. 27:26when you build one of these systems.
918. 27:27Now, the question you would have from
919. 27:29here, this last piece I want to cover is
920. 27:30what I call my 10x batch system. Cuz
921. 27:33really what you're wondering is, okay,
922. 27:34that was one rep, but how do I actually
923. 27:36like run the system over and over? How
924. 27:38do I know what to make over time? Do I
925. 27:40do this every time? How do I sequence
926. 27:42it? Let me just explain how I approach
927. 27:43this flow. You're going to run that
928. 27:45whole loop that I just said once and
929. 27:47that's going to produce 10 videos,
930. 27:48right? So, the topic, format, research,
931. 27:51that one pull should inform 10 videos.
932. 27:54You're going to go substance all the way
933. 27:55down. Substance all the way down. All
934. 27:56the way down. All the way down 10 times.
935. 27:57So that first part you're going to run
936. 27:58once and then you're going to have that
937. 28:00substance hook, script, editing flow 10
938. 28:02times. That's one batch of 10 videos.
939. 28:04You're going to post each one of those
940. 28:06videos every day back to back for 10
941. 28:08days. Now, once you post the 10, I want
942. 28:10you to look backwards at the analytics
943. 28:12and I want you to look at three
944. 28:13different metrics. These the only three
945. 28:14metrics that matter for content. Number
946. 28:16one, and I think the most important is
947. 28:18conversions per video or conversions per
948. 28:20view. So, if you have a lead magnet or a
949. 28:22low ticket product and you're actually
950. 28:24able to track via many chat the
951. 28:25conversions off the video, this is the
952. 28:27absolute best metric to measure and
953. 28:29you're going to look at how many
954. 28:30conversions did I drive per video. If
955. 28:32you can't track that or you don't have a
956. 28:33backend built out, the next best thing
957. 28:35would be to look at followers per video
958. 28:37or followers per view. How many
959. 28:39followers did this video drive? Now,
960. 28:41followers themselves don't matter in
961. 28:42terms of you getting reach, but a follow
962. 28:44from a human psychologically is a proxy
963. 28:47for an endorsement that is bigger than a
964. 28:48view. If someone presses follow, they're
965. 28:50like, "Wow, I really like this guy's
966. 28:52stuff. I really like this girl's
967. 28:53content." That is higher weighted than
968. 28:55them just watching something that was
969. 28:56given to them automatically. So, if you
970. 28:58can't look at conversions, you want to
971. 29:00look at followers gained per video and
972. 29:02use that to guide. The worst metric of
973. 29:04the three, but still way better than
974. 29:05everything else, is just total views.
975. 29:07You're going to look at how many views
976. 29:08you got relative to the field. So, what
977. 29:10I want you to do is after you post a
978. 29:11batch of 10, I want you to look
979. 29:12backwards and be like, did any of the 10
980. 29:14get at least a 5x above the average for
981. 29:18any of those three metrics. Did any of
982. 29:19them drive 5x more conversions, 5x more
983. 29:22followers, or 5x more views than the
984. 29:24average of the 10. Now, here's how
985. 29:25you're going to treat this. If none of
986. 29:26them got above a 5x, they were all at
987. 29:28the average, and ideally the average is
988. 29:31probably low, that means you're going to
989. 29:32rerun this process from scratch the next
990. 29:34batch, right? And typically when this
991. 29:36doesn't happen, there's usually like an
992. 29:37obvious thing that's wrong with your
993. 29:38videos or several obvious things. I
994. 29:40actually put a program together where if
995. 29:42you need help, I just literally review
996. 29:44your videos myself and I say like, "Fix
997. 29:45that. Fix that. Ooh, that's why that's
998. 29:47not working. Your hook's not working."
999. 29:48Every single call twice a week. I just
1000. 29:50go through everyone's videos as a group,
1001. 29:52one-on-one videos, but so everyone can
1002. 29:54see them. And I'm just like, "That's not
1003. 29:55working. That's not working. Oh, you
1004. 29:56messed that up. You messed that up."
1005. 29:57It's really helpful. So if you want
1006. 29:59that, I've linked it below. We've just
1007. 30:00opened a few spots. You don't have to
1008. 30:01have it forever, but it's really helpful
1009. 30:03for you to scale. If you get feedback
1010. 30:05from me, so that's if nothing is above a
1011. 30:075x. Now, if you get one or more that's
1012. 30:09above a 5x, but not above a 10x. It's
1013. 30:11kind of in this 5 to 10x range, that
1014. 30:13means you've got a winner, and you're
1015. 30:15going to take that topic and format, and
1016. 30:16you're going to tweak it slightly when
1017. 30:18your next batch, three of the 10 in the
1018. 30:20next batch should follow that topic and
1019. 30:22format. So, for me, in the example,
1020. 30:24hooks was the category, how to write
1021. 30:26better hooks was a topic, and the format
1022. 30:28was a breakdown. If I get a 5x, I'm
1023. 30:30going to be like, "Okay, how to write
1024. 30:31better hooks. Is there another
1025. 30:32contrarian take I could share? Could I
1026. 30:34tweak it somehow? Maybe try a different
1027. 30:35format?" Because you have data telling
1028. 30:37you that that video outperformed. Now,
1029. 30:39if you have a 10x or more, you're going
1030. 30:41to take the exact same topic, the exact
1031. 30:43same format, and you're going to run it
1032. 30:44again. That is like liquid gold. So,
1033. 30:46when you have a winner, five or 10x,
1034. 30:48you're going to run it three of your
1035. 30:49next 10. You're not going to run all 10
1036. 30:51like that cuz it's going to get way too
1037. 30:52repetitive for you and the viewer. But
1038. 30:54three of your next 10 are going to be
1039. 30:55like that. If you have three winners,
1040. 30:56well, that takes care of your next 10.
1041. 30:58This is called the 10x batch process.
1042. 31:00You run these batches of 10, you look
1043. 31:01backwards, you take a winner, and you
1044. 31:03carry it forward. What you're really
1045. 31:04looking to do is to have three different
1046. 31:06combinations of winners that just crush
1047. 31:0810x. That will inform your entire batch
1048. 31:11process. So, every time you make one of
1049. 31:1210, they just crush. All you have to do
1050. 31:14is continue running iterative loops of
1051. 31:1610, getting better from video to video,
1052. 31:18and taking forward the ones that worked,
1053. 31:20and you'll be good. Like I said, if you
1054. 31:22want help from me reviewing, this is
1055. 31:24really what I do. Like I get on the call
1056. 31:26and I just review everyone's videos
1057. 31:27one-on-one. They put their videos in the
1058. 31:29chat and I just say, "Cool, let me watch
1059. 31:31this." We all watch it together and I
1060. 31:32just give feedback live. If you want me
1061. 31:34to cook on your videos and just solve
1062. 31:35this for you, I've got a link below. Few
1063. 31:37spots open. All right, guys. That is all
1064. 31:38I've got for this video. This was a long
1065. 31:40one. You can tell in my voice I'm I'm
1066. 31:41cooked. To be honest, like this level of
1067. 31:43detail should not be given away for
1068. 31:44free, but I really do want to see all
1069. 31:46you guys win. Like I said in previous
1070. 31:47videos, my goal is I want a million
1071. 31:49people to make a million dollars from
1072. 31:50content. That's a trillion dollars in
1073. 31:52market value. in order to get you guys
1074. 31:53there, I need to give you every single
1075. 31:55thing I can. So, that's why I'm making
1076. 31:56these videos one of one for you. If
1077. 31:58you're watching this and you're like,
1078. 31:59"Man, that was amazing. I can't believe
1079. 32:01he gives this for free. What can I do to
1080. 32:02help?" If you're thinking that, this is
1081. 32:03what would really help me. Subscribe to
1082. 32:05the channel if you haven't already
1083. 32:06subscribed. If you have subscribed,
1084. 32:08especially like the video and leave a
1085. 32:10comment. And the comment I just want to
1086. 32:12be the chef emoji with the mustache.
1087. 32:14That's like my bat signal cuz I'm
1088. 32:16cooking, right? I'm cooking on these
1089. 32:17videos. Just drop the chef emoji cold.
1090. 32:20No other context but that. If you do
1091. 32:22that, you're a real one. I want to be
1092. 32:23able to look back on the videos and just
1093. 32:25see chef, chef, chef, chef, chef, chef,
1094. 32:26chef. Like, everyone's just chefing. If
1095. 32:28you do that, that means you got all the
1096. 32:29way through and you actually locked in
1097. 32:30this information. So, that's all I ask
1098. 32:32from you guys. I got a ton of stuff for
1099. 32:34free below. I mentioned several free
1100. 32:36things throughout the video. You can
1101. 32:37just grab those, the prompts, the visual
1102. 32:39database. You can use that. Like I said,
1103. 32:40I really want to see you guys win. So,
1104. 32:42hopefully you enjoyed that video. That's
1105. 32:43all I've got on this one and we will see
1106. 32:45you guys on the next one. Peace.