Claude Code & MCPs built my $145K marketing machine — Transcript

Full transcript

1. 0:00How can you use AI agents, MCPs, and a
2. 0:03bunch of different tools to make money
3. 0:05on the internet? Today, we walk [music]
4. 0:07through it all. Yes, you can vibe code
5. 0:10anything and right now, and that's
6. 0:12great, but how can you actually use AI
7. 0:14agents to get you customers 24/7? Well,
8. 0:18today we live build it. We actually spun
9. 0:20up 10 Claude code instances, and we show
10. 0:22you how you can do it to help you get
11. 0:25customers on repeat. I loved this
12. 0:27episode. It's my friend Cody Schneider.
13. 0:29He's an absolute legend when it comes to
14. 0:31vibe marketing and growth marketing.
15. 0:33This episode is saucy. And by the end of
16. 0:36the episode, you're going to feel pretty
17. 0:38confident you know what to do. You are
18. 0:40in for a treat. Enjoy the episode, and I
19. 0:42can't wait to see you in there.
20. 0:46>> [music]
21. 0:51>> Cody, by the end of this episode, what
22. 0:54are we going to learn? You're going to
23. 0:55learn how to build your first agents
24. 0:59that allow for you to go and build
25. 1:01personal software to do marketing,
26. 1:03sales, growth, customer experience for
27. 1:05yourself. And by the end of this, you're
28. 1:08going to come out of it with this whole
29. 1:10new tool set that allows for you to do
30. 1:13all of the middle work without touching
31. 1:15a keyboard. You're just going to use
32. 1:16your voice and have agents do work for
33. 1:19you in the background, man. It's going
34. 1:20to be crazy.
35. 1:22Okay, and can you list off a few of the
36. 1:24piece tools and pieces of software we're
37. 1:26going to use? Like
38. 1:27Yeah, absolutely. We're going to touch
39. 1:29Phantom Buster. We're going to use
40. 1:31Instantly AI. We're going to use
41. 1:33Verifonic. We're going to use
42. 1:36railway.com.
43. 1:38We're also going to use
44. 1:40a bunch of different other tooling
45. 1:42that's in my go-to-market stack. So,
46. 1:45we're also just going to use
47. 1:46like the Facebook Ads API as an example
48. 1:49as another
49. 1:50just like, you know, way that we're
50. 1:52going to interact via this this agent
51. 1:54harness Claude code. All right, and
52. 1:57we're going to live build it and
53. 1:58everyone, well, you're going to watch
54. 1:59you're going to watch the whole thing.
55. 2:00So, let's get into it.
56. 2:02Cool. Sweet, man. So, just to begin
57. 2:04with, do you do you know like GTM
58. 2:06engineering or like what it even means
59. 2:08or like where it comes from?
60. 2:10No, honestly, I don't. That's what I'm
61. 2:13here
62. 2:13>> like a buzz It's just a buzzword, right?
63. 2:15So, this is is actually like made up by
64. 2:16clay.com, which is hilarious.
65. 2:19And they originally did it as like a way
66. 2:22to explain somebody that like does
67. 2:24basically like cascading workflows for
68. 2:27like data enrichment to do outbound
69. 2:29sales motions over email or Slack or
70. 2:33you know, it could be like cold calling.
71. 2:35So, that was kind of the origin of this
72. 2:37was like it was just basically this term
73. 2:39that was given to it, but
74. 2:41it's quickly evolving into something
75. 2:44entirely different.
76. 2:46And so, let me screen share and I can
77. 2:47just like show you like what we're
78. 2:49seeing this work as now. But basically
79. 2:53like what we're like seeing is that the
80. 2:56Can you see this all right?
81. 2:58Yeah. Cool. So, how I'm thinking about
82. 3:00it now is like basically everything that
83. 3:02used to be the middle work that we would
84. 3:04do like
85. 3:06all of anything that I would do to touch
86. 3:08the keyboard, I'm now passing it on to
87. 3:11some type of agent harness, whether it's
88. 3:13Claude code or it's Codex or
89. 3:16any of these tools. And so, my job
90. 3:18suddenly turns into like I have ideas, I
91. 3:20pass them on to Claude code, and then
92. 3:23I'm basically polishing the end product,
93. 3:25and it enables me to do like things at
94. 3:29scale that that were just previously
95. 3:30impossible. And just to give you like a
96. 3:32taste of like what I'm talking about.
97. 3:34We're going to do this today. Like build
98. 3:35100 Facebook Ads, publish them to
99. 3:37Facebook.
100. 3:39Build [laughter] a dashboard to track
101. 3:40that, analyze the data within Claude
102. 3:43code, have it turn off the Facebook Ads
103. 3:45that are the low performers, have it
104. 3:47bump up the Facebook Ads that are the
105. 3:49best performers to a new ad set with its
106. 3:51own dedicated budget, and everything
107. 3:53that I just described that happening in
108. 3:54like literally, you know, 30 minutes.
109. 3:57>> [laughter]
110. 3:58>> And so, I'm Anyways, again, not really
111. 4:00sleeping.
112. 4:02So, this is kind of where it's at now,
113. 4:03and I'm going to talk through like this
114. 4:04whole setup process and actually how to
115. 4:06do this.
116. 4:07And then I'm going to talk about where
117. 4:09it's going, like how agents are the
118. 4:11natural evolution from this. Basically,
119. 4:14as soon as you like start you have this
120. 4:15epiphany like I can get this thing to do
121. 4:18work for me, then you suddenly have this
122. 4:21like you come to come to Jesus moment of
123. 4:23like, oh, I can just deploy this onto a
124. 4:25server, and now it's doing this task for
125. 4:27me in the background, and I'm building
126. 4:28out this personal software for myself,
127. 4:31for my job, for my my, you know, tasks,
128. 4:34etc. And this isn't some like hype thing
129. 4:37of like go do open claw and give it
130. 4:39access to everything. I'm talking about
131. 4:41like specific like jobs to be done
132. 4:44workflows that are custom made for how
133. 4:47you want to operate in your day-to-day.
134. 4:49So, that's kind of the high level, man.
135. 4:50Any questions I can try to answer? Happy
136. 4:52to go deeper on anything.
137. 4:53>> No. Um if you can teach me this by the
138. 4:56end of that episode, I mean, that's sort
139. 4:57of that That's I think the question that
140. 4:59a lot of people have in their heads
141. 5:01right now. Like, how do I
142. 5:04How can I do that, right? Cuz that's
143. 5:05going to be an unfair advantage.
144. 5:07So, yeah, let's let's go let's go
145. 5:09through it. Perfect. Let's jump into it,
146. 5:11man. All right. So, first off, what you
147. 5:13need to do if you're watching right now
148. 5:15is I want you to go and I need you to
149. 5:17create a folder that you're going to
150. 5:19start living out of. So, the one I live
151. 5:21out of is called Grasp Graph Growth
152. 5:23Agents. So, everything I do, now where I
153. 5:26start my work, it all exists within
154. 5:28here.
155. 5:29And the first thing that I'm going to
156. 5:31have you do is you're going to set up an
157. 5:34environment file. And this environment
158. 5:36file, it just holds all of your API keys
159. 5:39that you're basically going to be
160. 5:42working with. So, what I'm doing is I'm
161. 5:44basically having and I'm just not going
162. 5:46to show this just because it has
163. 5:47literally all of our API keys for
164. 5:49everything, but it has I can open up
165. 5:51this example one. So, it has
166. 5:55like Intercom, it has our SendGrid API,
167. 5:57my HubSpot API, my cal.com API, my
168. 6:00Perplexity API, my Facebook Ads API,
169. 6:03Inst Million Verifier, Instantly.
170. 6:06Everything that I live on top of that is
171. 6:08a part of my day-to-day growth stack,
172. 6:11this is like what I'm working with,
173. 6:13basically. Um
174. 6:15And so, what what this translates into
175. 6:17or like why I'm I'm why you start here
176. 6:20is you're basically starting to interact
177. 6:23with everything that you do on a daily
178. 6:25basis via the APIs. And this is actually
179. 6:27how I'm thinking about everything I do
180. 6:30now. And like how I buy software in
181. 6:31particular is how robust the API is.
182. 6:34It's funny, I was talking to a friend
183. 6:35recently, and he's like, if you're
184. 6:37looking at Salesforce versus HubSpot
185. 6:39right now, Salesforce, even though it's
186. 6:41like historically a more like clunky
187. 6:44CRM, it's actually the better product
188. 6:46for this AI foundation because it has a
189. 6:49more robust API, so you can do more with
190. 6:52it, basically.
191. 6:53And this is what this like turns into is
192. 6:55your all of the work that you're doing,
193. 6:58and we're going to do this together
194. 6:59today, is going to be happening from
195. 7:01this this like repository. And when I
196. 7:03when I say repo, all I mean is just this
197. 7:05like folder that we're living in that
198. 7:07has all of these files. And I'm going to
199. 7:09be using Claude code like throughout the
200. 7:11rest of the session to basically be
201. 7:12building out this personal software and
202. 7:14be building out
203. 7:16actually doing work, if that makes
204. 7:18sense.
205. 7:19So, that's kind of one component of it.
206. 7:21The last piece
207. 7:23is then I would strongly suggest get
208. 7:25suggest getting something like a Super
209. 7:26Whisper or any of these other
210. 7:29transcription softwares because it
211. 7:31enables you to just so quickly go
212. 7:33through the process of building out like
213. 7:35what you're trying to do on the
214. 7:36distribution side. And then optional is
215. 7:38just installing the Claude code front
216. 7:40end design skill. I've just found this
217. 7:42to be like one of those things that it,
218. 7:44you know, if we're going to generate a
219. 7:46UI, it's nice to have it look pretty.
220. 7:47So, All right. That is kind of the
221. 7:50foundational pieces. Now, let's actually
222. 7:52like, great, that's cool. You've just
223. 7:54built this. What do you actually do to
224. 7:55go I get started on this? So, the first
225. 7:57thing that I'm going to do to get
226. 7:58started is I'm going to go and I'm going
227. 8:00to have Claude code start responding to
228. 8:02people on LinkedIn for me
229. 8:05that have asked for an asset. So, I've
230. 8:07been doing all of these like giveaways,
231. 8:09basically. Here's one that's an email
232. 8:11triage. I wrote this giveaway you know,
233. 8:15Notion document. I'm now going to go in
234. 8:17and I'm going to get this agent to start
235. 8:19running for me in the background while I
236. 8:20have other work going on. So, I've got
237. 8:22the Claude code or sorry, I've got the
238. 8:24Claude Chrome extension installed. And
239. 8:27I'm going to go and I'm going to say,
240. 8:29so I'm working out of this that
241. 8:31directory, right? That I've already been
242. 8:33in.
243. 8:34And I already have built this basically
244. 8:37skill, and it's a piece of software that
245. 8:38will go and comment on everybody that
246. 8:41asked for this asset. So, I'm going to
247. 8:43say this right now. I'm going to say,
248. 8:46let's run we're going to transcribe it.
249. 8:48So, let's run the LinkedIn respond
250. 8:52software.
251. 8:54Keyword that you're looking for is
252. 8:55triage. I'm going to provide the Notion
253. 8:58documents and the LinkedIn post URL.
254. 9:02And then I'm going to select the post
255. 9:05URL, and I'm going to put that in. And
256. 9:07then I'm also going to select the Notion
257. 9:10document that I want it to do as well.
258. 9:11So, I'm going to give it that. It's
259. 9:13going to start running. So, what it's
260. 9:14going to do right now is it's going to
261. 9:16basically open this up. I'm going to
262. 9:17baby sit it for a moment while it while
263. 9:19it starts this process. And just make
264. 9:22sure that it starts on the right path.
265. 9:24And then once it's on the right path,
266. 9:25then I'm going to go and basically start
267. 9:27on the other things. And actually, while
268. 9:28it's thinking, let's just go to these
269. 9:29other places. So, next thing I'm going
270. 9:31to do is I'm going to go or sorry, that
271. 9:36just opened the LinkedIn profile. Let's
272. 9:37bring that back over here.
273. 9:39So, this is now running.
274. 9:43Perfect. And so, this should now start
275. 9:45commenting on those responding back to
276. 9:47those people. I'm just going to change
277. 9:49this to most recent just so that it
278. 9:50works backwards on this.
279. 9:52And then we're going to just let that
280. 9:54run in the background. So, while that's
281. 9:56happening, what I'm going to do now is
282. 9:59we're going to build a Facebook ads
283. 10:01generator. So, I've been doing this
284. 10:03where it's basically a and I'll show you
285. 10:06an example of what this looks like.
286. 10:07Let's just go over to LinkedIn and I can
287. 10:11give you an example of the output that
288. 10:13you're we're going to actually create
289. 10:15today. So, it's basically a bulk
290. 10:17generator of ad creative. We're going to
291. 10:19create this template and then I'm going
292. 10:21to go and do research based off of
293. 10:23Reddit
294. 10:25and other social media posts for the
295. 10:26pain points that people experience and
296. 10:28then we're going to go and bulk generate
297. 10:29all these variations. So, let's get that
298. 10:32started right now.
299. 10:33So, I've got again those API keys are
300. 10:37stored within here and I've also I
301. 10:39made that that skill. So, I'm going to
302. 10:41tell it right now, I want you
303. 10:45I want you to create a bulk Facebook ad
304. 10:47generator. It's going to be a 1080 by
305. 10:491080 pixel image.
306. 10:51What's going to happen is I'll give you
307. 10:54an example of what one of these ads
308. 10:56looks like and then we're going to go
309. 10:59and build a template around that and
310. 11:01then I'll basically create a or give you
311. 11:04variations of text both titles and
312. 11:07paragraphs that we I want to be
313. 11:10generated. It'll be a zip file that we
314. 11:11download for the beginning. Can you make
315. 11:15this into a UI as well so that we can
316. 11:17visually see
317. 11:18the creative.
318. 11:21For the first thing I just wanted to be
319. 11:22able to see is like what the actual
320. 11:24creative look like. For this you're
321. 11:26going to use
322. 11:28just react components.
323. 11:30So, I don't want you like just purely
324. 11:33build it with react components and then
325. 11:36also to actually change this react
326. 11:38components into a PNG that's
327. 11:40downloadable, we're going to use HTML to
328. 11:44canvas. It's just a
329. 11:47you know resource that you have
330. 11:48available for you on that. Ask questions
331. 11:50if you need. So,
332. 11:52I've just transcribed that.
333. 11:54I'm now going to put Claude into plan
334. 11:57mode and I'm just going to let that
335. 11:58start running in the background.
336. 12:00All right. So, while that's running in
337. 12:01the background, let's come back here and
338. 12:03let's see the work that it's doing. So,
339. 12:05it's going through these and I believe
340. 12:07it's now commenting. So, that's
341. 12:09happening. So, we'll just let that run
342. 12:11in perpetuity. Respond back to them. All
343. 12:14right. So, next I'm just going to just
344. 12:16click through this quickly.
345. 12:19I'll share it now. I'll share it
346. 12:22I'll share it after setup. And then
347. 12:24input method form based UI, let's do
348. 12:27both and then I'm going to hit submit.
349. 12:29All right. So, now that's working on
350. 12:30that in the background. I'm going to
351. 12:32open up another folder and I'm going to
352. 12:36start Claude code again within an
353. 12:37entirely another another window. So, I'm
354. 12:40going to do documents for slash graphs.
355. 12:44Let's go to agents
356. 12:47and then demo.
357. 12:50All right. So, the next thing that I
358. 12:51want to build as an example
359. 12:54is
360. 12:55I'm going to
361. 12:57basically pull information
362. 13:00so I just I I just did this actually so
363. 13:03we could like talk through
364. 13:05about what this ends up looking like but
365. 13:07basically scraped all of the podcasts
366. 13:09that were within the marketing category
367. 13:11and then built a workflow that goes and
368. 13:13cold emails them and then an agent that
369. 13:15responds back to book me on the podcast.
370. 13:18This ends up turning into way better
371. 13:21performing than I expected. This is what
372. 13:23my week That's crazy. So, what is
373. 13:25Instantly? Yeah, so Instantly is a cold
374. 13:28email software
375. 13:29and so this is just a part of my stack.
376. 13:31So, it's one of the things that like is
377. 13:34within that environment file that allows
378. 13:36for me to build on top of. And so, what
379. 13:39I'm
380. 13:40like how how I can think about this like
381. 13:43on the
382. 13:44is basically like my manual workflows
383. 13:46that I would do previously were just
384. 13:48like daisy chaining those together like
385. 13:49using this software. So, I'm going to
386. 13:51bring this into a new desktop and let's
387. 13:53just rebuild that whole thing. I'm going
388. 13:55to say um
389. 13:56you have the Refonic API key. I want you
390. 14:00to build a software that scrapes podcast
391. 14:03host emails from Refonic. It then sends
392. 14:06it to Million Verifier to verify the
393. 14:09emails and then it also will then send
394. 14:12it to an Instantly
395. 14:15campaign. I'll provide the Instantly
396. 14:17campaign that I want it to send to.
397. 14:21All right. So, I've got that now.
398. 14:24I'm going to put that in the plan mode
399. 14:25and let that run as well
400. 14:27and then we have that as its own window.
401. 14:31And so, while these two are working
402. 14:33again in the background we can then go
403. 14:37and actually do like some other work.
404. 14:40So, let me get this going and I'll just
405. 14:43say
406. 14:44cool. So, that's in plan. All right. So,
407. 14:47now this is my we're in this folder.
408. 14:49This is like my demo folder which I just
409. 14:51like every time I give this presentation
410. 14:53I just nuke. This is kind of to show you
411. 14:55like how to start it from zero to one.
412. 14:58This folder that we're in now is my
413. 15:00actual folder that I live out of
414. 15:02and I'm just going to show you some of
415. 15:03the things that like are like capable
416. 15:05with this. So, for example, I have it
417. 15:07attached to Notion and I've basically
418. 15:10given it a
419. 15:12an example of like how do we write a
420. 15:16like a Notion like giveaway, right? So,
421. 15:18I'm going to go and we're going to
422. 15:19create one of these together right now
423. 15:21because I need to actually accomplish
424. 15:22this. So, I'm going to give take this
425. 15:24URL
426. 15:25and then copy this over
427. 15:28and I'm going to say, okay. Um
428. 15:30write a or create a Notion document
429. 15:34uh
430. 15:36based off
431. 15:36>> Look at you typing with your hands.
432. 15:39>> The problem here we can do it in the the
433. 15:41transcription of. So, create a Notion
434. 15:43document based off of
435. 15:46our like current you know our
436. 15:48structure. Look for the skill that has
437. 15:49this. I'm going to provide context on
438. 15:51what that should include. You should
439. 15:54incorporate stuff that we have we
440. 15:57haven't within the repo like the
441. 15:58documentation that I have in the repo on
442. 16:00how to do this.
443. 16:03All right. And so, I'm going to copy
444. 16:04this over
445. 16:07and let that run and now it's going to
446. 16:08go and create me a Notion document just
447. 16:10like the one that we have being sent in
448. 16:12the background here currently.
449. 16:14All right.
450. 16:15So, in that folder I've already created
451. 16:18the bulk ad generator. So, I'm just
452. 16:21going to go in there just to show you
453. 16:23like what you can do with this once that
454. 16:25it's it's like you've actually gone
455. 16:27through this process of like zero to one
456. 16:30making this. So, this is the bulk the
457. 16:32creator as an example. So,
458. 16:34that's going to continue working on that
459. 16:36bulk Facebook ad generator in the
460. 16:37background. While that's happening,
461. 16:39let's go to graphs
462. 16:41growth
463. 16:43agents and then I'm going to start
464. 16:44Claude code within there.
465. 16:47And now I'm going to
466. 16:49start locally the bulk Facebook ad
467. 16:51generator. I just want to bring that up
468. 16:54within
469. 16:56my
470. 16:57local so that I can create some ads.
471. 17:02So, again it's already created the
472. 17:03software for me. I'm now coming back to
473. 17:05it and I'm going to talk through the
474. 17:06whole process of the actual creation of
475. 17:08this cuz it'll just like make more sense
476. 17:10momentarily. But the goal here is
477. 17:13basically what we're trying to do is I'm
478. 17:17trying to create as many different
479. 17:19variations of these ads as possible. So,
480. 17:22how did I make this ad? This is entirely
481. 17:25code and I think this is something to
482. 17:27like double click on. Um everything in
483. 17:30here is just it's just react components
484. 17:33like that. This is this is just a react
485. 17:34component and entirely built by code.
486. 17:36And so, I can make an infinite amount of
487. 17:39these
488. 17:40at scale and I've done this already. So,
489. 17:43you can see this bulk for slash or for
490. 17:45slash bulk for slash HTML. So, we're
491. 17:48going to go through this together
492. 17:49though. How do we actually like do this
493. 17:50process? So, I I would go and I would
494. 17:52give it an example and we'll do this
495. 17:54over here while this one is like working
496. 17:56on it. We'll come back to that. But I
497. 17:58would give it an example of what
498. 18:01like ad I'm trying to build. A way to do
499. 18:03this is if you don't know where to
500. 18:04start, go to Facebook ads library and
501. 18:07you can see what your like competitors
502. 18:09or other software companies are doing in
503. 18:12your category. This is actually how I
504. 18:14made this initial one was I found like
505. 18:16this before and after format and then I
506. 18:19basically had it build off of that
507. 18:21before and after. But this whole thing
508. 18:25like everything you see here
509. 18:27is just code. [laughter]
510. 18:29It's entirely code.
511. 18:31The other way you can do this is with
512. 18:32something like Nano Banana where you're
513. 18:34like going and bulk generating these,
514. 18:35right? But once I found and I built that
515. 18:39template, I can now make all those
516. 18:41variations. So, let's go back to Claude
517. 18:43code and let's say
518. 18:47okay, I want you to use the Facebook ads
519. 18:49API and I want you to go and scrape the
520. 18:52pain points that you see or sorry
521. 18:55actually let's restart that. I want you
522. 18:57to use the Perplexity API and go and
523. 19:00scrape Reddit for the pain points and
524. 19:03the outcomes that growth marketers wish
525. 19:06they could have from like something like
526. 19:08a Looker Studio
527. 19:10or any of these other business
528. 19:11intelligence softwares that they're
529. 19:13using. We've been
530. 19:15focusing on like the
531. 19:17the data analyst component of it of how
532. 19:19they can't get bandwidth or it's too
533. 19:21complicated to get started or they can't
534. 19:23unify their data all into one location.
535. 19:25You can also source from YouTube. You
536. 19:27can also source from Twitter if
537. 19:28necessary. So, I'm going to have it go
538. 19:31do research
539. 19:32and those pain points. For the ad sorry,
540. 19:35for the ad itself,
541. 19:37wouldn't we want to use Nano Banana Pro
542. 19:39like the best image model that exists?
543. 19:41Like why are we using code when you know
544. 19:44when we could
545. 19:44>> Yeah, I'm just doing this purely like
546. 19:46this was just a way I thought about
547. 19:47doing it like absolutely the Nano Banana
548. 19:50thing. The the only thing I found with
549. 19:53like Nano Banana is that
550. 19:56I sometimes have trouble like getting it
551. 19:58to stay on brand. And if I'm trying to
552. 20:00just like figure out the messaging
553. 20:02variations that I'm trying to go after,
554. 20:04this can be a faster way to do that.
555. 20:06Again, there's like a million different
556. 20:07ways to do this exact same thing.
557. 20:11I think yeah, if you're going to use the
558. 20:13mana Nano Banana, you should look at
559. 20:14something like um
560. 20:16I think it's like Kai AI, I believe.
561. 20:20Uh, Nano Banana.
562. 20:23Erk, yeah.
563. 20:25Kai dot AI. Um, and you can just like
564. 20:28basically it's one of these bulk buys.
565. 20:30Um, but we've been using that for these
566. 20:32bulk generations.
567. 20:33This that I'm doing cost me nothing.
568. 20:36Like it's literally on you know,
569. 20:39maybe a thousand tokens to do all of
570. 20:41these like these generations. And so
571. 20:43this that's a reason for it. So it's
572. 20:45like I can go we can go and create a
573. 20:47thousand ad variations right now, G. And
574. 20:50like this literally costs nothing.
575. 20:53>> [laughter]
576. 20:54>> Um
577. 20:55>> Yeah.
578. 20:55But again, we
579. 20:56>> that's a part of it, too, right? Is like
580. 20:59your goal is going to be come up with
581. 21:02the best ad creative that's going to
582. 21:03actually, you know, you put in a dollar,
583. 21:05you get three dollars out. Once you get
584. 21:08that
585. 21:09uh you can make it Well, there's there's
586. 21:11two schools of thought. One school of
587. 21:12thought is you need the best creative.
588. 21:15So you need to send it to Nano Banana
589. 21:16Pro to get scroll stopping creative.
590. 21:19Another school of of thought is like,
591. 21:21well, you actually have some pretty
592. 21:23{quote} ugly ads that just speak to the
593. 21:26IC the pain points that you can kind of
594. 21:29get a good understanding of this is
595. 21:30going to bring you a dollar 50
596. 21:33uh when you put in a dollar. And then
597. 21:35you can get it from a dollar 50 to three
598. 21:37dollars once you figure out the ad. So
599. 21:39you're kind of saying maybe it's best to
600. 21:41like
601. 21:42get as many ads as possible
602. 21:45start using the those, figure out the
603. 21:47one or two or three creative that
604. 21:49actually crushes and then go ahead and
605. 21:52go crazy with
606. 21:54spending tokens. Totally. Like this I
607. 21:57guess a different like I'm just trying
608. 21:58to find like the the format or the angle
609. 22:02that's going to be most receptive. I'm
610. 22:04going to remix that a thousand times,
611. 22:06you know, after it. Um,
612. 22:08the this is like the big piece of this,
613. 22:11especially with like how much we can do
614. 22:14um on the uh
615. 22:16like
616. 22:18the the Anybody can go and generate as
617. 22:20many of these as they want, right? Like
618. 22:22it's literally infinite. But identifying
619. 22:24those winners like you're talking about
620. 22:26now becomes the challenge that you're
621. 22:28going to face with all of this. And like
622. 22:30we're going to get to that in a second,
623. 22:31but the
624. 22:33the main thing I want to emphasize with
625. 22:35it is like this is malleable and
626. 22:37flexible and the spit the pace like just
627. 22:39think about you manually having to go
628. 22:41create 50 ad variations in Figma. And
629. 22:44like
630. 22:45>> [laughter]
631. 22:45>> you can just now make those and get
632. 22:47those live and test those. And as soon
633. 22:49as I find a winning like format from a
634. 22:52uh
635. 22:53a language that's being said
636. 22:54perspective, I can then go and remix
637. 22:57that into all these other different
638. 22:59templates. I can go and find like what
639. 23:01are different winning ad formats that I
640. 23:03can now port this to. But this is a way
641. 23:06to just like start immediately and then
642. 23:08get that basically out in the in in in
643. 23:10public. I also think that like the same
644. 23:12ideas can go into different formats.
645. 23:14Like we're all you know, already seeing
646. 23:15this where it's like cool, I made you
647. 23:17know, a static format. I'm now going to
648. 23:18port that over to a UGC format and I'm
649. 23:20sending that to the HeyGen API to pull
650. 23:23in that uh
651. 23:26you know, like to make that creative
652. 23:27like pull in pull it in as a video and
653. 23:30then bulk upload that to Facebook if
654. 23:32that makes sense. So I
655. 23:34like where I'm going with this or where
656. 23:36I'm seeing this head personally is like
657. 23:38I'm going to build these tools
658. 23:41that an agent is going to have
659. 23:43and then it's going to be able to run
660. 23:45this process in the background where
661. 23:47it's basically has the ability to make
662. 23:49new creative. It can publish that
663. 23:50creative to directly to Facebook, which
664. 23:52I'm going to show you in a second. Um,
665. 23:55it's going to then analyze what is
666. 23:57working and then it's going to like
667. 23:59basically turn off what isn't and
668. 24:00promote what is, right? And this is like
669. 24:02everything I'm going to show you today
670. 24:03is at like a small scale. Like this is
671. 24:04literally like weeks of realization
672. 24:07that's starting to happen like with
673. 24:08this. But I just again, I want to plant
674. 24:11the seed of like what is possible using
675. 24:13this tooling to like do your again, that
676. 24:16middle work that historically like you
677. 24:18wouldn't be able to to do. So yeah, it
678. 24:22Does Does there questions about that I
679. 24:23can try to answer? No. Let's
680. 24:25Let's keep cooking. Awesome. All right,
681. 24:27so this keeps like trying to scrape
682. 24:29things. I'm just going to be like um
683. 24:32Instead I want you to just brainstorm uh
684. 24:34pain points that people have uh with
685. 24:36data reporting, specifically uh the
686. 24:38unification of the data into multiple
687. 24:40locations.
688. 24:43Cool. So we're going to do that. Um
689. 24:45Once I have all those variations, I can
690. 24:47then say, "Okay, now go bulk generate
691. 24:50every one of those."
692. 24:52And then at that point, I can download
693. 24:55these as a CSV.
694. 24:57And
695. 24:58what I will do is just wait for that to
696. 25:02While that's happening, we'll just get
697. 25:03this to start downloading.
698. 25:05And that Facebook Ads API is going to
699. 25:06allow for us to bulk upload all of those
700. 25:09pieces of creative that we just
701. 25:10downloaded. So here's all those
702. 25:12variations. I'm going to say now, "Okay,
703. 25:14now uh use the bulk or let's do the
704. 25:18transcription."
705. 25:19Okay, now use the bulk uh Facebook Ads
706. 25:21Generator to go and create these
707. 25:23variations. Um,
708. 25:25put them in the forward slash bulk dot
709. 25:28HTML page when you're done with this.
710. 25:31So this is now happening. I'm going to
711. 25:32go back over to see what the other
712. 25:34things are going off of. All right, so
713. 25:38let's see this notion document based off
714. 25:40the current. We'll continue to let that
715. 25:41happen.
716. 25:43Um, I want to build this up.
717. 25:45Perfect. We're going to let that go as
718. 25:47well.
719. 25:48So I'm looking at the scaffolding to
720. 25:50build the bulk ad generator.
721. 25:52And we'll check in on this one to see
722. 25:53where it's like basically at in its
723. 25:55process of responding.
724. 25:57And it might have completed.
725. 25:59And it did complete. So that's done. It
726. 26:01ran for 15 minutes on its own in the
727. 26:03background.
728. 26:04These are coming back. Um, so while
729. 26:06these are all happening and I'm waiting
730. 26:07on them or they're waiting on me, I can
731. 26:09then open up another
732. 26:12one of these. I'll just click through
733. 26:13this uh to just kind of get it moving.
734. 26:16Um, but I would then open up another one
735. 26:18of these tabs and I would start on the
736. 26:19next project. So the next thing that I
737. 26:21want to do is I want to build a LinkedIn
738. 26:23um engagement scraper. Um, so I'm going
739. 26:27So basically uh people that engage on
740. 26:29the LinkedIn pro or the LinkedIn post, I
741. 26:31want to pull them out and then send them
742. 26:33to a um
743. 26:36uh basically add them into Instantly. So
744. 26:38we're going to go we're going to find
745. 26:40their LinkedIn profile using
746. 26:41PhantomBuster. We're then going to uh
747. 26:44you know, do that whole flow. So I'll
748. 26:45I'll do that in a second. Let me just
749. 26:47get this up.
750. 26:48We'll give that its own section.
751. 26:51By the end of this podcast, you're going
752. 26:52to have like
753. 26:53a hundred desktops.
754. 26:54>> This is literally how I'm working now.
755. 26:56>> [laughter]
756. 26:57>> This is like I'm just jockeying agents
757. 27:00across and then if I can automate them
758. 27:03and get them to do like like if I can
759. 27:06figure out, "Okay, this is the specific
760. 27:08lane that you can focus on."
761. 27:10Then I'm spinning that up onto a server
762. 27:12on Railway and I'll talk about that in a
763. 27:14second on like how you can on demand
764. 27:17create databases and on demand create
765. 27:20these servers so that this software
766. 27:21starts running in perpetuity. So
767. 27:24we're going to do this one.
768. 27:26All right, so I want to make a workflow
769. 27:28where it's a
770. 27:29I basically within Slack you'll do
771. 27:31forward slash LinkedIn post and anybody
772. 27:35in Slack will be able to just drop in a
773. 27:37LinkedIn post that they think is a good
774. 27:38fit. And then that's going to go and
775. 27:40it's going to use the PhantomBuster API
776. 27:43to extract all of the engagers. And then
777. 27:46it's going to take those LinkedIn
778. 27:48profiles. We're going to go and enrich
779. 27:49those with the Apollo API.
780. 27:52And then from there we're going to send
781. 27:53it to the MillionVerifier API. And then
782. 27:55finally we're going to add them to an
783. 27:57Instantly
784. 27:58uh campaign. Ask me questions if you
785. 28:01need.
786. 28:02All right, so I'm going to turn on plan
787. 28:04mode for that. Let that start running in
788. 28:05the background as well.
789. 28:07We'll come back to these. Let them
790. 28:09continue to cook.
791. 28:11Publish the landing page. Let me get
792. 28:13context on what this one is again.
793. 28:16This is the craziest part when you're
794. 28:18going from screen like uh screen to
795. 28:21screen and realizing that you're an
796. 28:23agent jockey. Yep. And then you're
797. 28:25trying to get context on each one.
798. 28:28You're like, "Okay, what was this one
799. 28:29doing again?" And I find that the
800. 28:31context switching is actually difficult.
801. 28:33I I I did as well, but now it's like now
802. 28:37it feels like I like that has expanded.
803. 28:39Like and again, this is just how I've
804. 28:41been working for the last like six
805. 28:43weeks. And it was like maybe I could
806. 28:45have like two or three of them in the
807. 28:47beginning and now it's like I'm
808. 28:48comfortable with like we could have 50
809. 28:50windows open. I'm about to literally go
810. 28:51buy a new computer cuz I'm like I need
811. 28:53more RAM. I need more
812. 28:55>> [laughter]
813. 28:56>> like ability to do this in the
814. 28:58background.
815. 28:59I'm just It sounds so stupid. Totally.
816. 29:02Yeah.
817. 29:03Like Like and comment this video so
818. 29:06that, you know, I could send some
819. 29:07YouTube AdSense revenue to Cody so
820. 29:09>> [laughter]
821. 29:10>> he can get some more RAM. Uh it's all
822. 29:12ridiculous. This guy needs some RAM. No,
823. 29:14man. We I it's like I I'm just realizing
824. 29:17like what this turns into. So okay, we
825. 29:19built we made all those pieces of ad
826. 29:21creative, right? We've got those
827. 29:23variations. And it's just text
828. 29:25variations.
829. 29:27All right. So now I'm going to go back
830. 29:30to Claude and I'm going to be like I
831. 29:32want
832. 29:33uh
833. 29:34Now I want to buckle a bulk upload all
834. 29:36of these ads as drafts into um a
835. 29:39Facebook ad set.
836. 29:41Um, here is the Here's where the um
837. 29:44folder is locally for the
838. 29:46um creative and I'm going to provide the
839. 29:48Facebook ads uh ad set URL to you in a
840. 29:51second.
841. 29:53All right, so I'm now going to go back
842. 29:55to finder
843. 29:57and I'm going to copy this
844. 30:02paste that in
845. 30:04and then let's go back to Facebook.
846. 30:08And I have this
847. 30:09ad set that I've already created for
848. 30:11this demo.
849. 30:12It's basically just here.
850. 30:15And I'm just going to paste in this URL.
851. 30:17And so now
852. 30:20here's that URL.
853. 30:22It's going to
854. 30:23basically
855. 30:25bulk upload all of those pieces of
856. 30:27creative into that ad set. So while
857. 30:30that's happening, I'm now going to go
858. 30:32and I'm going to create a dashboard
859. 30:34about this. So let's just pull out the
860. 30:37ad set ID.
861. 30:39Let's do the ad set ID and I'm going to
862. 30:42go over to Graft and I'm going to be
863. 30:45pull up Facebook ads as the data source.
864. 30:49And then I'm going to be say uh
865. 30:51this is the ad set
866. 30:54ID.
867. 30:58Make a dashboard showing clicks over
868. 31:00time.
869. 31:02Also have a scorecard that or sorry, I
870. 31:04didn't do the transcription. Make a
871. 31:06dashboard showing clicks over time as a
872. 31:09line chart. Also
873. 31:10within that line chart, can you include
874. 31:14the cost and the CPC as lines as well?
875. 31:17And then add a scorecard also that has
876. 31:20total spend, total traffic or total
877. 31:22clicks as another scorecard. So those
878. 31:24are two separate ones. And then I want
879. 31:27you to also show demographic data as a
880. 31:30bar chart
881. 31:31of showing the ages. So that's its own
882. 31:34separate chart. That's a bar chart
883. 31:36basically showing the impressions by the
884. 31:38age categories.
885. 31:41So I'm going to let that run in the
886. 31:42background. We'll come back to that in a
887. 31:45second. But now that I'm
888. 31:47have this campaign that's running
889. 31:50and I'm trying to track what's happening
890. 31:52within it, I can basically go and like
891. 31:56build out a tracking dashboard for this.
892. 31:59The other thing that I can do so once
893. 32:01it's got the ad set with destination URL
894. 32:04which you like just put them
895. 32:06as
896. 32:08a draft. So
897. 32:10while that's working, I can also analyze
898. 32:13what is happening
899. 32:15in that specific ad campaign.
900. 32:18And I can turn off the losers of that ad
901. 32:21campaign. So I'm going to show you how
902. 32:22to do that. So I do documents forward
903. 32:24slash Graft
904. 32:26forward slash
905. 32:27growth agents
906. 32:30and then Claude
907. 32:32and then I'm going to get that URL again
908. 32:35and I'm going to say uh
909. 32:38use the Graft MCP to pull in the data
910. 32:41for this ad set that I'm about to
911. 32:43provide from Facebook ads.
912. 32:45I want to look at the CPM data to see
913. 32:49which ones are the lowest performing
914. 32:51like the highest CPM price.
915. 32:55All right,
916. 32:56and then let's provide the ad set URL
917. 32:59again
918. 33:01let's go ad set
919. 33:04and then while that's happening, we can
920. 33:05come in and check on the other ones.
921. 33:08All right, so it's now built that entire
922. 33:11bulk Facebook ad generator
923. 33:13or sorry, which one is this is the
924. 33:16look one. Okay, so it's made the updates
925. 33:19to these ads. So these are an entirely
926. 33:22new ad set that it's basically pulling
927. 33:25in.
928. 33:26So I already did this, right? Of like
929. 33:29downloading these as a zip and bulk
930. 33:31uploading them. So we won't go through
931. 33:33that process again, but this is how easy
932. 33:35it is to basically make those
933. 33:36variations.
934. 33:38So
935. 33:39>> Not and not just variations. These are
936. 33:41variations based on pain points that
937. 33:43people have said publicly. Yes, exactly.
938. 33:46So it has pulled in basically
939. 33:49the social dialogue that's happened. So
940. 33:52the best ads that I'm seeing perform
941. 33:54right now are basically you're selling
942. 33:55outcomes or
943. 33:57you're you're talking to the pain
944. 33:58points, right? So I'm just guiding it to
945. 34:02focus on those things, pull me that
946. 34:03information and then build the ad sets
947. 34:06around those.
948. 34:08Well, that just happened. This in the
949. 34:10background just used the Graft MCP to
950. 34:12pull in all of the low performers. So
951. 34:16these are the let's look at the ones
952. 34:17that have the highest CPM. So these all
953. 34:19have high CPM. So I'm just going to tell
954. 34:21it turn these off.
955. 34:22I'm going to say uh
956. 34:25use the Facebook ads API to turn off
957. 34:29these ads with the this ad name.
958. 34:32And so now it's just pulled in this live
959. 34:34data from my data warehouse. And this
960. 34:38isn't an MCP that's interacting with the
961. 34:40Facebook ads API. I just want to like
962. 34:42emphasize this. So it's not have you're
963. 34:44not running into rate limits.
964. 34:46Like again, you can publish [laughter]
965. 34:48like
966. 34:49a thousand ads and are running those
967. 34:51variations. There is literally no way
968. 34:53like if you're spending enough, like
969. 34:54there is literally no way that you're
970. 34:56going to be able to analyze this data
971. 34:58without a data pipeline and a data
972. 35:00warehouse. And so this is like what
973. 35:02we've built, right? At Graft.
974. 35:04So anyways
975. 35:06the
976. 35:08just to get back to the MCP thing,
977. 35:09there's like this pagination problem. So
978. 35:11like we see this all the time where
979. 35:13people are like, yeah, I plugged in a
980. 35:14Facebook ads MCP I'm interacting with it
981. 35:15and then I realize that I'm only seeing
982. 35:17like 5% of the data that I think I'm
983. 35:19actually seeing, right?
984. 35:20Um
985. 35:21but so coming back to this
986. 35:24I've now said, hey, turn these off. So
987. 35:26it went and it paused those ads for me.
988. 35:28And so it's just just to walk through
989. 35:30what we just did just kind of reiterate
990. 35:32this. We just did ideation.
991. 35:34We just did bulk ad creation.
992. 35:38We just analyzed the data for the
993. 35:39performers.
994. 35:41We just turned those off and on based on
995. 35:43that.
996. 35:45And at this point, you're probably
997. 35:47starting to have the epiphany like, oh,
998. 35:48I can just turn this in a into a
999. 35:50repeatable process.
1000. 35:52And this is where I see all of this
1001. 35:54going basically is you're going to going
1002. 35:57to have these agents
1003. 35:59that are running on top of your live
1004. 36:01data.
1005. 36:02They're analyzing it, making decisions
1006. 36:05based off of like
1007. 36:07the the model. So like for example, how
1008. 36:09I would run this is I would have a test
1009. 36:11campaign where I'm basically testing new
1010. 36:12creative constantly. I would have a cron
1011. 36:14job that's on a daily basis basically
1012. 36:16going and turning off the low
1013. 36:18performers. And then the high
1014. 36:19performers, they get bumped up into
1015. 36:21their own ad sets with their own
1016. 36:22dedicated ad budget for CPA action. And
1017. 36:26then that whole thing could just run
1018. 36:27automatically in the background. And
1019. 36:29then to track that, I'd build out a
1020. 36:30dashboard that basically is showing me,
1021. 36:33you know, that information so I can come
1022. 36:35back to that like later on and see
1023. 36:37what's going occurring there.
1024. 36:38And then the other thing that I would
1025. 36:40then go do is
1026. 36:43like potentially have a conversation in
1027. 36:45the morning. Say for example, so I've
1028. 36:48got the Graft MCP
1029. 36:50in my Claude I
1030. 36:53like chat. So this is also technically
1031. 36:55on my phone. So like in the morning,
1032. 36:57I'll wake up and be like, how much
1033. 36:59traffic Let's just do this. How much
1034. 37:02traffic went to
1035. 37:03or how many new users went to the
1036. 37:05homepage of the website yesterday?
1037. 37:07I'll just say use
1038. 37:11the Graft MCP and Google Analytics 4.
1039. 37:16And we'll let that run.
1040. 37:17And so I can basically get a brief each
1041. 37:20morning whatever those KPI metrics are
1042. 37:22that I care about and have a
1043. 37:23conversation like with my data that's
1044. 37:26live and being seen continuously within
1045. 37:28the background. And you can give this to
1046. 37:30your whole team as well so that
1047. 37:32everybody on your team also has access
1048. 37:34to this both from within Claude code,
1049. 37:37within their [laughter] like whatever
1050. 37:39their harnesses that they use whether
1051. 37:41that's chat GPT or Claude. And then also
1052. 37:43the ability to like do that tracking
1053. 37:45within like dashboarding or
1054. 37:47conversations. Now
1055. 37:49coming back to all of this, right? We've
1056. 37:50just built out basically this whole like
1057. 37:53cycle of funnel. How would I now go
1058. 37:55deploy this? So this is where it gets
1059. 37:57like the most interesting. So what I'm
1060. 37:58doing right now, say I wanted to turn
1061. 38:00this into an agent.
1062. 38:02I'm using Railway for this and only
1063. 38:05reason is just because I saw a tutorial
1064. 38:07and that's how I've like figured out how
1065. 38:09to do this. So Railway has a really
1066. 38:11robust API key and say I wanted to spin
1067. 38:14up for example, this bulk ad generator
1068. 38:17so that my other team members could use
1069. 38:20this, right? They could come back and
1070. 38:22basically like use this software that
1071. 38:24I've created. I can just tell Railway
1072. 38:26via Claude code, hey, spin this up into
1073. 38:30a
1074. 38:31like
1075. 38:33a a server that I can access
1076. 38:35or I can just deploy this directly to
1077. 38:37Vercel
1078. 38:39or any of these workflows that I have.
1079. 38:40So say for example, we were talking
1080. 38:42about that LinkedIn
1081. 38:44funnel. Let's go see which one of this
1082. 38:47it is.
1083. 38:50All right, so this is the podcast
1084. 38:52software.
1085. 38:55This is the image generator.
1086. 38:58This is the LinkedIn.
1087. 39:00So say I wanted I want this to be
1088. 39:03accessible in perpetuity in the
1089. 39:05background. I can then take the software
1090. 39:07that I co-work on with Claude
1091. 39:10and I say, okay, deploy this to a server
1092. 39:13on Railway so that my whole team can
1093. 39:15basically use this action. Always be
1094. 39:17adding
1095. 39:18like whenever they come across a
1096. 39:19LinkedIn post as an example, they could
1097. 39:21be adding that into the queue so that it
1098. 39:23just automatically goes into the email
1099. 39:27like filtering. Sorry, the email like
1100. 39:30cold email process.
1101. 39:32And this even goes further. So like how
1102. 39:35I'm starting to use this G, and I'm
1103. 39:36curious like to
1104. 39:37>> [laughter]
1105. 39:38>> get your thoughts on this. So basically
1106. 39:39like I had to do some data analysis work
1107. 39:41the other day. Historically, I would
1108. 39:42have like downloaded the information,
1109. 39:44put it into Excel, and then I do a bunch
1110. 39:46of pivot tables.
1111. 39:47Now, instead, what I did was I I went to
1112. 39:50directly to the URL. I had it push it
1113. 39:53into a Postgres database that I on the
1114. 39:55fly created using the Railway API. It
1115. 39:58just pumped everything in there. I then
1116. 40:00did the analysis together with Claude,
1117. 40:03and then at that point, I basically
1118. 40:05pushed from that Postgres database
1119. 40:09the outputs to the the location that I
1120. 40:12wanted. What it used it would have it
1121. 40:14would have taken me probably 5 hours
1122. 40:16historically to like clean the data
1123. 40:18appropriately, and I smashed that out in
1124. 40:20like probably 20 to 30 minutes.
1125. 40:23And then as soon as I got done with that
1126. 40:24database, I just spun it down.
1127. 40:27>> [laughter]
1128. 40:29>> And this was the most like interesting
1129. 40:31part of this was that like it was
1130. 40:32basically like on the fly UI on the fly
1131. 40:37the epiphany I had was like on the fly
1132. 40:39UIs on the fly databases like on the fly
1133. 40:42software is going to become the standard
1134. 40:44for these people that are like, you
1135. 40:47know, working at the forefront of this.
1136. 40:48So,
1137. 40:50um yeah, man. I I we could probably, you
1138. 40:52know,
1139. 40:53sit here and watch me work for hours if
1140. 40:55you wanted, but that's I kind of
1141. 40:56everything I had that I wanted to show
1142. 40:58you today. The only other thing I'm just
1143. 41:01like keep getting asked like how do I do
1144. 41:03this how like show me more technical
1145. 41:05details. Um
1146. 41:07I bought the domain
1147. 41:09gtmengineeringcourse.com.
1148. 41:11I'm going to give this thing away for
1149. 41:12free to everybody who wants it. It'll be
1150. 41:15entirely public. I've already got a wait
1151. 41:17list of a hundred, but basically I'm
1152. 41:18just in the process of building this out
1153. 41:20with like step-by-step. And it's
1154. 41:22everything that I do I'm just going to
1155. 41:23document into one place, but anyways,
1156. 41:25just throwing that out there as the last
1157. 41:27thing, but any
1158. 41:29I'd be curious on like, okay, you see
1159. 41:31you just watched this and you're in like
1160. 41:33a role at some company.
1161. 41:35>> [laughter]
1162. 41:35>> How do you how do you defend against
1163. 41:37this like with your job or do you is it
1164. 41:39just like you need to learn this now and
1165. 41:41like I I want to hear your thoughts
1166. 41:42because I you're seeing way more than I
1167. 41:44am with everything. Here, I'll tell you
1168. 41:46my I'll answer your question, but I want
1169. 41:48to start by saying like my biggest
1170. 41:50takeaway from all of this. Yeah.
1171. 41:52So, my biggest takeaway of all this is
1172. 41:54when you connected it with Railway.
1173. 41:57And I I it's a glimpse into the future
1174. 42:00of autonomous marketing.
1175. 42:03So, marketing, you know, all you
1176. 42:07basically what you've done like all
1177. 42:08those sort of
1178. 42:11jobs to be done were jobs to be done
1179. 42:13that were literally done by human
1180. 42:15beings, right?
1181. 42:16Um and then you kind of stitched
1182. 42:18together, you know,
1183. 42:20if you've ever run ad campaigns before,
1184. 42:23you know how painful some of these
1185. 42:24things are.
1186. 42:26Like both
1187. 42:26>> Just uploading the ads alone like that.
1188. 42:29I was like I have literally spent like
1189. 42:33I I mean I'm just imagining uploading a
1190. 42:34thousand ad variations.
1191. 42:37Like
1192. 42:37>> Dude.
1193. 42:39>> [laughter]
1194. 42:40>> Like it just Don't give me PTSD on the
1195. 42:41pod. Straight up.
1196. 42:44I I did this early in my career. It was
1197. 42:45absolutely painful. It was painful,
1198. 42:48right? It was painful, and and it's not
1199. 42:50fun. It's not fun at all.
1200. 42:52Uh
1201. 42:52removing and figuring out low performers
1202. 42:56ad creative not fun. Not fun, and you
1203. 42:58need to be on it. Um so, the idea that
1204. 43:01you can, you know, make this
1205. 43:04an agent that's working 24/7 and that's
1206. 43:07managing all these different things is
1207. 43:10the dream.
1208. 43:11It is absolutely the dream. Autonomous
1209. 43:13marketing the dream.
1210. 43:14I think
1211. 43:16uh
1212. 43:17who are the winners and who are the
1213. 43:18losers of this? The winners are going to
1214. 43:21be
1215. 43:22uh
1216. 43:23you know, one-person businesses, small
1217. 43:26teams,
1218. 43:28um
1219. 43:29and then maybe your your head of
1220. 43:32marketing that currently you're getting
1221. 43:34paid a hundred thousand dollars a year.
1222. 43:37Now, all of a sudden if you can figure
1223. 43:38out how to do all these things, and this
1224. 43:39is where I'm answering your question. If
1225. 43:41you can figure out how to do all these
1226. 43:42things, um you know, you could make the
1227. 43:45case like, hey, triple, you know, triple
1228. 43:46my salary. Easily. Right? Like from a
1229. 43:49value perspective like if you can do all
1230. 43:51these jobs to be done and you're one
1231. 43:53person instead of ten,
1232. 43:55there is a case to be made that you've
1233. 43:58made, you know, you've added a
1234. 43:59tremendous amount of value to to your
1235. 44:01role.
1236. 44:02So, I think
1237. 44:04uh
1238. 44:04and then the unfortunate thing is I
1239. 44:05think a lot of a lot of these jobs to be
1240. 44:08done, and this is where I disagree with
1241. 44:09a lot of people, is I think that there
1242. 44:11are is going to be a lot of job loss.
1243. 44:13Uh real job loss like
1244. 44:16it just
1245. 44:17who anyone who just wants
1246. 44:19>> to be extremely rapid job loss, and then
1247. 44:23like I'm just thinking about like the
1248. 44:24early days of like what we saw, you
1249. 44:26know, in the Industrial Revolution in
1250. 44:28like the United Kingdom. Like they
1251. 44:31I mean you basically have this
1252. 44:32displacement, and then new roles get
1253. 44:33created, but like in that interim still
1254. 44:36a lot of
1255. 44:38turmoil.
1256. 44:38>> It's going to be chaos. It's going to be
1257. 44:39chaos. And I like I have a friend who
1258. 44:42runs a startup, and he texts me
1259. 44:43yesterday, and he's like, I think I'm
1260. 44:44going to fire 50 people, and that's like
1261. 44:4670% of his team, right?
1262. 44:49And I'm just over here, and I'm like,
1263. 44:51how, why, what, you know, tell me the
1264. 44:53reason. He's like, I think I can
1265. 44:54automate all of their jobs right now
1266. 44:56with like agent swarms.
1267. 44:58And I'm like, okay, what's an agent
1268. 45:00swarm, you know, cuz that's just this
1269. 45:01like throw like term that gets thrown
1270. 45:03around right now. And he's like, oh,
1271. 45:06it's just an agent that does like a
1272. 45:07specific thing, and then there's an
1273. 45:09agent that manages like that whole
1274. 45:11system, and then like imagine like five
1275. 45:13pillars under like another agent. And
1276. 45:15I'm like, oh, I built that. That's what
1277. 45:18an agent swarm is.
1278. 45:19And I think that this is the thing that
1279. 45:22like people aren't really cuz now it
1280. 45:24just runs in the background. So, like I
1281. 45:25have one that's just like crawling
1282. 45:26LinkedIn like as we speak. And it's like
1283. 45:29looking for like ICP, and then it
1284. 45:31enriches them, it writes a personalized
1285. 45:33email, and it cold emails them.
1286. 45:34[laughter]
1287. 45:35Yeah. And like
1288. 45:38I don't think people understand like
1289. 45:40what's about to happen in like the next
1290. 45:4212 months. So, and I'm excited about it
1291. 45:45cuz I think there's like again, what if
1292. 45:47you can build like it is incredible like
1293. 45:51you're so capable right now, especially
1294. 45:53if you have domain knowledge is the
1295. 45:54other thing that I'm finding is like
1296. 45:56just because you can like it it can be
1297. 45:58built doesn't mean that you can build it
1298. 46:00because you don't have the vocabulary.
1299. 46:02Like when I look at like my co-founder
1300. 46:03Max, right? And his technical
1301. 46:05vocabulary, how he can describe the
1302. 46:07problem to a coding agent is so much
1303. 46:10more sophisticated than I'll ever be
1304. 46:11able to do it. And so, the output
1305. 46:13quality that he can get from this is at
1306. 46:16a like a level that's in the top 10 top
1307. 46:181%. So, if we translate that to
1308. 46:20something else. Like say you studied
1309. 46:21graphic design for 20 years and like
1310. 46:22you've been working in the industry for
1311. 46:2320 years. The vocabulary you have to
1312. 46:26describe something is going to be so
1313. 46:28much more sophisticated than what I
1314. 46:30have. So, I'm like this happened the
1315. 46:32other day where I'm like I wanted to put
1316. 46:33texture on the back of an ad, and I was
1317. 46:35like, how the do I do that in the
1318. 46:36background? I kept trying to describe
1319. 46:38it, it came out terrible. And then I
1320. 46:39found this like person giving a
1321. 46:41description of like how do you make it
1322. 46:43have a TV type texture, right? And it
1323. 46:46was like all of these you know, specific
1324. 46:48like it was like a specific lexicon to
1325. 46:50describe the the qual
1326. 46:52literally one shots it, you know,
1327. 46:53immediately like what I was looking for
1328. 46:55after that. And I had you have that
1329. 46:57realization that this actually becomes
1330. 46:59like the superpower if you can
1331. 47:01incorporate these tools into what you're
1332. 47:03doing for work and have that domain
1333. 47:05expertise, that knowledge that's like on
1334. 47:07top of that. That actually is what makes
1335. 47:09you like incredible. And so, it's the
1336. 47:11same thing that we've always seen where
1337. 47:12it's like, oh, you have one or two
1338. 47:14skills with like a deep T, and then you
1339. 47:15like that what's makes you valuable.
1340. 47:17This is like that, but like, you know,
1341. 47:18times a thousand where it's like if you
1342. 47:20have the vocabulary,
1343. 47:22you know, [laughter]
1344. 47:23and six things,
1345. 47:25and you come to these this tooling and
1346. 47:27can basically express and explain like
1347. 47:29what you're needing or what you're
1348. 47:30looking for, it changes the entire
1349. 47:32system in my mind, but I
1350. 47:34again, I'd love to hear your thoughts on
1351. 47:36like where you like and also just like
1352. 47:39what you're seeing within your own
1353. 47:40companies that you own and all like, you
1354. 47:41know, within the market as well.
1355. 47:43Well, I I I think the reality is there's
1356. 47:46a lot of people even if they have a ton
1357. 47:48of domain expertise,
1358. 47:50they don't know the tools and how to use
1359. 47:53the tools optimally yet. Now, I think
1360. 47:56that the tools are going to get so good
1361. 47:58that
1362. 47:59it the UX is going to be so easy at some
1363. 48:01point. But for right now, like for
1364. 48:03example, if we if we
1365. 48:06counted off all the tools that you
1366. 48:08mentioned in this in this podcast, you
1367. 48:10probably mentioned 17
1368. 48:13no, more, 20 tools. I'll include some of
1369. 48:15those in the show notes. Yeah, I'll give
1370. 48:17you the list of thems that you can pass
1371. 48:18them on.
1372. 48:19Phantom Buster instantly. Like I'm not
1373. 48:22even talking about like Claude code and
1374. 48:24stuff like that. You you like you've
1375. 48:26done your research, you found the tools,
1376. 48:28and you know, I think that's why a lot
1377. 48:30of people listen to this podcast. They
1378. 48:31try to, you know, it's a way to for them
1379. 48:32to learn a lot of that stuff, but the
1380. 48:34point is
1381. 48:35I have a lot of respect for the people
1382. 48:37that
1383. 48:38understand they have domain knowledge,
1384. 48:41know that that domain knowledge is super
1385. 48:43valuable, and who are going out out
1386. 48:45there and trying things.
1387. 48:47Um but yeah, this, you know,
1388. 48:50the other sort of
1389. 48:52takeaway I have from this podcast
1390. 48:55is
1391. 48:56your insight around APIs, which I
1392. 48:59thought was really interesting. So, like
1393. 49:02in the old way in the old way of like
1394. 49:04SaaS tools and stuff like that, you you
1395. 49:06didn't I mean it was nice an API was a
1396. 49:08nice to have, you know. It was about how
1397. 49:10good is the software that you're using.
1398. 49:13Um how good is the UX? How good is the
1399. 49:15brand? How good is, you know, when you
1400. 49:17press this feature, how quick is it? How
1401. 49:19instant is it?
1402. 49:20But now when you're living in a
1403. 49:22terminal, for example, and you're using
1404. 49:25MCPs to talk to LLMs,
1405. 49:28um you kind of, you know, the nice to
1406. 49:30have is actually the the the UI.
1407. 49:34The nice to have is the SaaS. The nice
1408. 49:35to have is going to this website and and
1409. 49:37look pretty. Ultimately, what you care
1410. 49:39about is the output and the thing is
1411. 49:41running the agents are running, you
1412. 49:43know, 24/7. They aren't, you know,
1413. 49:46hogging tokens. The output is high
1414. 49:48quality. It's doing the thing that it's
1415. 49:50says it, you know,
1416. 49:52uh it should do. And I think uh Sam
1417. 49:54Altman said recently something about
1418. 49:56APIs. I think he said
1419. 49:58that every company is going to be an API
1420. 50:01company.
1421. 50:02Totally I align with this. Like now
1422. 50:04doing this. Like I there's a software
1423. 50:07like I just won't put them on blast cuz
1424. 50:08I know how big your audience is, but
1425. 50:10like like there's a thing you can do in
1426. 50:12their UI I can't do in their API and I'm
1427. 50:14literally about to churn because I'm
1428. 50:16just like this is critical for me and
1429. 50:19now it feels archaic for me to go and
1430. 50:21interact with your like UI to do
1431. 50:24this output like output that I'm that
1432. 50:25I'm like I need. And I think that this
1433. 50:28is going to be
1434. 50:29like I think there's going to be
1435. 50:30companies that are entirely just like
1436. 50:32like what what it
1437. 50:33I mean we literally have this
1438. 50:34conversation internally. Like do we
1439. 50:36build an like a UI?
1440. 50:38Is that even a thing that we do? Or do
1441. 50:40we just build the tooling that enables
1442. 50:42you to see like where we see the puck
1443. 50:44going? And I think that like you know,
1444. 50:47we had to come, you know, have a
1445. 50:48come-to-Jesus moment of like, "Oh, this
1446. 50:50is uh
1447. 50:51>> [laughter]
1448. 50:52>> like where we know the puck is going is
1449. 50:54entirely different than like where the
1450. 50:56normie like is right now and like the
1451. 50:58adoption cycle of this." And so like
1452. 51:00having to like meet there to ride this.
1453. 51:01But like I mean it's very clear, right?
1454. 51:04Like for example, like with the Graft
1455. 51:06MCP. That's a live data feed. So I may
1456. 51:08I'm like when I'm like show me my Google
1457. 51:10Ads and Facebook Ads paid ad spend,
1458. 51:12right? That's a live data feed that's
1459. 51:14happening underneath the hood. It's like
1460. 51:15a it's like an endpoint that I'm
1461. 51:17hitting. It's literally on the fly
1462. 51:18generating a live data endpoint that I
1463. 51:20can pull from my data warehouse from.
1464. 51:22So with that, I can basically build
1465. 51:25whatever I want [laughter] on top of
1466. 51:27that.
1467. 51:28And I'm like, "Okay, well
1468. 51:30what what's just go build a custom
1469. 51:31dashboard like for what we need." But
1470. 51:34what we're finding is that like there's
1471. 51:35like kind of different use cases. Like
1472. 51:37the
1473. 51:38I guess how what I'm trying to say is
1474. 51:40like you're basically making your agent
1475. 51:43so that or whatever it is your tooling
1476. 51:45is so that it fits into any harness. So
1477. 51:47whether you're working from Claude iOS
1478. 51:49or chat GPT desktop or Claude code on,
1479. 51:52you know, in your terminal or like
1480. 51:54cursor or, you know, even the UI it's
1481. 51:57unified across that and people can
1482. 51:59basically take that wherever they want
1483. 52:00and get the same outputs. And so from a
1484. 52:03product standpoint, this is how we're
1485. 52:04like you know, focused on it and kind of
1486. 52:06moving forward. And I think it
1487. 52:09Again, what it comes down to though is
1488. 52:11like everything that like we're talking
1489. 52:12about this tooling piece like unless you
1490. 52:14know what to do
1491. 52:16it's very hard to get it to like if I if
1492. 52:18I didn't know this these exact tool sets
1493. 52:20to use to like go and do these actions.
1494. 52:22Like here's how to do this LinkedIn
1495. 52:24thing as an example, there would be like
1496. 52:26a very low chance that it would be able
1497. 52:28to figure this out. It's totally
1498. 52:29possible it can. It's just like going to
1499. 52:31take longer. And then you can do this
1500. 52:33now with like Claude and with Perplexity
1501. 52:35where you're like, "I'm trying to do X.
1502. 52:37List five APIs that can help me do
1503. 52:39that."
1504. 52:39>> [laughter]
1505. 52:39>> But I think this is how people are going
1506. 52:41to start building this. And like I'm
1507. 52:43finding myself doing this where I'm like
1508. 52:44I have this like vision of a workflow
1509. 52:47and I'm starting with the final product
1510. 52:48and then I'm like working back and like
1511. 52:51you know, basically piecing together how
1512. 52:53does this work and then having the agent
1513. 52:55go and build that for me. And so then
1514. 52:57this comes back to like how do agents
1515. 52:59discover the the necessary
1516. 53:01infrastructure
1517. 53:03for and like how do you be the thing
1518. 53:04that it picks when it's going to build,
1519. 53:06you know, XYZ thing? And like again,
1520. 53:09these are the things I'm losing sleep
1521. 53:10over now as we're like getting deeper
1522. 53:11and deeper into this.
1523. 53:13Well, Cody, I appreciate you being so
1524. 53:15saucy with with sharing all this stuff
1525. 53:17with us. We do appreciate it.
1526. 53:19I need you know, it's been too long.
1527. 53:21It's been too long. You need to come
1528. 53:22back on again
1529. 53:24and and uh share some ideas and stuff
1530. 53:27like that. So would love to have you.
1531. 53:29Um and uh
1532. 53:31Have a ton of them, man. We could go for
1533. 53:33days about Chrome extensions right now.
1534. 53:34You can literally have Claude code one
1535. 53:36shot them and just turn on Facebook Ads
1536. 53:38in the background automatically. I also
1537. 53:40had an agent that was running an Etsy
1538. 53:41shop for a little bit. That was crazy.
1539. 53:43It just got banned two days ago, which
1540. 53:45was hilarious. So
1541. 53:47So people, please, you know, let's beg
1542. 53:49Cody to come back on the pod and uh have
1543. 53:52him on soon. This is an open invite,
1544. 53:54Cody. You can come on whenever you'd
1545. 53:56like. Share ideas. You get you
1546. 53:58definitely get my creative juices
1547. 53:59flowing. So I appreciate you so much and
1548. 54:01>> Appreciated you. Thanks for for your
1549. 54:02time as always, man. I always love
1550. 54:04coming on.
1551. 54:05Thank you.