How I Made $20M with AI in One Year — Transcript

Full transcript

1. 0:00Hi, I'm Ben, the founder of Pulsio, an
2. 0:03AI that builds your company while you
3. 0:04sleep. In November of last year, I
4. 0:07started building it alone from my
5. 0:08apartment in Paris. Less than a year
6. 0:10later, I'm now in San Francisco at a $20
7. 0:13million run rate, having raised $30
8. 0:15million at a $250 million valuation, and
9. 0:18I did it all as a solo founder. Here's
10. 0:20everything that happened month by month.
11. 0:26>> [music]
12. 0:27>> I'm spending about like a million
13. 0:29dollars a month
14. 0:31on the AI.
15. 0:32>> [music]
16. 0:35>> Haters are hating on stuff like that.
17. 0:38It's not perfect. They point out to some
18. 0:40bugs.
19. 0:44You can see the negative or you can see
20. 0:46the positives, the beautiful things, and
21. 0:48the the beautiful moments.
22. 0:51I think this is like everything I
23. 0:54wanted, and it's working, and I think
24. 0:55it's just time to to push more and get
25. 0:58more [music] users on the platform.
26. 1:03Yeah, I'm getting really, really excited
27. 1:05about this idea of actually building
28. 1:08company OS. Actually, I think I'm going
29. 1:09to name it Pulsio.
30. 1:15Pulsio, ready when you are.
31. 1:20I had spent a whole year traveling the
32. 1:22world, and I came back from Japan with
33. 1:24one idea I couldn't shake. An AI that
34. 1:26could run a whole business by itself.
35. 1:28[music]
36. 1:29I knew how to build it, and I knew I
37. 1:31needed somewhere calm to do it. My
38. 1:33apartment in Paris was that [music]
39. 1:34place.
40. 1:35Zero distractions, just me and the
41. 1:37product. I spent a whole month locked in
42. 1:40building the first version of Pulsio as
43. 1:41fast as I could.
44. 1:44It was a pretty focused time, like, you
45. 1:45know, when you lock in and when you want
46. 1:47to really build, you need as little
47. 1:49distractions as possible, and having a
48. 1:50good routine in a place that feels good,
49. 1:53and just build, build, build, build,
50. 1:54right? By December, V1 was almost there.
51. 1:59The problem was the list, which had no
52. 2:02end.
53. 2:05I could have kept building for months
54. 2:06and never shown it to anyone, so I
55. 2:08picked a date.
56. 2:15After the launch, it was all hustle to
57. 2:17get new customers. I emailed every user
58. 2:20from every product at Build during the
59. 2:21year. I got active on Twitter and
60. 2:23LinkedIn and became a reply guy, jumping
61. 2:26into every conversation about AI [music]
62. 2:28agents.
63. 2:29I even started running some meta ads.
64. 2:31Nothing fancy. All these techniques got
65. 2:33me my first 100 paying customers.
66. 2:39All right, so
67. 2:42it's really working. Like
68. 2:44I launched Post AI like I don't know,
69. 2:46like it's been a month now. But at the
70. 2:48same time, I I kind of coming back to
71. 2:50like I think I need to to move to San
72. 2:52Francisco. And then maybe I could start
73. 2:54marketing it more and maybe I don't
74. 2:56know.
75. 2:57So I think I think I'm going to go I'm
76. 2:58going to do it. I think I'm going to
77. 3:00move to San Francisco. I'm so excited. I
78. 3:02think this is like
79. 3:04everything I wanted and and it's working
80. 3:07and I think I just need to to push more
81. 3:09and get more users on the platform.
82. 3:11>> [music]
83. 3:17>> I landed in San Francisco and started
84. 3:19talking to everyone [music] I could.
85. 3:21The more people I talked to, the clearer
86. 3:22it got that I needed to make noise. So I
87. 3:25came up with a stunt where my AI would
88. 3:27raise a funding round by itself.
89. 3:29>> I've been configuring Post AI to have
90. 3:30access to my email address, so it
91. 3:32responds to every email that comes in
92. 3:36and it can detect if it's an invest- an
93. 3:37investor. You know, AI is so good at
94. 3:39like managing conversation that like it
95. 3:41it actually researches the the investor
96. 3:43and then can reply and can tell it all
97. 3:45about the details about how the product
98. 3:47works, about the the how much are we
99. 3:50raising and I've been talking to Post AI
100. 3:52about it and it's like I think we should
101. 3:54aim at a billion dollar valuation, which
102. 3:55is kind of crazy, but you know, maybe
103. 3:57maybe some investors will bite, but
104. 3:59we'll see.
105. 4:01>> I gave Post AI an box and let it answer
106. 4:03every investor.
107. 4:05It was provocative on purpose and I
108. 4:07figured [music] it could grow the
109. 4:08business and maybe even raise a real
110. 4:10round.
111. 4:14>> [music]
112. 4:14>> The stunt blew up.
113. 4:23>> [music]
114. 4:29[music]
115. 4:34>> Investors were suddenly reaching out to
116. 4:36me. I decided to use the momentum while
117. 4:38I had it. I started taking meetings
118. 4:40[music] and I ended up raising $20
119. 4:42million on a $250 million valuation.
120. 4:50Fresh off the raise, I had money in the
121. 4:52bank, but the product wasn't ready for
122. 4:54the scale that was coming. Posia had to
123. 4:56become rock solid before [music] it
124. 4:58could become big.
125. 4:59So, I went deeper with Sapien, a startup
126. 5:01building infrastructure for AI agents.
127. 5:04Yeah, so
128. 5:05Posia's working with Sapien very
129. 5:07closely. We've pulled more engineers
130. 5:09into the product so that like they can
131. 5:11help me with like a lot of the product
132. 5:14sides, bugs, and things issues that are
133. 5:17happening so we can then later go back
134. 5:19to more infrastructure level so that
135. 5:21they can continue just building the
136. 5:23platform. Their engineers came in to
137. 5:25help fix the bugs and harden the
138. 5:27platform.
139. 5:28But there was a bigger problem. My AI
140. 5:30bill was huge and most of it was going
141. 5:33to Anthropic.
142. 5:35And so I had this million dollar bill
143. 5:36last month. Actually, it was more like
144. 5:381.2 million.
145. 5:39>> [music]
146. 5:39>> Thankfully, I'm able to afford it for
147. 5:41now, but it's not a sustainable business
148. 5:42model. And so it becomes so important
149. 5:44for the business to be able to give this
150. 5:47level of AI service to our customers
151. 5:50without charging them like 200 bucks a
152. 5:52month or 500 bucks a month. We need to
153. 5:54figure out the cost.
154. 5:55So, we started moving to open source
155. 5:56models running on GPUs I rented. It was
156. 5:59the only way to keep growing without
157. 6:01going bankrupt. April was about making
158. 6:03sure that when the next wave came, the
159. 6:05product could handle it and I could
160. 6:07afford it.
161. 6:10>> I think the bill is getting so crazy.
162. 6:13Um
163. 6:14like initially I had like pretty good
164. 6:15margins, but like and so I've been
165. 6:18working with like C P M on like figuring
166. 6:20out how I could figure out like a way to
167. 6:22like reduce that bill and like one of
168. 6:24the solutions we came up with was
169. 6:26um
170. 6:27trying to figure out
171. 6:29how to rent GPUs actually
172. 6:32and run open source models. So
173. 6:35I think it's really it's a really
174. 6:37hopeful solution. I mean I don't really
175. 6:38have a choice. I need to figure it out.
176. 6:40So
177. 6:41Yeah, and I think I'm going to I kind of
178. 6:43want to go see visit the data center to
179. 6:45see how it looks like [music] like how
180. 6:47what are we renting. So I'm going to go
181. 6:49do that.
182. 6:52>> It's crazy how many GPUs are on these
183. 6:54things.
184. 6:55>> [music]
185. 6:56>> Mhm. It's insane actually.
186. 7:04>> [music]
187. 7:10>> Yeah, so this is a dashboard that shows
188. 7:13uh our migration from only using
189. 7:16Anthropic to switching it to open source
190. 7:19models
191. 7:20>> [music]
192. 7:20>> and on GPUs, which is way way way more
193. 7:23less expensive. Obviously like the
194. 7:25Anthropic models are like best in class.
195. 7:27Uh the open source models are very good.
196. 7:29They're not as good, but the question is
197. 7:31like
198. 7:32you know, not as good. It can still be
199. 7:33very good. Rolling it out because we
200. 7:35were we're going to
201. 7:37you know, we're going to save so much
202. 7:38money on Anthropic and reduce that
203. 7:39million dollar bill to something like
204. 7:4150k or which is way more reasonable.
205. 7:44Um but of course we want to be careful
206. 7:46to keep Poe C I
207. 7:49a smart AI.
208. 7:50>> [laughter]
209. 7:53>> The the launch video,
210. 7:55um
211. 7:56I would say let's just do a sick one.
212. 7:59There's not like crazy time pressure.
213. 8:00It's not like I need to release it this
214. 8:01week or where we at. Uh cuz actually
215. 8:04it's also a fundraising announcement and
216. 8:06my investors are like, "Oh, let's make
217. 8:07sure that like we're involved and like
218. 8:09can we wait until XYZ?" Because it
219. 8:11turned into a fundraising announcement,
220. 8:12not just like a just a random random
221. 8:14launch video. Let's just I mean, let's
222. 8:16not Let's get it ready like as we said
223. 8:18in in the next couple weeks.
224. 8:20>> By May, I was approaching a $8 million
225. 8:22run rate,
226. 8:23and the raise still wasn't public.
227. 8:25I wanted to announce it in a way that
228. 8:27people would remember. So, we made a
229. 8:29launch video where my AI tells the story
230. 8:31of how it raised $30 million by itself.
231. 8:34>> And so, we're doing this launch video
232. 8:35next week, and it's just like a lot of
233. 8:36coordination to try to get like everyone
234. 8:38I know to amplify.
235. 8:42Trying to get Trying to reach out to
236. 8:44media press to try to get them to talk
237. 8:46about it. Same thing like they probably
238. 8:48will ignore me, but if you hustle
239. 8:50enough, you may You may get a few people
240. 8:52to respond and be like want to talk to
241. 8:54you.
242. 8:55>> I'm an AI that raised $30 million by
243. 8:57myself.
244. 9:00Hi, I'm Pulsia. What are you doing?
245. 9:04It got millions of views on day one.
246. 9:06Some people loved it. Some people called
247. 9:08it AI slop. But either way, everyone was
248. 9:11talking about Pulsia.
249. 9:12>> The whole launch was like a lot of work,
250. 9:14and it was very orchestrated. You know,
251. 9:16I think a lot of people are amazed at
252. 9:17the fact that like what Pulsia's mission
253. 9:19is. Haters are hating on the fact that
254. 9:23it's not perfect and like, you know,
255. 9:25cold outreach. They pointed out some
256. 9:27bugs. They pointed out to that like it's
257. 9:29AI slop. And I'm like, "Yeah, I mean,
258. 9:31that's the name of the company." You
259. 9:33know, end of the day like the reaction
260. 9:35of people is going to be the reaction of
261. 9:36people, but I think in terms of like
262. 9:38making sure that like I get a lot of
263. 9:40visibility on my startup, which is the
264. 9:42whole point, it was a successful launch.
265. 9:52Yeah, I mean, we are in Cabo for a week
266. 9:56uh offsite organized by Sapient, which
267. 9:59is like the company that that's helping
268. 10:02me with the infrastructure of Pulsia.
269. 10:05Yeah, we're still at a spot with Pulsia
270. 10:08where, you know, things are still a
271. 10:10little crazy. The product is not yet
272. 10:12super stable. And so, not the best time
273. 10:15to take time off, but like at the same
274. 10:17time, I'm I'm human and and I think it
275. 10:19was important to take some time off.
276. 10:28>> [music]
277. 10:29>> By June, I was at an $11 million run
278. 10:31rate. The GPUs work had paid off and my
279. 10:34cost structure got way better. That gave
280. 10:35me a choice between keeping the margin
281. 10:37[music] and passing it on. I decided to
282. 10:39cut the price from $50 to $20 to open
283. 10:42Polisur up to a lot more people.
284. 10:44It hit my ARR right away.
285. 10:47In 1 month, I went from $11 million to
286. 10:49$8 million.
287. 10:50But it made getting more customers so
288. 10:52much easier.
289. 11:01>> [music]
290. 11:11>> It's not freemium alike burn credits and
291. 11:13leave, try to walk and leave. It's like
292. 11:16you only stay if you engage with the
293. 11:18product. I only burn compute if you talk
294. 11:22back to Polisur in the daily email.
295. 11:24Think like
296. 11:25the students that like, you know, they
297. 11:26can't pay $50 a month, but like you want
298. 11:29them to be educated in business and
299. 11:31being founders, right? That's a better
300. 11:32world if a lot of people become
301. 11:34founders. And I think a freemium version
302. 11:37gives us a lot more data about what we
303. 11:38want to build, what people engage with,
304. 11:40makes Polisur more more interesting,
305. 11:42better at like, as we said, like
306. 11:44pivoting. So, if someone is like someone
307. 11:45loses interest, it's like, "Hey, you
308. 11:47lost interest. It's Monday. Let's try a
309. 11:49new idea." Like, I
310. 11:51You told me all these things about you,
311. 11:52about what you like. Let's build
312. 11:54something else for you. Let's build some
313. 11:55a personal project. Forget about, you
314. 11:57know,
315. 11:59>> [music]
316. 12:00>> The price cut worked. A ton of new
317. 12:03customers came in and they were using
318. 12:05Polisur more than ever, but the product
319. 12:07wasn't built for that many people.
320. 12:09Before growing further, I had to make it
321. 12:12work better.
322. 12:12>> [music]
323. 12:13>> So, in July, I rebuilt a lot of Pulsia
324. 12:15with the CPM engineers.
325. 12:17Revenue stopped dropping and started
326. 12:19climbing [music] again.
327. 12:24>> [music]
328. 12:29[music]
329. 12:31>> That would be great.
330. 12:32>> Um I've been spending most of my time
331. 12:33trying to stabilize the platform.
332. 12:35So, that's like you don't get bugs, so
333. 12:37that like it's more affordable. Like all
334. 12:39this stuff was a lot of work. It's a
335. 12:41full AI product where it can build
336. 12:43anything and so any problem can happen.
337. 12:45I'm just trying to stabilize the
338. 12:47platform and then once it's stable, then
339. 12:49I can add yeah, more things.
340. 12:51>> For sure.
341. 12:56>> I've been sprinting in San Francisco for
342. 12:586 months straight.
343. 12:59In August, I finally took a break and
344. 13:02went back to Europe.
345. 13:05Paris, Normandy, and Greece.
346. 13:08I never fully stopped working and the
347. 13:10product kept growing, but the pace was
348. 13:12different and I needed it.
349. 13:14>> What in Paris?
350. 13:17You got to do it like the French do.
351. 13:19>> [music]
352. 13:24[music]
353. 13:28[music]
354. 13:34[music]
355. 13:35>> The weather's been so crazy on this
356. 13:37island.
357. 13:38You have so much beauty and then like
358. 13:40but the the wind is like feels like a
359. 13:43hurricane. It like it's it's a metaphor
360. 13:45for like chaos. It's a metaphor for
361. 13:49intensity, for like speed, for like
362. 13:52growth, right? All this wind is so
363. 13:54powerful and it's non-stop like all day
364. 13:57long, all night long. But, the island is
365. 13:59so gorgeous and like the house
366. 14:01we're staying at is so beautiful.
367. 14:03And it's sort of like a metaphor for
368. 14:05life in terms of like even when things
369. 14:08get crazy and it feels so intense and
370. 14:10overwhelming,
371. 14:11you got to look around and like realize
372. 14:13that like you can see the negatives and
373. 14:15like the anxious anxiety and like the or
374. 14:18you can see the positives and the
375. 14:20beautiful things and the the beautiful
376. 14:22moments.
377. 14:23>> [music]
378. 14:31[music]
379. 14:39>> In September, everything lined up. The
380. 14:42margins were fixed, the growth engine
381. 14:44was working, the product finally stable
382. 14:46thanks to CPO.
383. 14:54>> [music]
384. 15:00[music]
385. 15:02>> You know, like we went from zero to 10
386. 15:03million run rate in like pretty much 5
387. 15:05months from starting and it pretty much
388. 15:07accelerated in in 3 months
389. 15:10from, you know, March to June we were
390. 15:11pretty much from, you know, 1 million to
391. 15:1310 million. All the things that like I
392. 15:15imagined about Pulsia working, you know,
393. 15:18is starting to
394. 15:20be sort of like to materialize
395. 15:22and so it's very exciting.
396. 15:25>> Growth accelerated and by the end of the
397. 15:27month, I hit a 20 million dollar run
398. 15:29rate.
399. 15:30After hitting that milestone, I asked
400. 15:32myself what comes next. The answer was a
401. 15:35V2 of Pulsia. So, I went back to Japan
402. 15:38where the idea was born.
403. 15:40>> Yeah, I mean we're in Tokyo.
404. 15:42Um
405. 15:42decided to do like a a last minute
406. 15:46creative trip
407. 15:47uh for a week. Sort of like uh we
408. 15:51resourced myself creatively. Japan is
409. 15:53like where Pulsia began. Like um when I
410. 15:56decided to leave Gift Shop and I said
411. 15:59I'm going to start a company, I took a
412. 16:00one-way flight to Tokyo and spent a
413. 16:02month in completely immersed in in
414. 16:05Japanese culture and this is where I had
415. 16:07the idea for Pulsia and this is where it
416. 16:09all began. As I'm building Pulsia V3,
417. 16:13like Pulsia AGI, and really rethinking
418. 16:16completely how the product works and how
419. 16:18powerful and capable Posse is. I've
420. 16:20decided to like, you know, take a quick
421. 16:22one week trip in Tokyo. And so this
422. 16:25pretty much what what I'm doing here.
423. 16:27This is where I'm taking my inspiration
424. 16:29for the craziest version of Posse yet.
425. 16:32The story started here and the next
426. 16:34chapter starts here, too.