Claude Built the Ultimate Second Brain — Transcript

Full transcript

1. 0:00So, this is my second brain. It holds
2. 0:03all the knowledge that I have,
3. 0:04everything about YouTube videos,
4. 0:06everything about AI labs, the latest
5. 0:08papers, even my sponsors. Everything
6. 0:10that I know gets automatically logged,
7. 0:13connected, and summarized for me in my
8. 0:15second brain. And I didn't write most of
9. 0:19it anyway, really. The AI did. And as it
10. 0:21slowly aggregates all the information,
11. 0:23it's able to parse it, and I'm able to
12. 0:26ask it a question and get immediate
13. 0:27answers based on my contact and all of
14. 0:30the data that I have access to. For
15. 0:32example, these little clusters here are
16. 0:34different YouTube creators. For example,
17. 0:37this node is me. It connects me to all
18. 0:39of my videos, the various topics that
19. 0:41I've talked about, all my content and
20. 0:43performance on YouTube, on X,
21. 0:45newsletters, et cetera. Here's an
22. 0:47example of how I would use it. So,
23. 0:48here's Cloud Code, the desktop app.
24. 0:51Recently, I've started using the desktop
25. 0:52app for pretty much everything. Before,
26. 0:54I was running most of my stuff in the
27. 0:55command line interface. Now, it's pretty
28. 0:57much all here. So, I'm going to ask it
29. 0:59based on the data we have available,
30. 1:00what topics haven't I covered that are
31. 1:02very popular in the AI sphere. And
32. 1:04here's the answer: Grok, Deep Seek,
33. 1:06Copilots, Codex, et cetera. It's taking
34. 1:08into account the velocity of the
35. 1:10different videos, kind of which topics
36. 1:12are trending right now, and pulling from
37. 1:14the entire database to provide the best
38. 1:17possible answer. So, does this from its
39. 1:19own notes, the notes that it wrote and
40. 1:22that it keeps updated every single day.
41. 1:24It knows my sponsor deadlines better
42. 1:26than I do, and it keeps track of a
43. 1:27million different things that I could
44. 1:29not possibly keep track of. This whole
45. 1:32idea came from one of the most respected
46. 1:34names in AI. That is Andrej Karpathy.
47. 1:36And by the end of this video, you'll
48. 1:38know exactly what to do to build your
49. 1:40own. I also had to create some slides
50. 1:42for me, so I can do this presentation
51. 1:44without losing track of what I'm talking
52. 1:46about. So, first and foremost, what is a
53. 1:48second brain? This idea has been around
54. 1:50for some time. It went by many different
55. 1:52names, but the core idea was how do we
56. 1:54get all the stuff that we're dealing
57. 1:56with, all the little notes and ideas and
58. 1:58the to-do lists and meeting minutes and
59. 2:00just everything everything everything.
60. 2:01How do we sort of collect it all, then
61. 2:03organize it all, and have it available
62. 2:06for when it's actually actionable? And
63. 2:08back in the days you would have write it
64. 2:09down in a notebook, and you'd hope that
65. 2:11that page was there when you needed it.
66. 2:13Apps improved it a little bit, you could
67. 2:15write things down, you could save it.
68. 2:16All of us at some point had some system
69. 2:18for the intake of all the information
70. 2:21that life throws at us. But just saving
71. 2:23information, hoarding it, you know, all
72. 2:25the bookmarks you have saved in X. That
73. 2:27was never really the problem. The
74. 2:29problem was trying to maintain it,
75. 2:31trying to stay on top of it, and using
76. 2:33it when you needed to use it. Having
77. 2:35that information available to you when
78. 2:37you need it. So, the first brain, the
79. 2:39one in your skull right now, gently
80. 2:42floating there, it's great at thinking.
81. 2:45It's terrible at storing and organizing.
82. 2:47We forget most of the things that we're
83. 2:49supposed to do, most of the things that
84. 2:51we read. There's a some very small
85. 2:53percentage of the population that have
86. 2:54amazing memory, and they just remember
87. 2:56everything when they need to, and you
88. 2:57know, hooray for them, but that's not
89. 2:59most of us. And then we got LLMs. And
90. 3:03then these LLMs got good. And we're now
91. 3:06at a point where that LLM, well, it can
92. 3:09be a librarian for all of the data that
93. 3:11you have. So, you just throw all of your
94. 3:14notes and data and recordings and
95. 3:15everything everything everything into
96. 3:17this vault. Some of it you do manually,
97. 3:19some of it you set up various collection
98. 3:20processes for that, so it's done
99. 3:22automatically. But you just throw it in
100. 3:24the vault. The librarian, this AI, it
101. 3:27reads everything, it files it correctly,
102. 3:29it puts it on the right shelf, so to
103. 3:31speak. It connects all those things in
104. 3:34some way that makes sense. Some things
105. 3:36you might need for work, some things you
106. 3:38might need for the house, for what I'm
107. 3:40doing, I like to organize it by topic
108. 3:42clusters. And this AI, it keeps all that
109. 3:44information tidy every day, forever, on
110. 3:47autopilot. So, the idea, I got to give
111. 3:49credit to Andre Karpathy. So, he called
112. 3:51it the LLM Wiki. And before that it was
113. 3:53called the second brain, before that
114. 3:54there was a book called getting things
115. 3:56done, the GTD system, and it goes back
116. 3:58even even further. In fact, this
117. 4:00original idea predates LLMs. It predates
118. 4:04even computers. This idea of a machine
119. 4:07that organizes all of your thoughts and
120. 4:09ideas, that comes from 1945. So, as
121. 4:12Andrej Karpathy said, Obsidian is the
122. 4:14IDE. So, the IDE is your development
123. 4:15environment, it's kind of where you
124. 4:17shape software, where you build
125. 4:18software. If you're not familiar with
126. 4:20Obsidian, don't worry, you will be, and
127. 4:22I think you're going to like it. It's
128. 4:23free. Then the LLM is the programmer.
129. 4:26That's the thing that goes in there and
130. 4:28manipulates things and builds things and
131. 4:29creates structure. And the Wiki is the
132. 4:32code base. So, in plain English,
133. 4:33Obsidian, that's this app, just think of
134. 4:35it like a notebook. It's this app that
135. 4:37you look through to see your notes. Then
136. 4:39we have the LLM, the AI. It writes the
137. 4:41pages, interlinks them, and keeps them
138. 4:44updated. And the Wiki becomes sort of
139. 4:46the product, the code base. It's the
140. 4:48thing from which you draw all the
141. 4:49insights and the knowledge. It's a
142. 4:51growing library of interconnected pages.
143. 4:53Your job is to make sure that it's
144. 4:55getting fed with the various raw data
145. 4:57that you want in there, and then on the
146. 4:59other side, you ask it questions or you
147. 5:02have it deliver the insights to you in
148. 5:04some scheduled manner. Here's the thing,
149. 5:06humans, we've been dreaming about this
150. 5:07for a long, long time. This dream is 80
151. 5:10years old. Someone named Vannevar Bush
152. 5:13thought of this in 1945. He originally
153. 5:15called it the Memex, Memex. Memex sounds
154. 5:19cooler, I got to say. So, it's basically
155. 5:20a desk that recorded everything you've
156. 5:22ever read with trails connecting related
157. 5:24ideas. So, we had that idea 80 years
158. 5:26ago, we just never had anything capable
159. 5:29of maintaining this database, making it
160. 5:31useful on autopilot. We do now, it's
161. 5:33large language models. It's ChatGPT,
162. 5:36Grok, Gemini, Claude, you name it. All
163. 5:38right, so first and foremost, what is
164. 5:40Obsidian? Obsidian is a pretty simple
165. 5:43app that has been built on a radical
166. 5:45idea. This is going to blow your mind.
167. 5:47You know, like oh, your notes and the
168. 5:49stuff you write down. What if instead of
169. 5:51you putting it on somebody else's
170. 5:53computer like the cloud somewhere some
171. 5:55large enterprise. What if bear with me
172. 5:57here. What if you just kept them all on
173. 5:59your computer? Like they're your files,
174. 6:02your data, your notes. What if they were
175. 6:03just on your computer? If Obsidian
176. 6:06vanishes tomorrow, all the stuff that
177. 6:08you have saved it's still there. People
178. 6:10call this local first software and it
179. 6:12uses something called markdown. If
180. 6:14you're working for or LMs, you you've
181. 6:16probably heard about markdown. So,
182. 6:18they're markdown files, right? So, that
183. 6:20.md extension like claude.md, skill.md.
184. 6:24Those are your markdown files. Markdown
185. 6:26files are super simple. They're
186. 6:28basically just text plus a few symbols
187. 6:32that do something. You've probably seen
188. 6:34something like this, right? So, you just
189. 6:35type your text. If you want a heading,
190. 6:37you just use one hashtag for heading one
191. 6:40and the big one or two hashtags or three
192. 6:42for heading three. Two asterisks around
193. 6:44something makes it bold. One asterisk
194. 6:46around it makes it italicized. And
195. 6:48simple ways to add code blocks or links,
196. 6:50etc. It's super super simple. So, a
197. 6:5310-year-old can learn this in 10
198. 6:55minutes. So, this right here is
199. 6:57Obsidian. This is kind of what it looks
200. 6:59like. And this is the graph view. It's
201. 7:02pretty cool kind of a way to visualize
202. 7:04all of the things all of the topic
203. 7:06clusters, whatever you have saved in
204. 7:08there. Each one of these little dots is
205. 7:10a file. So, for example, here's this
206. 7:12cluster. Let's zoom in and see what this
207. 7:13is. It's my sponsor content flow with my
208. 7:17sponsors. These sponsors are not real.
209. 7:20They're fake. I can't actually show you
210. 7:22the real data. So, I had Cloud Code to
211. 7:24come up with some fake data just to be
212. 7:25able to kind of a showcase the idea, but
213. 7:27this is more or less exactly the system
214. 7:29that I use to keep track of things. Each
215. 7:31sponsor gets its own node and then all
216. 7:34the information that is needed is
217. 7:35uploaded to that. Based on that, we work
218. 7:37out what needs to be done, when the due
219. 7:39date is, any assets that I need to know
220. 7:42about is in there. For me, it's kind of
221. 7:44like homework. You know, if you think
222. 7:46back to the days you went to school, you
223. 7:47got homework, and you know, you hoped
224. 7:49that you remember about it and when the
225. 7:51due date is. I had struggles with that
226. 7:53sometimes. I would forget to write down
227. 7:55the due date. I would forget when things
228. 7:57were due. And sometimes I had trouble
229. 7:59kind of like breaking up the work into
230. 8:01manageable tasks, so that slowly over
231. 8:03time you completed the project. So, if
232. 8:05you've ever had trouble with the same
233. 8:06thing or paying the bills on time or
234. 8:08filing some specific paperwork that you
235. 8:10need to file, that's sometimes referred
236. 8:12to as the ADHD tax. And for a lot of
237. 8:14people, this is a very real thing that
238. 8:16they struggle with on a daily basis. If
239. 8:18you're fortunate enough to be able to
240. 8:20afford an executive assistant or
241. 8:22somebody that just like handles those
242. 8:23things for you, that's great. For most
243. 8:25of us, there wasn't an easy solution for
244. 8:27most of our lives until recently. Now, I
245. 8:30just have to find a way to get all those
246. 8:32important things and funnel them into my
247. 8:35second brain. Then, I work with my
248. 8:37favorite assistant, Claude Code or
249. 8:39ChatGPT, or whatever, to then set up
250. 8:41systems, automated systems that make
251. 8:43sure that I get notified, "Hey, this is
252. 8:46coming up. Maybe you should start
253. 8:47working on this." Also, I don't have to
254. 8:49open five different tabs and hunt for
255. 8:51different pieces of information all over
256. 8:53the place. Everything's connected. Let's
257. 8:55zoom out and I'll give you another
258. 8:56example. For example, let's zoom over
259. 8:58here. Each one of these things is a
260. 9:00paper or blog post from a Frontier AI
261. 9:02lab. They're something that I've talked
262. 9:04about in the previous videos that I
263. 9:06might need to talk about in the future.
264. 9:07So, for example, there's a page about
265. 9:08AlphaGo and Tree of Thoughts. Some of
266. 9:11you might have been following me from
267. 9:12those days, years and years ago, when we
268. 9:15covered Tree of Thoughts. Who remembers
269. 9:17that? And all of these are
270. 9:18interconnected and they're linked. So,
271. 9:20anything that has to do with meta AI or
272. 9:22Demis Hassabis or Sam Altman, they're
273. 9:24all connected to each other through
274. 9:25these nodes. If you're wondering what
275. 9:27this mess of a cloud is, these are the
276. 9:29various people that publish about AI.
277. 9:32For example, this is me. All those
278. 9:34little purple lines point to things that
279. 9:36are connected to me. All of my videos,
280. 9:38recently I started cataloging some of my
281. 9:40tweets, although I don't think this is
282. 9:41hooked up to that yet. We're We're in
283. 9:43the process of doing that. It also
284. 9:44connects to topics that I've talked
285. 9:47about. So, when we ask the question
286. 9:48like, "What topics haven't I talked
287. 9:51about in the last, whatever, 6 months?"
288. 9:52It has all that data. It's not guessing.
289. 9:55It's not going online and searching. It
290. 9:57knows. All the data is here. It also
291. 9:59knows what everyone else is posting. And
292. 10:01the notice that goes to a number of
293. 10:04these lines here. That connects all of
294. 10:05those videos, for example, to the topics
295. 10:08that they discuss, to the analytics
296. 10:10behind those videos, to what works, what
297. 10:13doesn't. And these red dots here, that's
298. 10:15what Claude decided to call beats, like
299. 10:18on the beat or my beat. By the way, I
300. 10:20have Claude naming a lot of these
301. 10:21things, so some of them look a little
302. 10:23bit weird, like it decided to call
303. 10:24something the armory. The armory is all
304. 10:27the things that it thinks I should
305. 10:28build. Things that would be useful and
306. 10:31helpful to me, but I need to sit down,
307. 10:33you know, plan it out, tell Fable or
308. 10:35whatever model I'm using to go ahead and
309. 10:37build it out. These are the things that
310. 10:38are on that list, the priority board,
311. 10:40analytics, demon, X wide funnel,
312. 10:43packaging lab, first responder pipeline,
313. 10:45retention miner, comment archive, clip
314. 10:47engine. Now, if you're wondering, "What
315. 10:49are these projects that they're talking
316. 10:50about?" For example, this is one of
317. 10:52them. So, this is what we call the X
318. 10:54data ingestion engine. We're getting
319. 10:56tons of data from X/Twitter about the
320. 10:59performance of various tweets from my
321. 11:01own account as well as some other
322. 11:03people's accounts. So, step one was to
323. 11:05build the engine, sort of the data
324. 11:07collection engine. By the way, if you're
325. 11:09wondering, "Oh, are you going to show us
326. 11:10how how you did that? I I I also want to
327. 11:12know how to do that." Yeah, sure. I
328. 11:14opened up Fable 5 on high effort. By the
329. 11:17way, this is, I think, one of the best
330. 11:19ways to use it. I found that this is
331. 11:20kind of the sweet spot. Not extra high,
332. 11:23not ultra Fable 5 high. So, I opened it
333. 11:25up and I said, "This is what I want.
334. 11:28Tell me what you need from me, what kind
335. 11:29of API keys, what services should I sign
336. 11:31up for, and then go build it. Once that
337. 11:34X engine is built, on top of that you
338. 11:36build the things that are actually
339. 11:38useful to you. So, for example,
340. 11:39something that alerts you when a new
341. 11:42trending topic is developing. Or, in
342. 11:44this case, as you can see, we sort of
343. 11:45broke down how well different format of
344. 11:48tweets work. What if it's a standalone
345. 11:50tweet with a native video, or a quote
346. 11:52and a video clip, quote plus image,
347. 11:54quote or link, or just bare text. What I
348. 11:56realized by looking at the data is that
349. 11:58the Twitter algorithm changed a few
350. 12:01months ago, and I didn't realize it. And
351. 12:03so, what that meant was that my
352. 12:05impressions used to keep going up and up
353. 12:07and up month over month. That huge line
354. 12:10in January 26, that was a few viral
355. 12:13hits, so that's not really kind of
356. 12:14representative, but that was a good
357. 12:16month. One of those tweets was shown in
358. 12:18a Fireship video. So, yes, as seen on
359. 12:21Fireship. He was a little bit sarcastic
360. 12:23about the tweets, but he's a little bit
361. 12:24sarcastic about everything, so and
362. 12:27totally love that guy. So, I I I I was
363. 12:29just happy to be mentioned. But, notice
364. 12:31there's a steep drop-off, right? You can
365. 12:33look at it, you can see it. You know
366. 12:36something happened. What? So, as you can
367. 12:38see here, I have one of this these
368. 12:40folders, X analytics. So, we have all of
369. 12:42our data that we're ingesting. Ingesting
370. 12:45is a special word that we use here to
371. 12:46basically say collect the data, take the
372. 12:48data into the vault, into our second
373. 12:51brain. Not just using that word because
374. 12:53I'm hungry, that's the correct
375. 12:55terminology here. We have all our
376. 12:56important accounts that we're keeping
377. 12:58track of a weekly scorecard of how well
378. 13:01my tweets are performing, and also the X
379. 13:04engine. The X engine is the actual thing
380. 13:06that runs it. So, right now we're in the
381. 13:08second brain, this is the amount of data
382. 13:11that we have. 22,000 posts archived,
383. 13:14almost 6 billion views represented,
384. 13:164,400 unique authors. There's a lot
385. 13:20there. How quickly can I reference one
386. 13:22of them, quickly pull out some piece of
387. 13:24information that I need? Instant. It's
388. 13:25offline, no API needed. How much did
389. 13:28this cost? Under 100 bucks. I don't know
390. 13:30the exact number. I know it's under 100
391. 13:32cuz I purchased 100 dollars in credits
392. 13:33and that was enough. I just don't know
393. 13:36exactly how much it spent. And also,
394. 13:38this isn't a static database. It's being
395. 13:40watched. So, for example, it lets us
396. 13:42compute velocity of the trending topic
397. 13:45or tweet, which post is getting 600
398. 13:46likes an hour right now, tracking
399. 13:48breaking AI news. It's also seen which
400. 13:51strategies started breaking down at that
401. 13:54algorithmic change that we saw a few
402. 13:55months ago. And it's benchmarking me
403. 13:58against some of the other accounts,
404. 14:00right? So, if something that I'm doing
405. 14:01is underperforming, it allows me to
406. 14:03pinpoint exactly what it is that I'm
407. 14:06doing wrong. And notice that it's
408. 14:08storing all these insights and updating
409. 14:10them and curating them. And it's a
410. 14:12living document which Claude decided to
411. 14:14call X growth playbook. Now again, the
412. 14:17reason I wanted to bring that up was
413. 14:18because I wouldn't have called it that
414. 14:20necessarily. The point isn't let's grow.
415. 14:23It's not a growth playbook. It's not a
416. 14:25growth hacks. I'm thinking of it more as
417. 14:26a don't shoot yourself in the foot, Wes,
418. 14:29playbook. The whole point is basically
419. 14:31to understand the kind of how the
420. 14:32algorithm functions, if it changes, so I
421. 14:35don't get caught up in the changes and
422. 14:37just lose all my views, etc. So, the
423. 14:40point isn't growth hacks. The point is
424. 14:42what are the best practices right now?
425. 14:44By the way, the next level up, I think,
426. 14:47is to turn it into something like this.
427. 14:49This is kind of what I'm building right
428. 14:50now. This also brings in ideas like, for
429. 14:53example, which apps are connected to
430. 14:57Claude. So, I have things like Obsidian,
431. 14:59Social Blade, X, the X API, etc. etc. As
432. 15:03your little sort of branching empire
433. 15:05starts growing a larger and larger, it
434. 15:07really helps to have just one place
435. 15:08where you can at a glance see, okay,
436. 15:10what are all of the things that are
437. 15:11connected? What are all the apps and
438. 15:13APIs that are hooked into the system?
439. 15:15This is pretty important from a lot of
440. 15:17different angles. Security, doing basic
441. 15:19security checks, it's important to have
442. 15:20this visualized somewhere. Saving money,
443. 15:23right? You can see at a glance what
444. 15:24you're paying for, what services you
445. 15:26need to cancel. By the way, all these
446. 15:27systems are getting pretty good at
447. 15:29actually doing computer use, running
448. 15:30their own browser. So, at some point,
449. 15:32they'll be able to cancel out of the
450. 15:34recurring billing for us. I've already
451. 15:35been testing it, trying to use the
452. 15:36browser within Claude code to do certain
453. 15:38tasks. It's pretty good so far. And I'm
454. 15:41planning to start ramping it up more and
455. 15:43more. Then we have our routine. So,
456. 15:44these are actually the things that are
457. 15:46running on a daily basis. So, if you
458. 15:48want to see all of the things that are
459. 15:49running the cron jobs, the things
460. 15:51ingesting new data into the second
461. 15:53brain, all those routines, everything is
462. 15:55there. Then we also have our various
463. 15:57skills. So, those are like the skill.md
464. 16:00files and everything, everything,
465. 16:02everything. Now, if you're wondering
466. 16:03what this doctrine is, again, I have
467. 16:05Claude naming a lot of these things. So,
468. 16:07bear that in mind. So, I told it to put
469. 16:09all of the things like the learnings
470. 16:10about the X algorithm, all of the sort
471. 16:13of final insights, all of the juicy
472. 16:15information that we're squeezing out of
473. 16:17this thing into a folder. I'm like,
474. 16:18"Call it something cool." And I was
475. 16:20like, "Oh, I know, the doctrine." And I
476. 16:22was like, "All right, whatever, Claude."
477. 16:23All right, so so far we've talked about
478. 16:24our first tool that you need. It's
479. 16:26Obsidian. That's this on the left. It's
480. 16:29free. It's wonderful. It's local-first.
481. 16:32And Obsidian is basically just a bunch
482. 16:34of markdown files. So, again, those
483. 16:36markdown files is just text plus a few
484. 16:38simple symbols that make it functional.
485. 16:41And it's really good for cross-linking
486. 16:43everything. So, for example, here is the
487. 16:45Karpathy's LLM wiki. That's kind of the
488. 16:48idea that kicked this whole thing off.
489. 16:49So, let's click on it. This is part of
490. 16:51the wiki. It's the database around that
491. 16:54subject, that topic. And markdown is
492. 16:56super simple. So, let's say I wanted to
493. 16:58add that he worked at OpenAI. We'll do
494. 17:00two hashtags for heading two. I'm going
495. 17:02to say, "Used to work at" and notice how
496. 17:05it turns into heading two. And I'll say,
497. 17:07"Andre used to work at" and I'm going to
498. 17:09say, "OpenAI." But, I will cross-link
499. 17:11those two documents. I'll do double
500. 17:13brackets. Notice how it pre-fills the
501. 17:15other two double brackets. Now,
502. 17:16everything that you type between those
503. 17:18two becomes a link. So, I'll type in
504. 17:20open AI. Notice it already gives me all
505. 17:22of the other pages that we have on open
506. 17:23AI the topic, open AI the entity, and
507. 17:26various transcripts that include open AI
508. 17:29in the title. So, in this case, we'll
509. 17:30say open AI the entity, and now that
510. 17:33links to that page. So, you do this
511. 17:35enough times and these stop being just
512. 17:38pages and they become a network. By the
513. 17:41way, when you don't have any tabs open,
514. 17:42if you hit control G, that opens up the
515. 17:45graph view that lets you visualize that
516. 17:48entire network. And if you hit animate
517. 17:49here, you can kind of see how page by
518. 17:52page by page through cross-linking the
519. 17:54whole thing takes shape. As you add more
520. 17:56and more data, more and more pages, both
521. 17:59raw pages that are just from the
522. 18:01internet or from whatever data you're
523. 18:02pulling in to actual summaries that are
524. 18:04made by the LLM to all of the different
525. 18:07stuff that you're adding to it, this
526. 18:08slowly becomes that kind of knowledge
527. 18:10graph. It slowly becomes your second
528. 18:13brain. Once you build this whole thing
529. 18:15and you hit that animate button, this is
530. 18:16just kind of rewarding just watching all
531. 18:19that information slowly come together.
532. 18:20We're not going to watch it cuz I have
533. 18:22too much stuff in here. It'll take
534. 18:23forever. But, this is your tool number
535. 18:25one, Obsidian. And your second tool is
536. 18:28Claude Code or chat GPT or Codex. Now,
537. 18:31if you've been following this channel,
538. 18:32you've seen me use these models through
539. 18:34a lot of different interfaces. For a
540. 18:36long time, I dealt more or less
541. 18:37exclusively with open Claude. I would
542. 18:39use Telegram to talk to it. I've used
543. 18:40the command line interface, tons of
544. 18:43different ways of interacting with it.
545. 18:44Currently, now with this new iteration
546. 18:46of Claude Code Desktop, which is what
547. 18:48you're seeing here, at this point, I'm
548. 18:50pretty much exclusively using this. They
549. 18:52added a lot of functionality to where
550. 18:53you really don't need to leave this at
551. 18:56all. It has a Claude Code. You can
552. 18:57switch over to the home tab, which has
553. 18:59your regular kind of ability to talk to
554. 19:00Claude, as well as Claude co-work. Or
555. 19:03you can stay in code and do a lot of
556. 19:05coding kind of on this side. So, what I
557. 19:07did was I created a second brain
558. 19:09directory or folder, and I just told
559. 19:11Claude to build everything in there. So,
560. 19:13now whatever new information we're
561. 19:14ingesting, it finds a place somewhere in
562. 19:16there. So, for example, recently
563. 19:18Anthropic released this, a global
564. 19:20workspace in language models. So, it's
565. 19:22basically talking about if Claude could
566. 19:25be conscious on some level, or they're
567. 19:27not suggesting that that's what's
568. 19:28happening. They're just finding a lot of
569. 19:30very interesting similarities in how
570. 19:32Claude's brain works and how LLMs work.
571. 19:35In some ways, it's very similar to how
572. 19:36the human brain works. So, this idea of
573. 19:38a global workspace is something that
574. 19:40exists in human brains. It's a mechanism
575. 19:42by which we sort of find things that are
576. 19:43unconscious and kind of bring it to the
577. 19:45surface so that we're able to interact
578. 19:46with it in our brains. And they're
579. 19:47finding something that is analogous or
580. 19:49similar in Claude. So, definitely kind
581. 19:51of a big deal of a of a publishing of a
582. 19:54paper. So, we want to ingest this into
583. 19:56our second brain. By the way, a lot of
584. 19:57this should be handled automatically.
585. 19:59Here, I'm just showing you how you would
586. 20:01deal with it, how you would do it
587. 20:03manually if you needed to. So, I'm going
588. 20:04to take this URL, or just copy this and
589. 20:07paste it, and we're going to go into
590. 20:08Claude, and we're going to say ingest,
591. 20:09and I'll just paste the link, and we'll
592. 20:12click go. Another really good feature of
593. 20:14Claude Coda Desktop is you can just
594. 20:16dictate your commands. Click this
595. 20:17microphone button and just say what you
596. 20:19want it to do. Now, by the way, one
597. 20:21recent thing that they've added is an
598. 20:24actual built-in browser. So, if you
599. 20:25click on that, you can actually just
600. 20:27type in whatever URL, and it will open
601. 20:29within this built-in browser within
602. 20:31Claude Coda Desktop. So, I can go to
603. 20:33google.com, for example, and I can
604. 20:35actually tell it to open up web pages,
605. 20:37interact with those web pages, whatever
606. 20:38you want. But here, we'll actually open
607. 20:40up a file. This is my second brain, just
608. 20:42a folder with a number of other folders
609. 20:44in it. At the bottom, I had to create
610. 20:46this. So, that is just this, this kind
611. 20:48of a visual representation of kind of
612. 20:51like the second brain 2.0 I'm trying to
613. 20:53build that is going to have all the
614. 20:54skills and routines and everything else
615. 20:56on top of it with a different
616. 20:57visualization. And notice here, as it's
617. 20:59building out, ingesting that content
618. 21:01from the Anthropic website, it's saying
619. 21:03now the ripple. So, they're
620. 21:04cross-linking all these pages, adding
621. 21:07more information about it. So, they're
622. 21:08adding it to the interoperability
623. 21:10concept page and updating all the other
624. 21:12entity pages. So, I give it one link, it
625. 21:15adds it, and now it's rippling through
626. 21:17and adding it and interconnecting it
627. 21:19within the network. All right, so that's
628. 21:20how we ingest information. That's how we
629. 21:23add information to our second brain. All
630. 21:24right, but this is where it stops being
631. 21:26just a research engine and starts kind
632. 21:29of running my life because your second
633. 21:31brain shouldn't just know about the news
634. 21:34and what's going on in the world, it
635. 21:35should know about your life. So, you've
636. 21:37probably heard about the Kanban board.
637. 21:39So, it's usually something that you have
638. 21:41maybe like on the wall, you have sticky
639. 21:42notes, and you move those sticky notes
640. 21:44from place to place. Each sticky note is
641. 21:46a project or a to-do item that kind of
642. 21:49goes through stages. So, maybe going
643. 21:50from to do to doing to done. In
644. 21:53Obsidian, it's very easy to create a
645. 21:55Kanban board. So, for example, we might
646. 21:57have a flow like this. If we're doing a
647. 21:59content calendar where videos get
648. 22:01produced from idea to research to
649. 22:04scripted to filmed, edited, and
650. 22:06published. Now, currently my process of
651. 22:08creating videos is a lot more chaotic,
652. 22:11let's say. And also, I don't script
653. 22:13them, and I apologize if I'm stating the
654. 22:15obvious. If you've ever seen me go on
655. 22:17some wild tangent and forget my original
656. 22:19idea, you probably can tell that none of
657. 22:21this is scripted. But now, to try to
658. 22:24keep up with the sheer amount of
659. 22:25information releases, I am trying to be
660. 22:27a little bit more organized about how I
661. 22:29release things, having certain ideas,
662. 22:31some from me, some that Claude or some
663. 22:34other chatbot comes up with
664. 22:35automatically based on the information
665. 22:36available on the trending news. So, you
666. 22:38might have tons of ideas ranked by some
667. 22:41metric, how relevant it is, how
668. 22:43interesting it is. So, let's say I want
669. 22:45to create one of these. So, recently I
670. 22:47published a video called the $20,000
671. 22:49revenue apps with one-person teams or
672. 22:52something like that. So, I would pick it
673. 22:54out of my list of ideas and I would move
674. 22:55it to kind of this packaging gate. So,
675. 22:58if it scores good on some metric about
676. 22:59how viable it is as a video idea, so it
677. 23:02gets put there. Once it's scored, we can
678. 23:04move it to research. And by the way, a
679. 23:06lot of this stuff can be automated, so
680. 23:07if I move it there, Claude can go ahead
681. 23:10and start working on it. So, here as you
682. 23:12can see, Claude already wrote some
683. 23:14suggested hooks for me. The first one is
684. 23:16"3 years ago I showed you a dad selling
685. 23:19Excel formulas for $25,000 a month. The
686. 23:21number today made me double-check my
687. 23:23sources." In that video, I used the hook
688. 23:26about 6 minutes in. The first 6 minutes
689. 23:28was me rambling. And then after 6
690. 23:30minutes or so, I got to the hook. Claude
691. 23:32tries and does a great job. I still find
692. 23:35ways to mess it up, but that's on me.
693. 23:36So, let's say once we've done all the
694. 23:38research, we move that to, you know,
695. 23:40scripting the video. Now again, I don't
696. 23:42script my videos, but I do like to have
697. 23:44these little cheat sheets with the
698. 23:46numbers and the claims, dates, things
699. 23:49like that written out. That ensures that
700. 23:51what I say on camera is accurate. So, I
701. 23:53tell Claude that I did a video 3 years
702. 23:55ago about this thing. I want to do a
703. 23:57follow-up. So, keep in mind, it has the
704. 23:59transcript of the video that I did 3
705. 24:01years ago. That's in the vault. It knows
706. 24:03every word I said on that video. Take a
707. 24:05look at this. We covered a product back
708. 24:07then, 3 years ago in 2023, that was
709. 24:09doing 20,000 a month. It was called
710. 24:11thumbnailtest.com, and it was AB testing
711. 24:14thumbnails. By the way, and this is why
712. 24:16I love Claude. He's insufferable. Look
713. 24:19at that. It says, "The thing your war
714. 24:21room now does for free." With that
715. 24:23grinning kind of a smiley face, like
716. 24:24it's up to something. So, it built that
717. 24:26AB testing thumbnail software for me,
718. 24:28and this is it. Just being kind of smug
719. 24:30about it. It's like, "Oh, yeah, like I
720. 24:31built that thing for you." I, as you can
721. 24:33imagine, did not ask for that to be in
722. 24:36the show notes, in the in the thing that
723. 24:38I'm going to use to prepare for my
724. 24:39video. For Claude to be like, "What's
725. 24:41up?" That was not asked for. But, notice
726. 24:43what it did here. So, it found what
727. 24:45happened to that case study that I did
728. 24:47in 2023. What happened to
729. 24:48thumbnailtest.com? Is it still making
730. 24:5020,000? Is it making more? It found that
731. 24:52it sold for six figures in 2024. By the
732. 24:55way, since then, YouTube actually
733. 24:57launched their own internal thumb
734. 24:59testing split testing tool. And as
735. 25:00Claude is saying here, the platform ate
736. 25:02the moat. And notice what it's saying
737. 25:04here, this is the exact platform risk
738. 25:07warning from your 2023 video. When
739. 25:09OpenAI announced Whisper, everyone
740. 25:11building that was gone. So, this is kind
741. 25:13of why having a second brain like this
742. 25:15is so important because it's going back
743. 25:17and checking my notes from 3 years ago.
744. 25:19It's also updating it from doing
745. 25:22internet search, kind of seeing what
746. 25:23happened since then to now. It's doing
747. 25:26all of that while while being smug about
748. 25:27it. What's not to love here? So, while I
749. 25:30don't use the content calendar in that
750. 25:32Kanban style dashboard, I'm planning to
751. 25:34do that a little bit more to kind of
752. 25:36automate more of the research and
753. 25:38information gathering. But, here is a
754. 25:41sponsor flow Kanban board. This I
755. 25:44actually do use to help me visualize
756. 25:46where I am in the process. These are
757. 25:48dummy names, kind of dummy sponsors,
758. 25:50they're not real. I can't put the actual
759. 25:51sponsors in there because often times
760. 25:53there's non-disclosure things. So, I I
761. 25:56can't use the real sponsors, but this is
762. 25:57literally what it looks like. We have
763. 25:59the script, the the sponsor approval,
764. 26:01recording, editing, and all the way once
765. 26:04it's approved into publishing. As I get
766. 26:06approvals, I just drag it over and this
767. 26:08updates its status. Once it's published,
768. 26:10I put it into the done category and I'm
769. 26:13done. This, by the way, can be very
770. 26:14easily hooked into some sort of a system
771. 26:16that notifies you on your phone through
772. 26:18a text message or email if you're
773. 26:20running behind on something. If you're
774. 26:22keeping up with things like this through
775. 26:23Slack, for example, it pull that
776. 26:25information in here as well. And
777. 26:27finally, it brings us to maybe the most
778. 26:29important piece of this whole thing,
779. 26:31kind of the point of the second brain.
780. 26:34It's called the doctrine. And again, I
781. 26:36have to remind you here, I I don't come
782. 26:37up with these names. This is all Claude.
783. 26:39I think it knows that I like those RPG
784. 26:41games, so it tries to kind of flavor
785. 26:43everything in that style. So, as it's
786. 26:46wrote here, right? So, this is the
787. 26:47output layer of the second brain. So, we
788. 26:49have raw data flowing into it, the wiki
789. 26:52organizes everything that's known, and
790. 26:54the doctrine is what comes out the other
791. 26:55end. It's Fable Analyze. So, this is
792. 26:57done by Fable 5, which I found is
793. 26:59incredibly good at this kind of deep
794. 27:02data analysis and coming up with
795. 27:04insights. So, it's Fable Analyze
796. 27:06receipts backed actionable strategy.
797. 27:08Every doc here answers, "What do we
798. 27:10actually do?" And every claim in here
799. 27:12traces back to the data that we've
800. 27:14collected. So, the war room gathers
801. 27:15intelligence. So, I have this mini PC
802. 27:18that's always on. So, it's kind of like
803. 27:19a Mac Mini. And it just sits there. It's
804. 27:21hooked up to Wi-Fi. It doesn't take up a
805. 27:23lot of electricity. It doesn't take up a
806. 27:24lot of room. It just kind of looks like
807. 27:26this, and I think it cost about 200
808. 27:28bucks on Amazon. And it runs 24/7. It
809. 27:30never turns off. It doesn't have a
810. 27:32screen saver. It's just like a little
811. 27:33box that's always on. I mean, so that's
812. 27:35sort of the war room, if you will. It
813. 27:37kind of just sits there, collects data.
814. 27:39It's looking at what's happening on
815. 27:40YouTube, on X, on various news
816. 27:43platforms. It's the 24/7 kind of home of
817. 27:46the agents that just gather data. Then
818. 27:48the Wiki remembers it, organizes it,
819. 27:51cross-links it, all the stuff that we
820. 27:53talked about before. And the doctrine
821. 27:54decides how we fight. Again, I'm sure we
822. 27:56could have used some corporate speak to
823. 27:58make these names and describe what they
824. 28:00do, but I think I would just like fall
825. 28:02asleep here at my keyboard. And then the
826. 28:04armory tracks what we're building next.
827. 28:06So, those future projects, those
828. 28:07nice-to-have, that that's all in there.
829. 28:10Largely selected and suggested by Fable.
830. 28:13Now, of course, at the end of the I'm
831. 28:14the one that's choosing what to focus
832. 28:16on, what to do, but a lot of the heavy
833. 28:18lifting, the analysis, the data
834. 28:20collection, all of that is handled by
835. 28:22Claude. By the way, the next big step
836. 28:24will be once we have kind of like our
837. 28:27to-do actions from the doctrine, we're
838. 28:28going to execute on them and collect
839. 28:31data about how it works. So, next, let's
840. 28:33say few quarters, 6 months, 12 months,
841. 28:35whatever. That will become its own sort
842. 28:37of flywheel, where we're putting
843. 28:38together strategies, we're executing on
844. 28:41them, we're seeing the results, and
845. 28:42we're updating in real time how well
846. 28:44it's working. So, the longer it runs,
847. 28:47the more it compounds, not just in terms
848. 28:48of the sheer data that's coming in, but
849. 28:50also in terms of the the learning that
850. 28:52the system is doing, both in terms of of
851. 28:54just what it knows, but also of making
852. 28:57strategies, executing them, and and
853. 28:59getting feedback. So, sort of that OODA
854. 29:00loop. So, for those who are not
855. 29:02familiar, so observe, orient, and
856. 29:04decide, and act, and then it becomes a
857. 29:06loop. So, observe is the data
858. 29:07collection, orient is the wiki and the
859. 29:09summaries, and in fact the the doctrine,
860. 29:11then deciding is like kind of like what
861. 29:13we're doing with that the act is the
862. 29:15actual action, the execution of that
863. 29:17strategy, and then we're taking that
864. 29:19data and we're adding it into the OODA
865. 29:21loop. By the way, since Fable designed a
866. 29:24lot of this, even if Fable does go away
867. 29:26eventually, we don't get it back, a lot
868. 29:28of the stuff that it's built will still
869. 29:29be helpful. So, a lot of this doesn't
870. 29:31necessarily rely on Fable to to
871. 29:33continue. A lot of the data of
872. 29:34collection is automatic. But, think
873. 29:36about this, as time goes on, this
874. 29:39system, what happens as better and
875. 29:41better models come out? Does the system
876. 29:43become better, worse, or stay the same?
877. 29:46I think we can safely say that the
878. 29:47system not only just gets better the
879. 29:48longer it runs, it also gets better and
880. 29:50better with stronger and smarter models
881. 29:53being released and and used to run the
882. 29:55system, to to improve the system. So,
883. 29:57let me show you how to build this for
884. 29:59yourself. And my advice to you is take
885. 30:02the time to do this. This might take
886. 30:03some time to set up. Maybe there's going
887. 30:06to be some new skills that you have to
888. 30:07learn. Learning can and probably should
889. 30:09be a little bit uncomfortable. There's a
890. 30:11certain feeling that comes with doing
891. 30:12new stuff. It's not just like pure joy.
892. 30:14There's there's a little bit of a
893. 30:16difficulty of a resistance. Just push
894. 30:18through that. Build this because once
895. 30:19it's in place, it starts compounding. It
896. 30:22starts growing. I honestly wish I did
897. 30:24this on day one. When Eric Karpathy
898. 30:26talked about it, I knew it was a good
899. 30:27idea. I should have jumped on it right
900. 30:29then and there. All right, so this is
901. 30:30how you build this for yourself. I don't
902. 30:33want to say it's super fast. Some of
903. 30:34these steps take time. For some of them
904. 30:36you have to wait for for Claude to build
905. 30:39some of it, to organize some of it. But,
906. 30:40you can probably do this in a single
907. 30:42afternoon. So, first and foremost, you
908. 30:44need two tools, Obsidian and Claude
909. 30:46Code. So, Obsidian is the note-taking
910. 30:49app, although it's a little bit more
911. 30:50than that. So, it's over here,
912. 30:51obsidian.md. Here it is, again, free to
913. 30:54start, most of it is free, it has a huge
914. 30:56community. It's a pretty cool tool, if I
915. 30:59do say so myself. There's a lot to like
916. 31:01here, and it's free without limits, no
917. 31:03sign-up required, no strings attached.
918. 31:05It's a cool tool by by cool people.
919. 31:08Then, get Claude Code Desktop. Again,
920. 31:10you don't have to get the desktop app.
921. 31:12If you're already settled in certain
922. 31:13routine, you know what you're doing, do
923. 31:15that. But, I got to say, if you haven't
924. 31:17tried it, they've really been making a
925. 31:18lot of good strides with it, and it does
926. 31:20seem like it's becoming that super app
927. 31:23that we've been waiting for. And I don't
928. 31:24know, that that might be the the final
929. 31:27form. At least for me, I'm really
930. 31:28wondering what else they can do to to
931. 31:30improve on it. Like, if you haven't
932. 31:32realized it, this is it. I'm using the
933. 31:34browser within it to search for the
934. 31:36stuff that I need. I can even ask Claude
935. 31:38to go and download and install it. By
936. 31:40the way, if I need to take this on the
937. 31:41road and use it from my phone, I just
938. 31:43type in {slash} remote control, I hit
939. 31:46enter, and then that allows it for me to
940. 31:48use it from the Anthropic or Claude app
941. 31:50on my phone. And as they say here, also
942. 31:52to view and control the session from
943. 31:53claude.ai/code.
944. 31:55So, you're able to remote control this
945. 31:57from anywhere. All right, so, you got
946. 31:59Obsidian, you got Claude Code, both are
947. 32:01free to start. My recommendation is you
948. 32:04do purchase a subscription, either
949. 32:06Anthropic or OpenAI, or whatever chatbot
950. 32:08you think is best. But, at this point, I
951. 32:10feel like you kind of need one, if you
952. 32:13understand kind of the significance of
953. 32:15what these companies are are building. I
954. 32:17would say it's time to invest if you
955. 32:19don't yet have a subscription. Then, we
956. 32:20make the vault. We do that by opening up
957. 32:23Obsidian. When you open up the first
958. 32:24time, the button is create new vault or
959. 32:26something like that, and that will get
960. 32:28you started. Inside, you make three
961. 32:30folders, inbox, raw, and wiki. Inside
962. 32:33the wiki, you can make folders like
963. 32:34concepts, entities, summaries, plus two
964. 32:36empty notes, index, and log. Here's the
965. 32:38thing, I didn't do any of this. I told
966. 32:40Claude to build me this thing. It made
967. 32:43all of the folders, all of the files,
968. 32:45everything, everything, everything. By
969. 32:46the way, quick note, notice how flat the
970. 32:48structure is. So, we don't have 50
971. 32:50subpages beneath each page. Everything's
972. 32:53pretty flat. This is not me being lazy
973. 32:55or Claude being lazy. This is by design.
974. 32:58Also, notice that the folders aren't
975. 33:00topics. So, there isn't a folder called
976. 33:02AI news. The folders are the different
977. 33:04layers. What goes in, what it knows, and
978. 33:07what it concludes. In fact, some of
979. 33:09these things like templates, I shouldn't
980. 33:10even have it on here. And the topics,
981. 33:12those topics, they live in the links.
982. 33:14So, OpenAI is a topic. It's the link
983. 33:17that we use to cross-link all of the
984. 33:18different pages that have anything to do
985. 33:21with OpenAI. Creating too many
986. 33:23subfolders, those kind of a deep nested
987. 33:25structures, it becomes a nightmare for
988. 33:27LLMs. Keep it very, very flat. Next step
989. 33:30to creating this would be to write the
990. 33:32rulebook, aka Claude.md.
991. 33:34Now, again, I didn't write this, Claude
992. 33:36did. By the way, I'll have a template
993. 33:38down below that you can just download
994. 33:40and give to your agents, and it will
995. 33:42execute everything for you. But the
996. 33:44Claude.md file, that's the rulebook.
997. 33:47That's the employee rulebook. Every
998. 33:48morning, Claude wakes up and reads the
999. 33:50rulebook and goes to work. So, for
1000. 33:52example, the raw files, those are the
1001. 33:53immutable source documents. So, those
1002. 33:55are the things directly from the source.
1003. 33:57We don't change them. The wiki is the
1004. 33:59LLM wiki. The summaries, the entities,
1005. 34:01concepts, those are the compounding
1006. 34:03knowledge maintained by our AI
1007. 34:05librarian. So, just start there. Later,
1008. 34:08you can add all of the other things like
1009. 34:10I added the doctrine, etc. The inbox are
1010. 34:12quick captures for me waiting to be
1011. 34:14processed. So, this is going to have to
1012. 34:16be in a different video, but there are
1013. 34:17ways to do, for example, voice notes
1014. 34:19where you dictate something or certain
1015. 34:20emails, or even creating a little Chrome
1016. 34:22plugin to whenever you see something
1017. 34:24that you want to add to this, you just
1018. 34:25talk it into your phone, or you just
1019. 34:27record your voice, or you just click a
1020. 34:29button so that it goes to the inbox and
1021. 34:31then later gets processed by Claude.
1022. 34:33Then, the next step is optional, but if
1023. 34:35you wanted to have that graph, that data
1024. 34:37view, that's a plugin in Obsidian. Same
1025. 34:40with the Kanban board. The next step is
1026. 34:42optional, and that's adding two plugins,
1027. 34:45Dataview and Kanban. Dataview builds
1028. 34:47those automatic tables. Kanban is that
1029. 34:49Kanban view where you drag those
1030. 34:51stickers across the board. You can skip
1031. 34:52those on day one if you want. This works
1032. 34:54very well without them, but you can find
1033. 34:56those in settings, and they have core
1034. 34:58plugins and they also have a community
1035. 35:00plugins. You have to turn them on, so
1036. 35:01approve the fact that those can be used,
1037. 35:03and I'm using here Dataview and Kanban.
1038. 35:04And then, you start feeding your second
1039. 35:07brain. You start ingesting data. If you
1040. 35:09do it 10 times, around 10, those dots
1041. 35:11start to become a web. So, take one
1042. 35:14afternoon to set this up, then daily
1043. 35:16just start adding maybe one link a day
1044. 35:18or whatever you think is best. Ideally,
1045. 35:20you also set up some automation, so it
1046. 35:22pulls the data that you care about. You
1047. 35:24can do it for your personal tasks, for
1048. 35:25your business or job or school. You can
1049. 35:28do it for your health or whatever you
1050. 35:30want. Check out the link below, so I'll
1051. 35:33have a PDF that kind of explains it, so
1052. 35:35you can read it or just hand it to
1053. 35:37Claude Code or whatever chatbot you're
1054. 35:39using and tell it to set it up for you.
1055. 35:42So, what we build is a Wikipedia that
1056. 35:44you care about. It's maintained entirely
1057. 35:47by AI. It can run your work life, plus
1058. 35:50create certain actionable playbooks from
1059. 35:52your own data, things that you care
1060. 35:54about. Everything's stored on your
1061. 35:55computer as basically text files. It's
1062. 35:58on your computer, you own it forever.
1063. 36:00You're not tied down to any application,
1064. 36:02any model. As new things come out, this
1065. 36:05stays useful. Obsidian or Claude Code,
1066. 36:08they don't control those files. Those
1067. 36:09files are text files. No one can lock
1068. 36:12them down or take them away. Why this
1069. 36:14matters is because notes that get
1070. 36:17maintained this way, they actually get
1071. 36:19used. They're useful. They also don't
1072. 36:21take a lot of bandwidth for you to to
1073. 36:22figure them out and organize them. This
1074. 36:24was an 80-year-old dream at this point
1075. 36:26that is finally possible. It's your own
1076. 36:28personal library with a librarian that's
1077. 36:31never sleeps. So, make sure you're
1078. 36:32subscribed to this channel because more
1079. 36:34stuff is coming that's going to utilize
1080. 36:36this and build on top of this. If you
1081. 36:39have any questions, comments, tips,
1082. 36:41leave them below. And if anything didn't
1083. 36:42make sense, definitely let me know so I
1084. 36:44can kind of troubleshoot and hopefully
1085. 36:46improve the next time that I'm talking
1086. 36:48about this. If you made this far, thank
1087. 36:49you so much for watching. I will see you
1088. 36:51in the next one.