YouTube2Text

Master 96% of Codex in under 35 minutes — Transcript

by Nate Herk | AI Automation · 8,664 words · 1,175 segments · language en · Watch on YouTube

Full transcript

  1. 0:00Today I'm going over the 18 core Codeex
  2. 0:02concepts that you actually need to know
  3. 0:03in order to start using it right away
  4. 0:05and getting value from it. It doesn't
  5. 0:06matter if you're not technical at all or
  6. 0:07you've never used Codex before. By the
  7. 0:09end of this video, you'll understand
  8. 0:10exactly how this thing actually works so
  9. 0:11that you can start using it right away.
  10. 0:13So, let's not waste any time and just
  11. 0:14get straight into this one. All right,
  12. 0:15so I've split these 18 core concepts
  13. 0:16into four different parts and they get
  14. 0:18cooler as we continue to go on. So, part
  15. 0:20one, we're going to start off here is
  16. 0:22foundations. So, we're going to kick off
  17. 0:23with concept number one, which is
  18. 0:25projects. Now, a lot of you guys, I'm
  19. 0:27assuming, have used chatgbt or claude in
  20. 0:28the web before where you can talk to an
  21. 0:30LLM and it can, you know, give you an
  22. 0:32answer. But the problem you face is that
  23. 0:34every time you're talking to it or you
  24. 0:35start a new chat, you have to kind of
  25. 0:37familiarize it with you and your
  26. 0:38business and what you're working on that
  27. 0:40week or that quarter. So, with projects,
  28. 0:41you can basically keep everything
  29. 0:42organized into one project so that as
  30. 0:45you keep working inside of it, it only
  31. 0:46gets smarter about knowing who you are.
  32. 0:48So, right here on the lefth hand side,
  33. 0:49you can see that I've got a few projects
  34. 0:50going on. I've got Herk 2, I've got
  35. 0:52Hyperframes, I've got Trading Challenge,
  36. 0:54I've got AIS demo, and I typically am
  37. 0:56always just working inside of my Herk 2
  38. 0:57project. My Herk 2 project is basically
  39. 0:59what I call my AI operating system. And
  40. 1:01what a project is is actually very
  41. 1:02simple. It is just a collection of
  42. 1:04folders and files. So right here, this
  43. 1:06is inside of my file explorer on my
  44. 1:07computer and it's called Herk 2. And
  45. 1:09this is exactly the project that I have
  46. 1:11open in Codeex. When I say right here,
  47. 1:13I'm inside of Codex, right? This is the
  48. 1:15file path that I'm working in, which is
  49. 1:16if I open this up, my Herk 2 project. So
  50. 1:19what that means is this project has
  51. 1:21rules about me. It has, you know,
  52. 1:22credentials. It has all my projects. It
  53. 1:25has all my YouTube videos. This thing
  54. 1:26has so much knowledge about me. So now
  55. 1:28when I come into the chat and I say,
  56. 1:30"Hey, can you just give me a rundown of
  57. 1:32what we've been working on this past
  58. 1:34month?" This thing basically gets
  59. 1:35familiarized with my project and then it
  60. 1:37looks at everything that we've been
  61. 1:38working on together. It even looks
  62. 1:40through all of my previous chats that
  63. 1:41I've had with Codeex. So here you can
  64. 1:43see that it looked back from the past
  65. 1:44month and it said, "Here's what we've
  66. 1:46worked on. We've been improving your
  67. 1:47Herk 2 and your AISOS. We've been
  68. 1:50working on YouTube content and strategy,
  69. 1:51AI model testing and demos, apps and
  70. 1:53product experiments, websites, branding,
  71. 1:56business operations. So, this is no
  72. 1:58longer just an AI chatbot. This is now a
  73. 2:00co-founder. This is a personal
  74. 2:02assistant, which brings us on really
  75. 2:03nicely into concept number two, which is
  76. 2:05something called an agents.md.
  77. 2:08And don't let that MD intimidate you
  78. 2:10here. If I go over to the right hand
  79. 2:11side and I open up my files, you'll
  80. 2:13notice if I scroll down, I have
  81. 2:15something right here called an
  82. 2:15agents.mmd. And when I open this up,
  83. 2:18it's literally just a markdown file.
  84. 2:20That's what the MD means. And it's just
  85. 2:22rules. Like this is a very simple thing
  86. 2:23that you all can read and it's not
  87. 2:25technical at all. So this is basically
  88. 2:27like the rules for your project. So you
  89. 2:29remember inside of my Herk 2 project, I
  90. 2:31have this agents onMD. Well, if I was
  91. 2:32inside of my Hyperframes editor project,
  92. 2:34there would be a different agents.m MD.
  93. 2:36So this thing basically just sets the
  94. 2:37ground rules for the project that you're
  95. 2:39working in. So right here you can see
  96. 2:40that this one says, "Hey, you are Nate
  97. 2:41Herk's AI operating system. Your job is
  98. 2:43to help him spend less time on
  99. 2:44operations so he can focus on making
  100. 2:46YouTube videos." And then I give it
  101. 2:47things like core operating rules for how
  102. 2:49to work with me. I give it some stuff
  103. 2:50about how to be safe on my desktop. And
  104. 2:52then the real bulk of my agent MD is
  105. 2:54what I call routing. So this is a
  106. 2:56routing map. It basically means, okay,
  107. 2:58so yeah, Nate's project is huge. There
  108. 3:00are hundreds, probably thousands of
  109. 3:02files and folders inside of this
  110. 3:03project. So let me just explain to you
  111. 3:05how this actually works. If you need
  112. 3:07business stuff, you go to the wiki. If
  113. 3:09you need corporate structure, you go to
  114. 3:10the corporate structure section inside
  115. 3:11the wiki. If you need his voice, you go
  116. 3:13here. If you need course knowledge, you
  117. 3:14go here. If you want active projects,
  118. 3:15you go here. So everything that I'm
  119. 3:17doing, we're routing it back somewhere
  120. 3:19so that my agents.mmd file tells this
  121. 3:22project where everything is so that
  122. 3:24every time I open up a new chat, like I
  123. 3:25said, I don't have to reexplain things.
  124. 3:27And basically the way that this works,
  125. 3:28if we picture this being a chat thread,
  126. 3:30you know, you come over here and you
  127. 3:31say, "Hey, you say, "Hey, Codex, I need
  128. 3:33you to help me do this." And before
  129. 3:35Codex actually even like reads this
  130. 3:37message, what it does is it first reads
  131. 3:39the agents.mmd. So I'm just going to put
  132. 3:41a here and that stands for agents.mmd.
  133. 3:43But codeex will basically read this and
  134. 3:45then it will read your message and then
  135. 3:46it will respond with something like, you
  136. 3:48know, hey, how can I help you today? So
  137. 3:49the kind of stuff that you want to put
  138. 3:50inside your agents.mmd is the type of
  139. 3:52stuff that you want codeex to know about
  140. 3:54you every single time you're talking to
  141. 3:56it in that project. And now moving on to
  142. 3:58concept number three, we have something
  143. 4:00called the agent loop. So you'll hear
  144. 4:02about something like codecs or claude
  145. 4:04code or Hermes agent. These are all
  146. 4:07called agent harnesses. And these
  147. 4:09harnesses have their own agent loops.
  148. 4:11And real quick, if you're wondering
  149. 4:12about codecs versus like work, chatbt
  150. 4:15work, I always use codecs. The majority
  151. 4:17of my work is not building products or
  152. 4:19building software, but I only use codecs
  153. 4:21because it's just the most powerful. And
  154. 4:22once you get through this video, you'll
  155. 4:23realize how easy it is to use. But
  156. 4:25anyways, the reason why this is called a
  157. 4:26harness is because they have this thing
  158. 4:28called the agentic loop, which at the
  159. 4:30highest level basically just means when
  160. 4:31you ask the agent a question, it has a
  161. 4:34bunch of tools at its disposal and it
  162. 4:36basically thinks and it reasons and it
  163. 4:38uses tools and then it responds to you
  164. 4:40and then it thinks and then it reasons
  165. 4:41and it takes action. It uses tools,
  166. 4:43responds to you and that's basically the
  167. 4:44agent loop. And what's really awesome
  168. 4:46about that is you can see it in action.
  169. 4:48So right here you can see that this said
  170. 4:49worked for 25 seconds. And if I open
  171. 4:51this up, we can basically see what it
  172. 4:53did, right? So like it hid this response
  173. 4:55and it gave us our output, but it had a
  174. 4:57reasoning loop inside. It first said,
  175. 4:59"Okay, I'll check our recent tasks and
  176. 5:00project nodes, then I'll pull together
  177. 5:02the main themes from the past month."
  178. 5:03And when it did that, it ran all these
  179. 5:05commands. It ran this, which you don't
  180. 5:07have to know what this means, but
  181. 5:08basically what this did is it looked
  182. 5:10inside something right here called a
  183. 5:11memory.md. So very similar to the
  184. 5:13agents.md, but this is about memories
  185. 5:15rather than just like rules. It then
  186. 5:18looked at other chats. And so we can
  187. 5:19basically watch it do things. So, to
  188. 5:21show you guys a real example of this,
  189. 5:23hey, could you just check my most recent
  190. 5:24YouTube video that I uploaded and give
  191. 5:26me three comments that you thought were
  192. 5:28funny from that video.
  193. 5:31Now, as I shoot this off, what I want
  194. 5:32you guys to pay attention to is what
  195. 5:34we're watching. So, first it said
  196. 5:35thinking. Now, it says, "I'll find your
  197. 5:36latest upload, read through the
  198. 5:38comments, and pick three that made me
  199. 5:39laugh." And now we can see what it's
  200. 5:40doing. So, it's looking through the
  201. 5:42memory. It's looking at these things.
  202. 5:44It's running these commands. And these
  203. 5:45commands are basically right here. It's
  204. 5:47reading, you know, files to get
  205. 5:49established and to get oriented with
  206. 5:51what it needs to do. And so, you can
  207. 5:52actually learn a lot about how codeex
  208. 5:55works and how to best work with it by
  209. 5:57when you ask it a question, you just
  210. 5:58kind of watch what it does. So, you can
  211. 6:00see what it had to do was it had to run
  212. 6:01a command in order to actually go talk
  213. 6:04to YouTube, look at the data, and pull
  214. 6:06it back. And now we can see three
  215. 6:08comments. We have this one, never seen
  216. 6:10anyone so excited to lose money. Trading
  217. 6:12with AI is very easy. I just prompt
  218. 6:13Claude only take profitable trades. And
  219. 6:15if I want to lose money, I can do it
  220. 6:17myself faster and better. And what
  221. 6:18you'll notice is how quick that all
  222. 6:20happened. Now, I do have my Astra on
  223. 6:21fast mode, but I will talk about that
  224. 6:23later because that is a core concept
  225. 6:24that's coming up. All right, and moving
  226. 6:26on to the last concept of part one. It's
  227. 6:29a slashgoal. And this is honestly one of
  228. 6:32my favorite things about AI ever. It's
  229. 6:34the ability to do something called a
  230. 6:36slash goal. And right here, you can see
  231. 6:37that that says set a goal to keep
  232. 6:39pursuing. Now, AI agents are super goal
  233. 6:41oriented, which is super cool. So, let's
  234. 6:43say you have Mr. AI agent right here and
  235. 6:45you basically say, "Hey, here is your
  236. 6:47goal. You know, this is what you need to
  237. 6:49do." And you give it this objective
  238. 6:50goal. What happens is it will basically
  239. 6:52just keep working and keep working and
  240. 6:54keep working until it actually hits that
  241. 6:57goal. And it's cool because even if you
  242. 6:58kind of like give them a roadblock, they
  243. 7:00are going to be so goal-oriented that
  244. 7:03they're going to keep retrying different
  245. 7:04things and they're going to just
  246. 7:06basically keep going until they're able
  247. 7:08to succeed and hit that goal for you.
  248. 7:10Now, obviously, the more objective a
  249. 7:11goal, the easier it is to prove that.
  250. 7:13So, if you say, "Hey, can you, you know,
  251. 7:15work on this until you've pulled in 257
  252. 7:18sources and then write the report."
  253. 7:20That's an objective goal, right? 257
  254. 7:22sources and written report. But you can
  255. 7:24also set goals that are a little bit
  256. 7:25more emotional. Like your goal could
  257. 7:27maybe be more like, you know, until
  258. 7:29you're fully satisfied or until you've,
  259. 7:31you know, verified this over and over
  260. 7:33and you feel good about it. And in that
  261. 7:34case, you're still going to, you know,
  262. 7:37get there and it will still decide when
  263. 7:39it's done. But as you get more objective
  264. 7:41with your goal, the goal prompts are
  265. 7:43just more effective. But let me show you
  266. 7:45one goal that I actually did that you
  267. 7:46guys have already been witnessing in
  268. 7:48this specific example is I went to my
  269. 7:50hyperframes project. And hyperframes
  270. 7:51basically lets your coding agents build
  271. 7:53HTML and then render it as video. So
  272. 7:55essentially it creates video. But what I
  273. 7:57did here is I shot it off a/goal prompt.
  274. 7:59You can see this one says sent as goal.
  275. 8:01So basically what I did is I said I need
  276. 8:03you to help me create some motion
  277. 8:04graphics for a YouTube video. So all the
  278. 8:05motion graphics you've seen so far were
  279. 8:07from this exact prompt right here. I'm
  280. 8:09going over 18 different Codex concepts.
  281. 8:11I want, you know, an intro. I want
  282. 8:12transition cards. I want all this. So, I
  283. 8:14basically defined that I wanted one
  284. 8:15intro scene and that I wanted 18 cards.
  285. 8:18And then you can see that it said I
  286. 8:19finished and I visually checked, meaning
  287. 8:20it actually took screenshots to make
  288. 8:22sure everything looked good before it
  289. 8:23decided to be done. And now we have all
  290. 8:25these animations. But what's cool about
  291. 8:27the goal prompts is that models nowadays
  292. 8:29are getting so good, the AI models are
  293. 8:31getting so good that it's actually
  294. 8:33better to give them a goal and then step
  295. 8:35out of their way. If you really just try
  296. 8:36to like micromanage them, you're kind of
  297. 8:38just like muzzling their capabilities.
  298. 8:40So, if the situation isn't super super
  299. 8:42risky, give them a goal and you can get
  300. 8:43out of the way. Now, there's one quick
  301. 8:45thing that I felt the need to address,
  302. 8:46which is basically the question, "But
  303. 8:48Nate, what if we've been using cloud
  304. 8:49code for a long time?" Well, I'm really
  305. 8:51glad that you asked. It's actually
  306. 8:52really simple. Really, the main
  307. 8:54difference is the agents.mmd, thecodex,
  308. 8:56and the aagents. Because in cloud code,
  309. 8:59you basically have like your claw.md and
  310. 9:01you have yourclaw. So when I first
  311. 9:02started switching over from cla code to
  312. 9:04codeex. All I did was I said, "Hey
  313. 9:05codeex, take a look at this herku
  314. 9:07project, I've been building this up
  315. 9:09using claude code for the past couple
  316. 9:10months. I need you to help me get this
  317. 9:12codeex ready." And basically what that
  318. 9:14means is I made a copy of my claw.md and
  319. 9:16called it agents.mmd.d. So I have two of
  320. 9:18those now. One one agents.mmd one
  321. 9:19claude, but they're basically the same
  322. 9:21file. And then I made a copy of all my
  323. 9:22claude skills and put them in the agents
  324. 9:24folder. And I made a copy of all of my
  325. 9:26like claude settings files and I put
  326. 9:28them in my agents. And the cool thing is
  327. 9:30you don't have to do any of that
  328. 9:31manually. You'd literally just say,
  329. 9:32"Hey, Codeex, look at the documentation,
  330. 9:34analyze my project, and make this codeex
  331. 9:36accessible. Make it codec ready." So,
  332. 9:38anyways, if you were feeling a little
  333. 9:39bit of doubt about all that kind of
  334. 9:41stuff, that's the way I did it. And now
  335. 9:43you have a bunch of local files and
  336. 9:44folders that you can use on any agent
  337. 9:46harness whenever. It's very flexible.
  338. 9:48It's very tool agnostic, and that's what
  339. 9:49you want to be building at the end of
  340. 9:50the day. So, anyways, had to address
  341. 9:52that real quick. Let's get back to the
  342. 9:53video. All right. Awesome. So, let's
  343. 9:54move on to part two of these core
  344. 9:56concepts, which is environments. So the
  345. 9:59first piece of environments is number
  346. 10:00five, which is local versus cloud. So
  347. 10:03there's a couple things to talk about
  348. 10:04here. The first thing is if someone
  349. 10:06says, "Oh, is this running locally or
  350. 10:08are these local files?" All that means
  351. 10:10is does that exist only on the machine
  352. 10:12that you're currently using? Whether
  353. 10:13you're on a MacBook right now or you're
  354. 10:15on a desktop PC like I am right now.
  355. 10:16When you say local, it just means it's
  356. 10:18only accessible by you. So right here,
  357. 10:20if I open up my files, these are
  358. 10:22basically local files. Now, they're
  359. 10:23syncing with one drive when I want them
  360. 10:25to. So that would kind of push them into
  361. 10:26the cloud, but otherwise it's completely
  362. 10:28local. So like if you're working on a
  363. 10:29Word doc on your computer and you forgot
  364. 10:32to save it and then email it to yourself
  365. 10:34on so you could work on it on a
  366. 10:36different laptop, that was a problem
  367. 10:37because it was locally on that other
  368. 10:39machine. But something like Google Drive
  369. 10:40is always in the cloud obviously. But
  370. 10:42this also goes beyond just files. This
  371. 10:44goes to like things that are produced.
  372. 10:46So right here you can see that it
  373. 10:47actually created me this little site for
  374. 10:49all these motion graphics. So if I open
  375. 10:50up this site, you can see that I can
  376. 10:52view all of these videos. I can click
  377. 10:54into the different ones and it threw
  378. 10:55this together for me really nice and it
  379. 10:57served this to me on a local host. So
  380. 10:59right here you can see that this URL is
  381. 11:01127.0.0.1
  382. 11:03and the port of this local host is 8008.
  383. 11:06Now realistically that's not super
  384. 11:08important. You don't really need to
  385. 11:09understand what exactly this means. But
  386. 11:11this is just a local host. But this is a
  387. 11:14little confusing because it looks like
  388. 11:15this is a web address that if I, you
  389. 11:16know, emailed to you, you could open
  390. 11:18this up and have the site. But what
  391. 11:19would happen is this would show on your
  392. 11:21screen or it would show on your computer
  393. 11:23as nothing actually lives here. So like
  394. 11:25if I copied this and I pasted this into
  395. 11:26a new browser but I changed the number
  396. 11:28to 7. I don't think I have anything
  397. 11:31running on port 8087
  398. 11:34on my local machine even though you
  399. 11:35might. So nothing actually loads up. And
  400. 11:37it's quite funny. I remember seeing
  401. 11:38these these tweets where I think it was
  402. 11:40a joke but it was like uh someone was
  403. 11:42pretending they were a beginner and they
  404. 11:44were like wow cloud code and codeex it's
  405. 11:46so cool. Look what I built in one day.
  406. 11:47and then they, you know, attached a
  407. 11:49bunch of local host addresses and
  408. 11:51obviously no one else could open that up
  409. 11:53because they're not on that local
  410. 11:54device. So anyways, that is the
  411. 11:55difference between local and cloud. Now,
  412. 11:57the other piece here is that when you're
  413. 11:58spinning up these new chats, you can
  414. 12:00kind of have these be either local,
  415. 12:02meaning right here, you know, I'm
  416. 12:03working in my local files and folders.
  417. 12:05And that's great because codeex can
  418. 12:07actually like edit these files, right?
  419. 12:09It can delete things, it can make new
  420. 12:10ones, it can edit them, it can move them
  421. 12:12around, it can go to my downloads, it
  422. 12:14can go to my desktop, it can do anything
  423. 12:15on my computer really. or you could
  424. 12:17actually have these work in the cloud.
  425. 12:18And what you'll notice here is when you
  426. 12:19choose cloud, you no longer have the
  427. 12:21ability down here to actually choose the
  428. 12:23model you were using and the permissions
  429. 12:25because this sets up sort of like a
  430. 12:27cloud sandbox environment. Now, I'll be
  431. 12:29honest, I hardly ever do this because
  432. 12:31when I'm working locally, as long as I
  433. 12:33keep my machine on, as long as I keep my
  434. 12:35PC on, it feels pretty cloudlike. And
  435. 12:38here's what I mean by that. Right here,
  436. 12:39you can see that I'm mirroring my phone
  437. 12:40and I'm in the Chat GBT app. And if I go
  438. 12:42over here and I click on this right
  439. 12:44here, which says remote, this lets me
  440. 12:46see all of my actual codeex chats that
  441. 12:50are right here. So if I go to Herk 2 and
  442. 12:52I go to summarize past month, this is
  443. 12:54exactly what we were just looking at
  444. 12:55right here. You can see this is
  445. 12:56literally the same conversation. It has
  446. 12:58those three chats and I could come in
  447. 13:00here and I could say hi. And this gets
  448. 13:02sent in my actual codeex right here on
  449. 13:03my laptop or sorry on my desktop as well
  450. 13:06as my phone. So, as long as my machine's
  451. 13:08on, even though this is running locally
  452. 13:09and I'm out on a walk, I'm out at
  453. 13:11dinner, I could even be in a different
  454. 13:12state and I could control my desktop
  455. 13:15from my phone right here. So, let's say
  456. 13:17for some reason I left a file locally on
  457. 13:19my computer, I could just say, "Hey, can
  458. 13:20you email that file to me?" And then I
  459. 13:22could use it on my laptop. So, anyways,
  460. 13:23just wanted to show off that remote
  461. 13:25capability. And that's going to bring us
  462. 13:26into concept number six, which is work
  463. 13:29trees. Now, honestly, this is something
  464. 13:31where if you're not really building
  465. 13:32software or, you know, production apps,
  466. 13:34this isn't going to matter to you too
  467. 13:36much. Like I don't hardly ever use work
  468. 13:38trees, but it is good to know about
  469. 13:40since you will see these in the
  470. 13:41interface. So right here you can see
  471. 13:42that if I was to go to um the local
  472. 13:45setting, it says that you can work in a
  473. 13:47new local work tree which says create a
  474. 13:49copy of Herk 2 to work in parallel. So
  475. 13:52basically because I'm working in Herk 2,
  476. 13:54if I was going to do something like
  477. 13:55risky or if I wanted to try and
  478. 13:56experiment with something or for
  479. 13:58example, if I was building software and
  480. 13:59I wanted to try to change the way that
  481. 14:01the onboarding flow worked, but I wanted
  482. 14:02to like not interrupt the main branch,
  483. 14:05especially if other people on my team
  484. 14:06are working on that main branch, I could
  485. 14:08use a work tree to basically create a
  486. 14:09copy and I could test in there. And then
  487. 14:12what you can do later is you can merge
  488. 14:13it back into the main. So it's basically
  489. 14:15just a duplicate that lets you test. But
  490. 14:17like I said, there's hardly any times
  491. 14:18for knowledge work that I ever use work
  492. 14:20trees. And if I ever am building
  493. 14:21software and stuff, a lot of times Codex
  494. 14:23will say, "Hey, by the way, you might
  495. 14:24want to use this. You might want to do
  496. 14:25this in a new work tree." So, I wanted
  497. 14:26to call it out because it's important
  498. 14:28because you might see things like this
  499. 14:29master branch and you might see other
  500. 14:30work trees that might have been spun up
  501. 14:32automatically. But, I will say it's not
  502. 14:33something that I'm actively thinking
  503. 14:34about or think that you need to be
  504. 14:36actively thinking about in order to get
  505. 14:37the most out of codeex. So, wanted to
  506. 14:39bring that up. But, let's move on to
  507. 14:40number seven, which iscodeex.
  508. 14:43Now, this is really interesting. We
  509. 14:45talked about how when you're inside of
  510. 14:46your project, you have things like an
  511. 14:47agents.mmd. Now, if I go to my files,
  512. 14:50there's also something in here called a
  513. 14:52codeex. So, if I open this up, what
  514. 14:54you'll see in here is I have some agents
  515. 14:56and I have some workflows, but I also
  516. 14:58have a config. ML file, which is
  517. 15:01basically like some local settings. And
  518. 15:03there are basically going to be two
  519. 15:04different places where a codex file
  520. 15:06lives. And that's either going to be
  521. 15:08inside of your project. So like in this
  522. 15:10example, this is my codeex inside of my
  523. 15:12Herk 2. But there's also going to be one
  524. 15:14which is user or global, which means
  525. 15:16this lives under your actual user. And
  526. 15:18these are settings that hold things like
  527. 15:21memories and sessions and automations.
  528. 15:23These are things that will apply to your
  529. 15:25global codeex. So whenever you're in
  530. 15:27codeex, regardless of what your project
  531. 15:28you're actually in, the docex will have
  532. 15:31personal settings and app data and stuff
  533. 15:32like that. Now once again, I think this
  534. 15:34is something it's important to
  535. 15:35understand where that lives and what it
  536. 15:37does. But this isn't a folder that I'm
  537. 15:38actively thinking about and actively
  538. 15:40having to like maintain. But I did want
  539. 15:42to call that out because sometimes if
  540. 15:44you need to have a certain setting
  541. 15:45applied or something like that, that is
  542. 15:47where it lives. Now, the cool thing
  543. 15:48about this is if you're ever confused,
  544. 15:50you say, "Hey Codeex, what did you do
  545. 15:52there?" Or, "Do we need to add this to
  546. 15:53mycodex or do we need to move this to a
  547. 15:55different folder?" You can just ask it
  548. 15:57questions because it can search through
  549. 15:58its own documentation and understand how
  550. 16:00to do things like correctly. But I do
  551. 16:02think that understanding the difference
  552. 16:03between project scope things and user
  553. 16:05scope things that's definitely important
  554. 16:07especially as we get into the agents
  555. 16:09which is going to come up later. And
  556. 16:10that is a great segue into part three of
  557. 16:12the core concepts which is control and
  558. 16:14customization. So number eight here is
  559. 16:16about different AI models. So when you
  560. 16:20are starting a new chat you can choose
  561. 16:22the model you want to use. You can also
  562. 16:24change the model in between messages.
  563. 16:26So, I could shoot off a message on Astra
  564. 16:27and then I could, you know, change this
  565. 16:29to something else and then I could shoot
  566. 16:30off another message and I can keep
  567. 16:31changing if I want. Now, if you're
  568. 16:33watching this video way later, maybe the
  569. 16:34model names look different, but
  570. 16:35essentially the theory is the same. You
  571. 16:37have different models to choose from and
  572. 16:39they all have different prices and
  573. 16:41strengths. So, like right now, GBD6
  574. 16:42Astra is just the absolute strongest,
  575. 16:44but it's going to cost the most. It's
  576. 16:46going to eat the most of your weekly
  577. 16:47limit. Right here, you can see that I'm
  578. 16:48at 100% because it just got reset. But
  579. 16:51as you chat with Astra, it will drain
  580. 16:52your weekly usage faster than you know
  581. 16:555.6 Soul would, which would drain your
  582. 16:57usage limit faster than 5.6 Tarot would.
  583. 17:00Now, right here, I just want you to look
  584. 17:01at the API pricing so I can explain to
  585. 17:03you how these models are actually
  586. 17:05billing you. So, they build by tokens,
  587. 17:07and they usually build by 1 million
  588. 17:09tokens. Meaning, on Astra, for every 1
  589. 17:11million input tokens, it will cost you
  590. 17:14$10. And for every 1 million output
  591. 17:16tokens, it will cost you 50 bucks. And
  592. 17:17here you can see the different pricing
  593. 17:19for these different models. And you can
  594. 17:20see how it gets cheaper as you go down.
  595. 17:22Now, this is API pricing, which means if
  596. 17:24you were building an automation that was
  597. 17:26programmatically using these models,
  598. 17:28that's how much it would cost you. But
  599. 17:30this usage inside of Codex, that's not
  600. 17:31programmatic. This is usage limit or
  601. 17:34subscription limit. So, when you pay 20
  602. 17:35or 100 or 200 bucks a month for Codeex,
  603. 17:38this is what you're actually your usage
  604. 17:39limit is eating up is that subscription.
  605. 17:42And you're getting way more inference.
  606. 17:43Like, this is a $200 a month
  607. 17:44subscription. And if I use all of my
  608. 17:47weekly usage every single week, I'm
  609. 17:50getting about $14,000 worth of inference
  610. 17:53every month and I'm only paying 200
  611. 17:54bucks for that. So the subscriptions are
  612. 17:56much cheaper than if you were using the
  613. 17:57API pricing. But if you ever do go over
  614. 17:59your weekly limit, you'd have to buy
  615. 18:01extra codeex credits. I don't exactly
  616. 18:03know why they price it like this, but
  617. 18:04you know, you can see they just give you
  618. 18:06codeex credits instead. So let's say I
  619. 18:08went over my weekly limit. I could go
  620. 18:10here and I could click on buy credits.
  621. 18:12This would bring up this screen which
  622. 18:13lets you buy credits by like the
  623. 18:15thousands or something like that. And
  624. 18:16that is what I wanted to show you guys.
  625. 18:18So like 2500 credits, 5,000 credits,
  626. 18:2125,000 credits. And this is how it would
  627. 18:23charge you based on those tokens. Now,
  628. 18:25if you've never heard of tokens before,
  629. 18:26I know that might be a little confusing.
  630. 18:28Think about it as roughly four
  631. 18:29characters or roughly 3/4s of a word.
  632. 18:32And there's an OpenAI tokenizer right
  633. 18:33here where if I paste in some text, it
  634. 18:35will show you how many tokens this would
  635. 18:37be. So this paragraph for example was
  636. 18:40about 240 characters or sorry 274
  637. 18:42characters and about 56 tokens. And it's
  638. 18:44interesting because sometimes
  639. 18:45punctuation is a token, right? Like this
  640. 18:47period is a token, this comma is a
  641. 18:49token, this word's a token, um this
  642. 18:51word's a token. And this is basically
  643. 18:53just kind of breaking down what it might
  644. 18:54look like for these tokens. But anyways,
  645. 18:56you can see this was 56. And what else
  646. 18:58you might notice is that I earlier said
  647. 19:00input tokens or output tokens. Input
  648. 19:02tokens are whatever you're feeding in.
  649. 19:04So your prompts or if it's reading like
  650. 19:05a PDF or a word doc those are input
  651. 19:08because they're going into the model and
  652. 19:09output tokens are anything that it
  653. 19:11outputs. So even its reasoning here is
  654. 19:13output tokens. This entire thread is
  655. 19:15output tokens and that's why the output
  656. 19:17tokens are more expensive. Okay. So
  657. 19:19let's move on to number nine which is
  658. 19:20about effort. So within each model not
  659. 19:23only do you choose the model but you can
  660. 19:24also choose the effort level. So right
  661. 19:26here I can move Astra down to medium or
  662. 19:28I can move it down to low or I can move
  663. 19:30it up to extra high or if I go like
  664. 19:32ultra I have to allow full access
  665. 19:34because that can get you know pretty
  666. 19:36autonomous. But anyways the point being
  667. 19:38as you move up on these effort levels
  668. 19:40you get more quality like you get more
  669. 19:43intelligence but you also get more cost.
  670. 19:46It's more expensive to run it at higher
  671. 19:47effort levels. And that's why when you
  672. 19:49see these benchmarks, when there are new
  673. 19:50models released, you'll see the model,
  674. 19:52like for example, Cloud Fable 5, you can
  675. 19:54see this mark was on low and this one
  676. 19:57was on medium and this one was on high.
  677. 19:59So, it shows how one model can behave
  678. 20:01differently and also cost a different
  679. 20:02amount based on the effort level that it
  680. 20:04was being used and tested at. Now, it
  681. 20:06actually is pretty nice to play with
  682. 20:08these because sometimes Astra High is
  683. 20:10just so much more powerful than what you
  684. 20:11need to do. Like if you were using GBD6
  685. 20:13Astra High to help you write an email,
  686. 20:15that's overkill and you're paying way
  687. 20:17more than you should realistically to
  688. 20:18write an email. I could probably come
  689. 20:19down here and use 56 Terra and be just
  690. 20:21fine with that and it's a lot cheaper.
  691. 20:23So definitely try to play with the model
  692. 20:24and play with the effort based on your
  693. 20:26task. But a lot of times if you're just
  694. 20:27doing a lot of knowledge work, I would
  695. 20:28just chuck it on sole and I'd maybe just
  696. 20:30go to medium and just call it a day
  697. 20:32there. Now the other thing about effort
  698. 20:33is you can do something called fast
  699. 20:34mode. This obviously makes the model
  700. 20:36respond quicker and do things quicker.
  701. 20:37And it it's 1.5 speed, but it costs you
  702. 20:39more usage, so it will eat your weekly
  703. 20:41subscription faster as well. All right,
  704. 20:43and moving on to number 10, we have
  705. 20:45permissions. Basically, right down here,
  706. 20:47you can see that this is orange and it
  707. 20:48says full access. This is basically
  708. 20:50Codeex's yolo mode, which means that
  709. 20:52it's not going to stop and ask you a
  710. 20:54bunch of permissions and questions. Can
  711. 20:55I do this? Can I do this? If you go on
  712. 20:57ask for approval, it's going to always
  713. 20:59ask to edit external files and always
  714. 21:01ask to use the internet. So that's if
  715. 21:02you want to sit here and really sort of
  716. 21:04like manage it and you're wanting to
  717. 21:05make sure everything's safe, you know,
  718. 21:07maybe start off like this so you can
  719. 21:08just understand, oh, okay, it's asking
  720. 21:09me if it can run this command or it's
  721. 21:11asking me if it can open up my browser
  722. 21:12and then you just get familiar with what
  723. 21:14it looks like for Codeex to actually run
  724. 21:15that agentic loop. You could also go on
  725. 21:17approveof for me. So it's only going to
  726. 21:19ask you if actions are detected as
  727. 21:21unsafe like maybe deletes or maybe
  728. 21:22certain sorts of like API calls or
  729. 21:25things like that. So that's how these
  730. 21:26permissions work. And if you really want
  731. 21:27to get granular about, oh, I want to
  732. 21:29like make sure these certain actions
  733. 21:30never happen, you can ask Codex about
  734. 21:32that and it can work things into those
  735. 21:34config files like we talked about
  736. 21:35earlier with the codeex that it can
  737. 21:37certainly like block things out. But now
  738. 21:38let's move on to number 11, which I
  739. 21:40think is my favorite one, which is
  740. 21:42skills. Skills are so so important.
  741. 21:44They're basically reusable workflows.
  742. 21:46They're recipes that let you do
  743. 21:48something once and then teach Codex how
  744. 21:50to do it the same way every single time.
  745. 21:52So now you can just run all these
  746. 21:53skills. So like I said, a skill is
  747. 21:55basically just a recipe. So, let's say
  748. 21:57you come in here and you make, you know,
  749. 21:59a chocolate chip pancake and you
  750. 22:00followed a recipe to make that pancake.
  751. 22:02You would basically say, "Okay, if I
  752. 22:04ever want to make chocolate chip
  753. 22:04pancakes again, I'm going to use this
  754. 22:06recipe because I know that I'm going to
  755. 22:07get a good result." But let's say that
  756. 22:09you burnt these pancakes and you said,
  757. 22:11"Okay, well, what I need to do is make a
  758. 22:12quick update in this recipe to say, hey,
  759. 22:14you know, cook it up 30 seconds less on
  760. 22:16each side." And the next time you run
  761. 22:17the skill, you basically see, okay, are
  762. 22:19these better? Are these good? Do I need
  763. 22:20to make any more feedback in here? Or
  764. 22:22now, is this skill just the way I like
  765. 22:24it? And then your agents are able to use
  766. 22:26these skills. So here's an example where
  767. 22:28I basically gave it a YouTube video. So
  768. 22:29you can see that this is literally just
  769. 22:30a local file path of a YouTube video.
  770. 22:32And I said that I needed a video
  771. 22:34description, timestamps, and a LinkedIn
  772. 22:35post. It went ahead and it ran its
  773. 22:37agentic loop and it gave me a
  774. 22:39description, timestamps, and a LinkedIn
  775. 22:40post. And then I said, "Okay, now I need
  776. 22:42you to turn this into an X article." And
  777. 22:43put the draft into my X as a draft. And
  778. 22:45it knew exactly how to do this with a
  779. 22:47thumbnail, with a title. It also put
  780. 22:49screenshots from my actual video and it
  781. 22:50sprinkled them throughout my whole X
  782. 22:52article. And this is because it used a
  783. 22:54bunch of skills to be able to know how
  784. 22:56to do this. So right here, if I asked
  785. 22:57it, what skills did you use? It used my
  786. 22:59YouTube description skill, my timestamp
  787. 23:01skill, my LinkedIn post skill, my ex
  788. 23:02article skill, my ex article from video
  789. 23:04skill, my format X article skill. It
  790. 23:06used all of these skills because I've
  791. 23:07taught it how to do something good once.
  792. 23:09And then every time I run these skills,
  793. 23:11I just give feedback. Hey, that was
  794. 23:12good. You know, I liked this, but I
  795. 23:14didn't like this. Update the skill. And
  796. 23:15now what's cool is you can just invoke
  797. 23:17these either with natural language. It
  798. 23:18will be smart enough to do it with
  799. 23:19natural language, or you can do it as a
  800. 23:21slash command. So, I have one in here
  801. 23:22that I can call like the audit. And you
  802. 23:24can see this will run my audit or I have
  803. 23:26one in here called the 3D brain. And
  804. 23:28this runs the 3D brain. Man, I think I
  805. 23:30planned this out so good. This is going
  806. 23:31to bring us on to number 12, which is
  807. 23:34agents. So, we talked about codecs and
  808. 23:36agents is basically the exact same. It
  809. 23:39just holds different things. So, right
  810. 23:40here, you can see that the agents holds
  811. 23:42your skills and these once again can
  812. 23:44either be project level skills or they
  813. 23:45can be user scope skills. So, they can
  814. 23:47be global skills. So if I come into this
  815. 23:50chat right here which is inside of my um
  816. 23:53you know Herk 2 project and I go to open
  817. 23:55up the files and I go to my agents
  818. 23:58folder which is right here we have a
  819. 23:59folder called skills and this is a bunch
  820. 24:01of skills as you can see. Now if I open
  821. 24:03up an actual skill so for example let me
  822. 24:05go to my YouTube description skill. This
  823. 24:08is an actual skill.md. So once again MD
  824. 24:10just means markdown file. So what we
  825. 24:12have here is some metadata and this is
  826. 24:14what codeex will read to understand okay
  827. 24:16do I need to invoke the skill or not.
  828. 24:17So, it has the name and it has basically
  829. 24:18when to use the skill. And then all the
  830. 24:20skill is is just instructions. It's,
  831. 24:22hey, for Nate's YouTube descriptions,
  832. 24:25keep them short, two to four sentences,
  833. 24:26paragraph form, no bullet points, no m
  834. 24:28dashes. Here's the process. You read the
  835. 24:30video outline. You identify the core
  836. 24:32topic. You write a short paragraph. And
  837. 24:33that's the skill. It's a very, very
  838. 24:35simple prompt. But if I go to my ex
  839. 24:36article one, they can also be way more
  840. 24:38in-depth. So, this one is way longer.
  841. 24:40There's a lot of other rules and a lot
  842. 24:42of other criteria. There's way more
  843. 24:43steps. There's way more things to do.
  844. 24:45And also what's cool is skills can
  845. 24:47reference other things. Skills can
  846. 24:48reference other skills. They can
  847. 24:50reference other agents. They can
  848. 24:51reference Python files or context files.
  849. 24:54So in this example, this skill
  850. 24:55references my voice guidelines like my
  851. 24:57LinkedIn style guide as well as my X
  852. 24:58article style guide. But that is where
  853. 25:00your skills live inside of the agents
  854. 25:02folder. And then moving on to number 13.
  855. 25:04Here we have plugins. And this is where
  856. 25:06a lot of the magic actually happens
  857. 25:08because you can super easily come over
  858. 25:10here in the Codex desktop app. You can
  859. 25:12go to plugins and you can connect a
  860. 25:13bunch of things. You can see all the
  861. 25:14stuff that I have installed like Alpaca,
  862. 25:16Gmail, ClickUp, Clay, Google Drive,
  863. 25:18GitHub, Canva, and the list goes on and
  864. 25:21on. You can also search for different
  865. 25:23plugins. So, let's say you were really
  866. 25:24interested in connecting something like,
  867. 25:26I don't know, let's see, do they have a
  868. 25:27LinkedIn plugin? They have a LinkedIn
  869. 25:29plugin where we can find the right
  870. 25:30profession. We can grow our business
  871. 25:31with ads. We have all these other sorts
  872. 25:33of plugins, too. So, if you want to try
  873. 25:35to connect your different apps, just
  874. 25:36come in here and see if you can connect
  875. 25:38them using these plugins. Maybe they
  876. 25:39have Composio, they don't have
  877. 25:41Composeio, and if they don't have a
  878. 25:42plugin here, you can still connect it
  879. 25:44using AENV, using an API key. But
  880. 25:46plugins are so much easier because you
  881. 25:48can just basically sign in once. They've
  882. 25:50got ones for creativity, for developer
  883. 25:51tools, for business and operations, for
  884. 25:53data analytics, communication, and
  885. 25:55they're basically always growing this
  886. 25:57plug-in library. All right, so that now
  887. 25:59brings us on to part four, which is
  888. 26:01tools and scale. Okay, we've got five
  889. 26:03left here. Number 14, we have the
  890. 26:05browser. Now, this is definitely one of
  891. 26:07my favorite things about Codeex. If I
  892. 26:10open up the tab over here, we can open
  893. 26:11up a browser, which is really cool
  894. 26:13because I can basically just control
  895. 26:14this. So, like I could go to school.com.
  896. 26:16And what you'll notice is when I log
  897. 26:18into school.com over here, I'm just
  898. 26:20going to zoom out a little bit. You can
  899. 26:21see right here that I'm already logged
  900. 26:22into my school account. So, I could
  901. 26:23manage my different communities, which
  902. 26:25also means that Codeex can help me
  903. 26:27manage my communities from here. Because
  904. 26:28when you save these login inside of the
  905. 26:30Codeex browser, it then saves them. So,
  906. 26:32if you say, "Hey, could you go to that
  907. 26:33account and pull the report for me?" or
  908. 26:35hey could you go into school and could
  909. 26:36you make this post for me it can do it
  910. 26:38through the browser if there's not a
  911. 26:40plugin or if there's not like an API or
  912. 26:42MCP server and the browser use of codeex
  913. 26:44and specifically with GBT6 Astra is the
  914. 26:47best that I've ever used ever I'll play
  915. 26:49a clip right here where I asked it to
  916. 26:51basically open up Canva and paint me and
  917. 26:53draw me using the tools in there and I
  918. 26:55gave it just a picture of me and it was
  919. 26:56able to replicate it in a way that I
  920. 26:58thought was really really good. So silly
  921. 27:00example, but it definitely demonstrates
  922. 27:01the vision and the browser use
  923. 27:02capability of this agent. All right, so
  924. 27:05number 15 we have sites. So if you go
  925. 27:07over here to explore and you click on
  926. 27:08sites, this is very very cool because it
  927. 27:11basically lets you put things out there
  928. 27:13onto the cloud. Remember earlier we
  929. 27:15talked about local versus cloud. Codex
  930. 27:17can build something for you and it can
  931. 27:18be a local host but then you say, "Hey,
  932. 27:19can you just put that on a site real
  933. 27:20quick so that you could have your team
  934. 27:22log in or so that you could share it
  935. 27:23with, you know, other people and make it
  936. 27:25publicly accessible." Like right here,
  937. 27:26you can see this London and Paris one is
  938. 27:28everyone whoever has this URL could go
  939. 27:30ahead and open this up. Whereas this one
  940. 27:32is just me. So it would literally have
  941. 27:34to be logged into my account in order to
  942. 27:36be able to open up this site. And what
  943. 27:38else is amazing about that is you can
  944. 27:39open up the analytics for the site. So
  945. 27:41you can basically host things. You can
  946. 27:43connect your own domain and this is
  947. 27:45basically going to replace something
  948. 27:46like a Verscell or like a I don't know a
  949. 27:49Squarespace, wherever you wanted to host
  950. 27:50your sites normally. You can now just do
  951. 27:51it right here from the Codex desktop app
  952. 27:53which is pretty cool. You can also even
  953. 27:54connect a database. So on the back end
  954. 27:56if you had to store like user login data
  955. 27:58or you know permissions or settings or
  956. 28:01you know a database of customer records
  957. 28:03whatever it is you could actually store
  958. 28:05that here too and it's really cool. All
  959. 28:06you need to do to actually use a site is
  960. 28:08you just basically have to do either say
  961. 28:10hey can you turn that into a site or you
  962. 28:11could do slash site and then it will
  963. 28:13basically know that you can you know
  964. 28:14host a site or build a site. So great
  965. 28:16touch here from codeex. All right so
  966. 28:18number 16 we have sub aents. So, sub
  967. 28:21agents are really cool because it lets
  968. 28:23you delegate work out to a bunch of
  969. 28:25different CEX agents. So, right here in
  970. 28:26this main chat, you can see that I'm
  971. 28:28talking to GBT6 Astra, which is a
  972. 28:30really, really smart and intelligent,
  973. 28:32but also expensive model. So, let's say
  974. 28:34I wanted to do a bunch of research, but
  975. 28:36I didn't want to waste Astra's brain on
  976. 28:37that. Or let's say I wanted to do like a
  977. 28:39bunch of different testing and just do a
  978. 28:40lot of stuff at once. Astra here that
  979. 28:42I'm talking to can delegate workout to
  980. 28:44tons of sub aents. Hey, I need you to
  981. 28:46search through X and search through
  982. 28:48YouTube and also dig through my own
  983. 28:49comments. So, just find, you know, what
  984. 28:50are people talking about right now in
  985. 28:52the AI space? Is there any big news or
  986. 28:54drama? I don't want you to do this
  987. 28:55research though. I want you to basically
  988. 28:57just delegate a bunch of sub agents out
  989. 28:58to do the work. And I want all of these
  990. 29:00sub aents to be using 5.6 Terra as the
  991. 29:03model. So, shoot off those agents and
  992. 29:05let me know what they find. So, this is
  993. 29:06really cool because I can specify that I
  994. 29:08want those sub aents, those little like
  995. 29:10researchers and workers to use different
  996. 29:13models. So right here it's using its
  997. 29:14agentic loop and it's going to spin up
  998. 29:16these agents. And whenever you see these
  999. 29:18little colorful things, these are
  1000. 29:20different agents that it's spun up. So
  1001. 29:21right here, this one's called X. And
  1002. 29:23when I open up X, it is actually this
  1003. 29:25sub agent working right here. I can
  1004. 29:27watch this sub agent go through its
  1005. 29:29agentic loop and I can watch it like
  1006. 29:31pull in data and everything like that.
  1007. 29:32So this is the same thing as over here,
  1008. 29:34but this one is just a different model
  1009. 29:36and it's not the main session that
  1010. 29:37you're currently talking to. You can see
  1011. 29:38we have two others that started working.
  1012. 29:40We have this one called YouTube Pulse.
  1013. 29:41And once again, it's going through the
  1014. 29:42same exact aentic loop. We have this one
  1015. 29:44called X drama. We have this one called
  1016. 29:46audience comments. And we can basically
  1017. 29:48watch all of these work, but we don't
  1018. 29:50have to manage them because once these
  1019. 29:52agents are done, the main session here
  1020. 29:54will say, "Okay, cool. All five are
  1021. 29:55done. They reported back to me. Here's
  1022. 29:57what they found." And this is really
  1023. 29:58good for gathering different
  1024. 29:59perspectives or researching out a bunch
  1025. 30:02of things in parallel. And these
  1026. 30:03intelligent models are really, really
  1027. 30:05good at delegation. So sub agents are
  1028. 30:07definitely something that you need to be
  1029. 30:08using. So, while these sub aents are
  1030. 30:10fanned out and they're working, let's
  1031. 30:11move on to number 17, which is scheduled
  1032. 30:14tasks. Scheduled tasks are awesome
  1033. 30:16because this basically lets you have
  1034. 30:18your codecs working autonomously for
  1035. 30:19you. So, on the lefth hand side, if you
  1036. 30:21click on scheduled, you can basically
  1037. 30:22ask codecs to schedule things for you,
  1038. 30:25reminders, updates, or run full sessions
  1039. 30:28because all a task basically is is it
  1040. 30:30injects a prompt into a codec session
  1041. 30:32like this. So it uses your same local
  1042. 30:34files, it uses your same skills, it uses
  1043. 30:36your model, it uses everything, the
  1044. 30:38agentic loop, it just does it
  1045. 30:39automatically for you. So in here, you
  1046. 30:41see that I could create one and I could
  1047. 30:43set it up manually. I could have it run
  1048. 30:45locally or I have could have it run on
  1049. 30:46the cloud. I could also choose what
  1050. 30:48project to work in. So Herk 2 or, you
  1051. 30:50know, Hyperframes editor. I can also
  1052. 30:52choose what chat. So every one of these
  1053. 30:54routines could be a new chat or it could
  1054. 30:56be a new chat for this task or it could
  1055. 30:58be even in an existing thread that we
  1056. 31:00already have going. And then you can
  1057. 31:01basically choose when to schedule it. So
  1058. 31:03it could be daily, it could be hourly,
  1059. 31:05it could be, you know, weekly, or you
  1060. 31:07could get it really, really custom in
  1061. 31:08here. And the cool thing is you can
  1062. 31:09describe everything in here manually, or
  1063. 31:11you could basically just ask Codex to
  1064. 31:14make one for you. So, for example, I
  1065. 31:15have these like seven or eight different
  1066. 31:18ones that are running for um trading.
  1067. 31:20They're literally trading $10,000 of my
  1068. 31:22real money. And I have this set up as
  1069. 31:24scheduled task so that they can
  1070. 31:25basically all work together, talk to
  1071. 31:26each other, check the market, make
  1072. 31:28trades, things like that. And of course,
  1073. 31:29you guessed it, these run in a different
  1074. 31:31project right here called trading
  1075. 31:33challenge. And they all run inside of
  1076. 31:35this challenge thread. You can tell that
  1077. 31:36this has scheduled tasks in it because
  1078. 31:38there's a little clock icon next to that
  1079. 31:40thread. And moving on to the very last
  1080. 31:42concept we have for today, number 18,
  1081. 31:44this is voice mode. Now, everything that
  1082. 31:46you've learned so far, skills, sub
  1083. 31:49aents, projects, agents, all of this can
  1084. 31:51be controlled by using your voice to
  1085. 31:53have a real conversation with codecs.
  1086. 31:55With your voice, you can manage
  1087. 31:57different codeex threads. You can
  1088. 31:58delegate different agents, you can run
  1089. 32:00skills, you can set up schedule tasks,
  1090. 32:01you can build sites, you can do anything
  1091. 32:03here with your voice. And all you have
  1092. 32:04to do is come down to the bottom left
  1093. 32:06and click on start new voice chat. So,
  1094. 32:08forgive me here as I'm awkwardly like
  1095. 32:09hunching down because my headphones
  1096. 32:11don't reach far enough. But watch this.
  1097. 32:13Hey, Codex. So, I need you to help me
  1098. 32:15out real quick inside of my Herk 2
  1099. 32:16project. Can you start a new thread to
  1100. 32:18make a thumbnail? I'm making this
  1101. 32:20YouTube video about 18 CEX concepts. So,
  1102. 32:22if you could make me a thumbnail for my
  1103. 32:24YouTube channel where you see me sitting
  1104. 32:26at a desk working on a laptop and I'm
  1105. 32:28pointing to the right and there's like
  1106. 32:3018 different little AI agents. Um, yeah,
  1107. 32:33just start start with that. And then
  1108. 32:34once you have finished that thumbnail,
  1109. 32:37could you send it to the thread that is
  1110. 32:39currently working inside of my HERK 2 as
  1111. 32:41well and it is doing research on YouTube
  1112. 32:43and X and um looking at my channel and
  1113. 32:46stuff like that. Could you just send
  1114. 32:47that thumbnail to that thread?
  1115. 32:48>> All right, let me check on that. Okay,
  1116. 32:50I've got your thumbnail brief queued up
  1117. 32:52with you with the laptop pointing toward
  1118. 32:53the 18 ages using your existing photos
  1119. 32:56for reference. And now found your
  1120. 32:58research thread. So I'll send the
  1121. 32:59finished thumbnail there. Now, do you
  1122. 33:00guys see how this is all coming
  1123. 33:01together? It started a work tree chat.
  1124. 33:04It was doing the agentic loop. It also
  1125. 33:06created a goal down here. It literally
  1126. 33:07created its own goal because it knew
  1127. 33:10that we had this, you know, task for it
  1128. 33:12and it started working. And you know,
  1129. 33:13the voice chat's still going. So, I
  1130. 33:15don't know what it's going to think of
  1131. 33:16this, but I just wanted to call that out
  1132. 33:17because all of what we just talked about
  1133. 33:19is all coming together because even you
  1134. 33:20can see over here, it started making
  1135. 33:22that actual prompt. Like, it shot off
  1136. 33:25this prompt right here and it started
  1137. 33:26off this new thread. And now, this is
  1138. 33:29working in a work tree and it's going to
  1139. 33:30create that thumbnail for us and then
  1140. 33:32it's going to send it over into this
  1141. 33:34actual chat and we'll see that actually
  1142. 33:35happen. But as you keep navigating, like
  1143. 33:37you can go to the web and you can go to,
  1144. 33:40you know, other things and you can keep
  1145. 33:41this voice working. So as you're just
  1146. 33:43working on your laptop, it'll show you
  1147. 33:45what's going on and you can keep talking
  1148. 33:46to it. And you can also do all this on
  1149. 33:48your phone as well because of the whole
  1150. 33:49remote thing, you can use voice mode
  1151. 33:51when you're on your phone, too. Okay, so
  1152. 33:53that finished up and you can see what
  1153. 33:54happens is it actually can send tasks or
  1154. 33:58messages to each other, right? So, what
  1155. 34:00happened was this was the original
  1156. 34:01thread that it created to make the image
  1157. 34:03and it made this and then it sent it to
  1158. 34:05this thread and said, "Hey, Nate wanted
  1159. 34:06me to send this to you. Here it is." So,
  1160. 34:08with just your voice, I know that was a
  1161. 34:10very simple example, but you're able to
  1162. 34:12coordinate a bunch of different codec
  1163. 34:14threads, manage them, multitask while
  1164. 34:16you're on the go, while you're flipping
  1165. 34:17between different apps. Very cool. But
  1166. 34:19anyways, guys, that is going to do it. I
  1167. 34:20hope you enjoyed. I hope you learned
  1168. 34:21something new. But if you want to keep
  1169. 34:23diving into more stuff with Codex, and
  1170. 34:24I've made so many videos, I'll tag a
  1171. 34:26playlist right up here that has just a
  1172. 34:28bunch of Codeex content. So, hope to see
  1173. 34:29you guys over there. Thanks for sticking
  1174. 34:30to the end, and I'll see you in the next
  1175. 34:32one. Thanks everyone.

About this transcript

This page contains the full transcript of Master 96% of Codex in under 35 minutes by Nate Herk | AI Automation, generated from the public captions YouTube serves with the video. The transcript has 8,664 words across 1,175 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.