YouTube2Text

/roast-room: How I Use Claude to Think Like the Top 1% — Transcript

by Nate Herk | AI Automation · 3,577 words · 498 segments · language en · Watch on YouTube

Full transcript

  1. 0:00So, right here you can see that I've got
  2. 0:01Jeff Bezos, Paul Graham, Alex Hormozi,
  3. 0:03Gary V, Dan Martell, Gino Wickman, and
  4. 0:04Warren Buffett all discussing a business
  5. 0:07question that I had. You basically shoot
  6. 0:08off a question with different options,
  7. 0:10and all of these different council
  8. 0:11members will kind of come together,
  9. 0:13roast your idea, and help you think
  10. 0:15about it in a different way. Because
  11. 0:17each one of these seats has one specific
  12. 0:19question that they're focusing on. And
  13. 0:20now you can see what's happening is
  14. 0:21they're actually debating with each
  15. 0:22other. So, Jeff Bezos was debating with
  16. 0:24Gary V. Same thing with Paul Graham and
  17. 0:26Gary V. Now, I've got Alex Hormozi going
  18. 0:28after Gary V, too. And well, everyone's
  19. 0:30deciding to pick on Gary V in this roast
  20. 0:32room. This is crazy. There we go. Warren
  21. 0:34Buffett goes after Hormozi. But, the
  22. 0:36point is, like I said, everyone has
  23. 0:37their own question they focus on. So,
  24. 0:39Warren Buffett's question is, "What does
  25. 0:41it do to cash, risk, and long-term
  26. 0:43value?" Alex Hormozi's question is,
  27. 0:45"Will people pay? How much? And how do
  28. 0:48we close?" And Gino Wickman's question
  29. 0:50is, "Can the business run this without
  30. 0:52the founder holding it all together?"
  31. 0:53Now, after all three of those rounds
  32. 0:55happen, we can see that we have a final
  33. 0:56verdict, which was option C. And then we
  34. 0:58can also download a PDF, which allows us
  35. 1:00to read this, and we can sort of look at
  36. 1:02this more in depth, and we could also
  37. 1:04share this with our team if we wanted
  38. 1:05to. Now, this is really just a visual
  39. 1:06layer on top of what actually goes on in
  40. 1:08Claude code, which is basically just
  41. 1:10running a bunch of sub-agents based on
  42. 1:12these different council members that I
  43. 1:14have created. So, you don't need this
  44. 1:15localhost open in order to run and use
  45. 1:17this, but it's just a nice visualization
  46. 1:20of what's going on. And this is
  47. 1:21completely customizable, meaning you can
  48. 1:23switch out different council members,
  49. 1:24and you can build other ones, too. In
  50. 1:26fact, inside of this repo that I'm going
  51. 1:28to give you guys, there already are a
  52. 1:29few other ones, too. Hey Claude, what
  53. 1:31are the other council members inside of
  54. 1:33our roast room? And so, that's Glydo
  55. 1:35actually shooting off a prompt right now
  56. 1:36to Claude code, as you can see, and then
  57. 1:38it'll show us who the other council
  58. 1:40members are. Claude code comes back and
  59. 1:42says there are 11 seats. We have the
  60. 1:43default seven that you're seeing here,
  61. 1:44and then we also have Naval, Cody
  62. 1:46Sanchez, Sara Blakely, and Leila
  63. 1:48Hormozi. But, of course, you can build
  64. 1:50as many other ones as you want. So,
  65. 1:52that's a new feature of Glydo where
  66. 1:53we're able to actually orchestrate
  67. 1:55different chats between different coding
  68. 1:56agents. If you want to try it out, you
  69. 1:57can get a free month using the link in
  70. 1:59the description. But anyways, just
  71. 2:00wanted to show this off to you guys and
  72. 2:01let me break down how this actually
  73. 2:02works. So, this is a repo called the
  74. 2:04Roaster Room, as you guys can see. And
  75. 2:06basically, we have these seven different
  76. 2:07advisors as well as four extras, and I'm
  77. 2:09probably going to keep adding more in
  78. 2:11the mix as well. But what I wanted to
  79. 2:12show you is the advisors. What are these
  80. 2:14council members? Where do they actually
  81. 2:16live? They're actually just a bunch of
  82. 2:17sub agents. So, if I go into the dot
  83. 2:19Claude, actually let me zoom in a little
  84. 2:20bit more. You can see that we have
  85. 2:22agents. And in here, we have one
  86. 2:23markdown file for each of these sub
  87. 2:26agents. So, Bezos, Blake Lively,
  88. 2:27Buffett, Gary V, Graham, and all the way
  89. 2:29through. And if we open one up so you
  90. 2:31can see what it is, it's literally just
  91. 2:32a markdown file that holds information
  92. 2:34about the sub agent. So, for Jeff Bezos,
  93. 2:36for example, how he thinks, what are the
  94. 2:39sources that he references, how he makes
  95. 2:40decisions, all that sort of stuff. So,
  96. 2:42that when I actually shoot off a prompt
  97. 2:44to the Roaster Room, it's going to go
  98. 2:46through the different council members
  99. 2:47and it sort of adopts their personality
  100. 2:49of how they think and how they make
  101. 2:50decisions, what they, you know, and it's
  102. 2:52able to pull from things like their
  103. 2:54actual conversations, their lectures,
  104. 2:56their books, their principles, things
  105. 2:58that they put out there publicly,
  106. 2:59articles, stuff like that. So, what
  107. 3:01you'll notice is they will all have
  108. 3:02their own agent file, which isn't really
  109. 3:04that long. It's some things like how you
  110. 3:06debate, how you talk, your lens. But the
  111. 3:09really interesting stuff here is the
  112. 3:10brain, right here. So, we can see all
  113. 3:12the sub agents. But if I actually go
  114. 3:14into the skills in this repo, we can see
  115. 3:17that there is a Roaster Room folder that
  116. 3:20has brains. And once again, there's a
  117. 3:22folder for every single council member
  118. 3:23that you want to spin up. So, if we for
  119. 3:26example, open up Leila Hormozi's brain,
  120. 3:29we can see that she has hot, books,
  121. 3:31frameworks, positions, principles,
  122. 3:32sources, stories, and voice. And it's
  123. 3:34pretty much the same thing that all of
  124. 3:35these other council members have as
  125. 3:36well, but they all obviously look
  126. 3:39different. The content is different. So,
  127. 3:40if we open up Leila's hot, what happens
  128. 3:43is we can see her pillar is can your
  129. 3:45team and your leadership carry this? And
  130. 3:48who has to change? It shows us what this
  131. 3:50was built from. So, for example, Leila
  132. 3:51was built from 48 sources, about half a
  133. 3:53million words from 2021 to 2026, 14
  134. 3:56interviews where she's the main guest,
  135. 3:58and 34 videos of her own on leadership
  136. 4:01and things like that. Every person here
  137. 4:03has eight rules they decide by,
  138. 4:04frameworks, decision loops, things like
  139. 4:06that. Obviously, any books that they've
  140. 4:08written, frameworks that they talk about
  141. 4:10frequently and that they quote, their
  142. 4:11positions on certain things, pricing,
  143. 4:13discounting, raising money, debt, their
  144. 4:15principles, different sources that they
  145. 4:16actually pulled from, specific stories
  146. 4:18that they might talk about in videos,
  147. 4:20interviews, things like that. And every
  148. 4:22single one of these brains for the
  149. 4:24individual people is built the exact
  150. 4:25same way. If we open up Dan Martell, we
  151. 4:27can obviously see that he's got books in
  152. 4:29here as well. We can see that he's got
  153. 4:30frameworks that he's using all the time.
  154. 4:32You can see the buyback principle, the
  155. 4:33buyback rate, time and energy audit, the
  156. 4:35drip matrix. All of this stuff should
  157. 4:36sound familiar to you guys if you have
  158. 4:38followed, you know, Alex Hormozi or Dan
  159. 4:40Martell, value equation, grand slam
  160. 4:42offer, all of these things that they
  161. 4:44actually talk about all the time. A
  162. 4:45quick break to tell you about the
  163. 4:46sponsor of today's video, Fish Audio.
  164. 4:49Now, anyone who's tried AI voiceovers
  165. 4:51knows that the words come out right, but
  166. 4:53the delivery doesn't always because it
  167. 4:55might read a joke or a warning and an
  168. 4:57apology in the exact same tone of voice.
  169. 4:59So, Fish Audio has a new model that is
  170. 5:01their most emotive yet, and it's called
  171. 5:03Drama 3. You direct it the way that you
  172. 5:06would direct an actual voice actor,
  173. 5:07which means that you can write the
  174. 5:08direction in plain English inside the
  175. 5:10text like slowly realizing something's
  176. 5:12wrong, and everything after that tag
  177. 5:14will follow that direction. So, one line
  178. 5:16can start calm, and it can end in a
  179. 5:18panic.
  180. 5:19>> Thanks for calling. Your order's on its
  181. 5:21way.
  182. 5:22Wait. That's not the right address. Who
  183. 5:25approved this workflow?
  184. 5:27>> And the tags work in other languages,
  185. 5:29too. It also generates full
  186. 5:30conversations. You can give each speaker
  187. 5:32a voice, you can mark who's talking, and
  188. 5:34every character keeps their own voice
  189. 5:36across the entire scene.
  190. 5:37>> Just run it one more time.
  191. 5:39>> Of course. That'll be attempt 47.
  192. 5:44>> So, if you're putting it on video, the
  193. 5:45streaming endpoint sends word-level
  194. 5:46timestamps back with the audio. So, your
  195. 5:49captions line up exactly with the
  196. 5:50speech. Now, on the build side, you set
  197. 5:52the model to Jama 3 in the Fish Audio
  198. 5:54API. They also have an MCP server, too,
  199. 5:56so you can connect it to Cloud Code or
  200. 5:58CodeX, and they also have an official
  201. 5:59N8N node. So, if you're building
  202. 6:01something that talks, like a voice agent
  203. 6:03or a game character or just narrating
  204. 6:05your videos, then try Jama 3 on the Fish
  205. 6:07Audio API with the link in the
  206. 6:09description. Huge thanks to Fish Audio
  207. 6:11for sponsoring this part of the video.
  208. 6:13Now, let's get back to it. And then, I
  209. 6:14added this new skill called New Roster,
  210. 6:16which basically just lets you build new
  211. 6:19rosters. So, for example, if I came in
  212. 6:20here and I was like, "Hey, new roster,
  213. 6:23and we wanted to do, you know, Elon
  214. 6:25Musk." And we shot that off, this will
  215. 6:27basically create Elon Musk's brain and
  216. 6:30his agent file and all of his sources.
  217. 6:32It will build him out and also build him
  218. 6:34a little cartoon character that will
  219. 6:35show up in the actual roster room
  220. 6:37itself, exactly the same way that all of
  221. 6:39these other roster room council members
  222. 6:41have been built out. So, you can keep
  223. 6:43scaling this thing up, and obviously,
  224. 6:44I'm going to let this thing build out
  225. 6:45Elon, and then I'm going to add that
  226. 6:47into the repo so you guys can all use
  227. 6:48Elon as well. So, while this is building
  228. 6:50out Elon, as you can see, let us do a
  229. 6:53new session. So, open up the roster
  230. 6:55room. That's pretty much all we have to
  231. 6:56do is ask for it, and I'm just going to
  232. 6:58switch this back to Opus here, and it's
  233. 7:00going to open up the roster room. We'll
  234. 7:01open it up in the local browser inside
  235. 7:03of the desktop app, and then we will
  236. 7:04shoot off a prompt so you guys can see
  237. 7:06how this actually works. Now, once
  238. 7:08again, you don't need this visual
  239. 7:09interface opened up. It will obviously
  240. 7:11run the sub agent roster room just the
  241. 7:13same, but it's nice to see. But, one
  242. 7:15thing that I will say about actually
  243. 7:16spinning up a roster room is typically,
  244. 7:18you want to put in a decision. You want
  245. 7:20to put in options, so two or three
  246. 7:22options. You want to give it context so
  247. 7:24that they can actually reason through it
  248. 7:26and make a better call, right? So, let
  249. 7:29me just shoot off a quick example. We
  250. 7:31are looking at hosting another in-person
  251. 7:33event in the summer of 2027, and we're
  252. 7:36trying to figure out where to do it.
  253. 7:37We've been, you know, analyzing
  254. 7:39different options like San Francisco,
  255. 7:42Salt Lake City, Austin, Chicago. We
  256. 7:45obviously want to be able to get good
  257. 7:46sponsors there, local incentives, get a
  258. 7:49lot of people there, and you know, so
  259. 7:51airports should be important, um,
  260. 7:54public transportation, good restaurants,
  261. 7:57good activities to do outside of just
  262. 7:58having the event, things like that. So,
  263. 8:00help me figure out where this should be.
  264. 8:02And do this on the roast room council.
  265. 8:04So, this natural language request should
  266. 8:06be just fine, but you could also
  267. 8:07couldn't come in here and do a {slash}
  268. 8:09roast room, and then you could just
  269. 8:10invoke this like that. But, we will go
  270. 8:13ahead and just get rid of that for now
  271. 8:14and shoot that off, and it should be
  272. 8:16good enough. One thing that this will do
  273. 8:17is it will help you figure out what
  274. 8:18would be the best council members to
  275. 8:20have for this specific debate. As you
  276. 8:23can see, it's going to load the skill,
  277. 8:24it's going to gather event and business
  278. 8:26context, so it will ask you questions if
  279. 8:28it doesn't feel like it yet has enough,
  280. 8:29and then it will actually go ahead and
  281. 8:30run the debate for you. So, you can see
  282. 8:32it is going to ask me a couple questions
  283. 8:34here. It has four questions that is
  284. 8:35asking me. And by the way, we already
  285. 8:37did book this event in Salt Lake City,
  286. 8:40June 16th through 20th. So, if you guys
  287. 8:42want to attend, then definitely check it
  288. 8:44out. But, for the sake of the example,
  289. 8:46I'm just going to pretend that none of
  290. 8:48this has happened yet. So, I'm just
  291. 8:49going to say not signed, open choice. It
  292. 8:51says, "What size and budget should the
  293. 8:52council plan around?" And I actually
  294. 8:54just skipped through those questions,
  295. 8:56but the last one says, "Who sits at the
  296. 8:57table?" And it gave me a recommendation,
  297. 8:59and it also let you play with it if you
  298. 9:01wanted to put different people in. So,
  299. 9:02that's part of it is thinking about who
  300. 9:05should actually sit on the council for
  301. 9:06your specific debate. But, now it's
  302. 9:07going to be it. It's basically going to
  303. 9:08run, it's going to process this. Even if
  304. 9:10this was closed, like I said, we would
  305. 9:12have sub agents that would be going
  306. 9:13through and doing this no matter what.
  307. 9:16And the reason I decided to do this with
  308. 9:17sub agents because when I actually
  309. 9:18started doing these roasts, I started
  310. 9:20doing this with agent teams, but the
  311. 9:22reason I'm doing sub agents now is
  312. 9:23because I found after doing lots of
  313. 9:25testing that agent teams was more
  314. 9:26expensive for a minimal amount of actual
  315. 9:29value add. So, doing sub agents works
  316. 9:31pretty well because they start to
  317. 9:33actually talk to each other and stuff
  318. 9:34like that either way. So, it works just
  319. 9:37fine. Now, in this case, you can see
  320. 9:38that we took out Dan Martell because his
  321. 9:40is more about like protecting the
  322. 9:41founder's time and we put in Naval. And
  323. 9:43his question is, does this build
  324. 9:44leverage and equity you own or does it
  325. 9:46just rent out your time? So, you can see
  326. 9:48it was great that we brought in Naval
  327. 9:50for this specific type of roast room.
  328. 9:52And obviously, last time you guys saw
  329. 9:54the PDF that it creates, which is pretty
  330. 9:56high-level, but all of this stuff is
  331. 9:57stored locally and you're able to dig
  332. 9:59in, you can ask more questions about it,
  333. 10:01you can inspect all that stuff if you
  334. 10:02want to look at the debate more deeply.
  335. 10:04So, it's not lost data, of course. Show
  336. 10:06me where all of the debates are held
  337. 10:09later if I want to be able to dig in and
  338. 10:11ask more questions about it. So, as
  339. 10:12they're going through round one and
  340. 10:13writing their opening positions, I can
  341. 10:14just show you real quick. If I go to the
  342. 10:16files and I go down here to projects and
  343. 10:20then we scroll down to a project called
  344. 10:21Founder Council right here. You can see
  345. 10:24that there is different sessions in here
  346. 10:26and these sessions will show us
  347. 10:27different roast rooms that we've run
  348. 10:29since creating this visual interface and
  349. 10:31we're able to actually then dig in. So,
  350. 10:33if I look at this one, we can see that
  351. 10:34there was a brief, there was a round
  352. 10:35one, round two, and a verdict. And this
  353. 10:38will hold all of the session logs from
  354. 10:39each of these people. Because as they
  355. 10:41start to write their positions, we'll
  356. 10:42see that all of their reasoning comes
  357. 10:44through even when they start to debate
  358. 10:46and we can view it right in this
  359. 10:47interface as well. So, now you can see
  360. 10:48they're starting to come through with
  361. 10:49their opening positions and we'll see
  362. 10:51below their name what they voted for.
  363. 10:53So, Gary came in with voting for A and
  364. 10:56gave it an eight out of 10 score. Gino
  365. 10:58did the same and you can see A is Salt
  366. 11:00Lake City and B San Francisco, Austin,
  367. 11:02Chicago. This one, Alex Hormozi, also
  368. 11:05voted for Salt Lake City. So, everyone
  369. 11:06voted for Salt Lake City except for it
  370. 11:09looks like Naval went for B, San
  371. 11:11Francisco. And then like I said, we
  372. 11:13could click in and we could look at
  373. 11:14exactly why, we could see the risk, we
  374. 11:16could see what they would do, and then
  375. 11:17we can keep seeing what everyone's going
  376. 11:19to do. And now what's going to happen is
  377. 11:21they're going to read each other's
  378. 11:22positions and then they'll start to go
  379. 11:24after each other and try to poke holes
  380. 11:26in each other's opinions. In this case,
  381. 11:28they all are in pretty much good
  382. 11:29agreement, which is nice, but if they
  383. 11:32did all completely agree, then the
  384. 11:34chair, or sort of like Claude Code, the
  385. 11:36facilitator of this debate, would make
  386. 11:38one of them argue for a different side.
  387. 11:40So, they kind of play devil's advocate,
  388. 11:41and we get to see more perspective as
  389. 11:42well. But, in this case, we didn't have
  390. 11:44a unanimous decision, so the debate will
  391. 11:46continue as expected. Now, what I think
  392. 11:48is really fun about this is we're able
  393. 11:51to leverage the subject matter expertise
  394. 11:53of other people. And humans are really
  395. 11:56good at their domains, right? So, like
  396. 11:59for example, not build a world-class
  397. 12:02copywriting agent. It would be tough for
  398. 12:03me because I don't have a copywriting
  399. 12:04background. But, what I could do is go
  400. 12:06find someone who I think is the best
  401. 12:08copywriter in the world. I could extract
  402. 12:10their frameworks, examples of their
  403. 12:11copy, how they think, how they decide
  404. 12:15what to write, and then I could put them
  405. 12:16into this council. And when I have
  406. 12:18questions about copy, or you know, how
  407. 12:20to market something, I could leverage
  408. 12:22them inside of this roast room. And it
  409. 12:24doesn't mean it's going to be perfect,
  410. 12:25right? It doesn't mean that whatever
  411. 12:26this roast room decides, I blindly take
  412. 12:28and accept. But, just getting in this
  413. 12:30mindset of thinking about who has done
  414. 12:32something that you want to do. Who has
  415. 12:34done something and proven it. Now, how
  416. 12:36can you go out there and study what they
  417. 12:38did and how they got there? And
  418. 12:39basically just try to follow that path.
  419. 12:42I think it's a lot better to do that
  420. 12:43than just trusting Claude's gut, or
  421. 12:45trusting your gut sometimes. Like,
  422. 12:47following someone that's actually walked
  423. 12:48the path you want to walk is a pretty
  424. 12:50good signal, right? So, doing these
  425. 12:52councils helps me just look at things
  426. 12:54through different perspectives, and you
  427. 12:56know, think about things that I might
  428. 12:57not have thought about because Paul
  429. 12:58Graham brought up this idea, or because
  430. 13:00Naval brought up that idea. And so, even
  431. 13:02outside of the bounds of this roast
  432. 13:04room, for example, I think that just
  433. 13:06thinking like that when you're building
  434. 13:08out other skills, or you're trying to
  435. 13:09automate different processes, I think
  436. 13:10thinking like that
  437. 13:12is a really good way to think. But,
  438. 13:13anyways, you can see they're actually
  439. 13:14coming through now. They're deciding on
  440. 13:17final votes. And you can see that we're
  441. 13:18probably going to win, you know, the
  442. 13:20Salt Lake City. It's 7.9 out of 10 right
  443. 13:22now, and that is probably the verdict
  444. 13:24that we're going to get. But, of course,
  445. 13:25we could go through and we could read
  446. 13:27everyone's key points. As you can see
  447. 13:28here is Hermosi's key point, here's
  448. 13:30Wickman's key point. You can see here
  449. 13:32Bezos conceded to Gary Vee and Graham
  450. 13:34about them being right on certain
  451. 13:35things, but also disagreed the most with
  452. 13:38Naval. So anyways, that's how you could
  453. 13:39sort of start to read through some
  454. 13:40things in more depth. We can see that we
  455. 13:42have the PDF now, which I'll go ahead
  456. 13:43and download real quick. And now we're
  457. 13:45able to look at the fact that Salt Lake
  458. 13:48City won, which is actually pretty cool
  459. 13:49because that is where we decided to have
  460. 13:51this first in-person event. Now anyways,
  461. 13:54that is done. We could obviously do more
  462. 13:57rounds if we wanted to. We could bring
  463. 13:58another council members. And now going
  464. 14:00back to check on Elon, you can see that
  465. 14:01it created the actual portrait for him
  466. 14:03and it created these different views so
  467. 14:06that we can see what it will look like
  468. 14:07when he's actually talking. And then if
  469. 14:09I go to the repo, we can see that we now
  470. 14:11have Elon added in here. If I go to the
  471. 14:13brain, or not the brain, the agents, we
  472. 14:15can see right here we have Musk. And
  473. 14:17then of course in the actual brain,
  474. 14:18because it's supposed to do that through
  475. 14:20the new roster skill, we have a brain
  476. 14:22for Musk. Once again, it has hot books,
  477. 14:25frameworks, positions, principles,
  478. 14:26sources, stories, and a voice.md. So we
  479. 14:29are now set up and we are able to use
  480. 14:31Elon Musk inside of this Roaster
  481. 14:33Council. If you guys want to grab this
  482. 14:35GitHub repo and play around with it,
  483. 14:36then just go to my free school
  484. 14:37community. This will be linked in the
  485. 14:38description. Go to the classroom and
  486. 14:40then grab it in from the AI skills or
  487. 14:42from the all YouTube resources. All of
  488. 14:44my other free GitHub repos, skills,
  489. 14:46prompts, guides, everything is in the
  490. 14:48free school community. So anyways guys,
  491. 14:50that is going to do it for this one. I
  492. 14:51hope that you enjoyed and you learned
  493. 14:52something new. And if you did, please
  494. 14:53give it a like. It helps me out a ton.
  495. 14:55And as always, I appreciate you guys
  496. 14:56making it to the end of the video and
  497. 14:57I'll see you on the next one. Thanks
  498. 14:58everyone.

About this transcript

This page contains the full transcript of /roast-room: How I Use Claude to Think Like the Top 1% by Nate Herk | AI Automation, generated from the public captions YouTube serves with the video. The transcript has 3,577 words across 498 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.