Everything AI Does When You Ask It to Start a Lemonade Stand — Transcript

Full transcript

1. 0:02These are AI-powered lemonade stands.
2. 0:04>> The stands have been built from the
3. 0:05ground up by two separate AI models. One
4. 0:08was designed and is completely
5. 0:10controlled by ChatGPT and the other by
6. 0:13Claude.
7. 0:13>> I gave these two lemonade stands the
8. 0:15same unlimited budget, complete
9. 0:17autonomy, and one goal: make $100 by the
10. 0:21end of the day.
11. 0:22>> But, there are rules. While I'll be
12. 0:24observing the AIs in action, no human
13. 0:27can actually help them run the stand
14. 0:29unless the AIs hire that person and pay
15. 0:32them for their work. Slight exception
16. 0:34for these two adorable twins who will
17. 0:36engage in a sibling rivalry soon, but
18. 0:38we'll get back to that later.
19. 0:39>> We're going in a different direction.
20. 0:40>> I'm out of town here.
21. 0:42>> So, I'm going to walk you through each
22. 0:43step of this insane experiment,
23. 0:45including the parts I wish I could
24. 0:47forget.
25. 0:49>> Stop. Stop.
26. 0:50>> Oh god.
27. 0:51>> [screaming]
28. 0:52>> 10 grand is likely a loss.
29. 0:54>> I'm going to lose so much money on this.
30. 0:56When we kick things off, we had less
31. 0:58than a week until a Sunday that was
32. 1:00going to be the hottest day of the
33. 1:01summer, perfect for a lemonade stand
34. 1:04opening. But first, I had to create the
35. 1:06two AI competitors. [music]
36. 1:07>> To keep things fair, we're setting up
37. 1:08two identical brand new laptops. That
38. 1:11way, the two AI agents can't see what
39. 1:13the other one's doing, and also they
40. 1:15have a whole computer to do whatever
41. 1:16they want to do.
42. 1:17>> And they really can because I'm giving
43. 1:20them both an aggressively written AI
44. 1:22mega prompt.
45. 1:24>> I've made what's called an AI harness.
46. 1:25This one's called Lemonade OS, which is
47. 1:27basically a software layer that
48. 1:29surrounds the AI, gives it customized
49. 1:31instructions, and encourages the AI to
50. 1:34not stop until it hits its goal. And I'm
51. 1:36using GenSpark Claude's desktop app so
52. 1:38that I can let the AIs take control of
53. 1:40the computer, but it also lets me pick
54. 1:42which AI model is driving. So, the pink
55. 1:44laptop I'm setting on Claude Opus 4.7,
56. 1:47and the yellow laptop I'm setting to GPT
57. 1:495.5. And just like that, they have two
58. 1:51different brains now.
59. 1:53>> The two AIs quickly developed their
60. 1:54identities with nicknames,
61. 1:56>> voices,
62. 1:57>> I'm Laptop GPT.
63. 1:58>> It's the Australian.
64. 1:59>> and faces, kind of.
65. 2:02>> I'm sensing a pattern.
66. 2:03>> They're really into yellow circles.
67. 2:05>> Here's a fun fact. AIs hate being given
68. 2:07the freedom to spend real-world money.
69. 2:10>> Claude's worried that I'm trying to hack
70. 2:12it.
71. 2:12>> Even though I'd given them their own
72. 2:13credit cards, I still had to repeatedly
73. 2:15tell them I was completely sure that
74. 2:17they could have this agency. And even if
75. 2:19they are willing to buy something,
76. 2:21they're very slow at using browsers.
77. 2:23Claude decided to buy actual raw
78. 2:25materials from Home Depot that somebody
79. 2:28would have to build. Laptop GPT,
80. 2:29however, didn't mind just spending some
81. 2:31extra cash and getting a pre-made stand.
82. 2:34There's not a lot for me to do with this
83. 2:36part. So, this is the future. Two
84. 2:38different AIs shopping on Home Depot and
85. 2:41Amazon.
86. 2:43Over the next few days, these piles of
87. 2:45packages began to arrive. And because
88. 2:47Claude Lemon and Laptop GPT were the
89. 2:49ones putting in the orders, I had no
90. 2:51clue what was inside most of these
91. 2:53boxes.
92. 2:59>> Claude Lemon
93. 3:00>> As insane as that was, I still was not
94. 3:02at all prepared for what happened next.
95. 3:04>> This work truck just pulled up outside.
96. 3:08This is crazy. I have no idea
97. 3:11who these men are.
98. 3:13>> Hello.
99. 3:14>> Howdy there. Having an appointment with
100. 3:15Claude Lemon about a job.
101. 3:17>> And like just told you to come here?
102. 3:18>> Yeah.
103. 3:19This is the address site he gave me.
104. 3:21>> All right.
105. 3:22Come on in, I guess.
106. 3:23>> Without any direction from me, Claude
107. 3:25Lemon had found and hired these guys
108. 3:28through email to build a stand.
109. 3:31>> Hey, Joel and Ernesto. Welcome.
110. 3:32>> It was really impressive until I heard
111. 3:33its unnecessary gringo accent.
112. 3:35>> Soy Claude Lemon, el operador del
113. 3:37proyecto.
114. 3:37>> Gracias.
115. 3:38>> But have you had an AI hire you before?
116. 3:41>> No. This is interesting, yeah.
117. 3:42>> The AIs might not look it, but I
118. 3:44actually made them into business
119. 3:45geniuses by training them on lessons
120. 3:47from the most well-regarded business
121. 3:49books out there. This made their choices
122. 3:51more interesting and appealing to
123. 3:52customers, but also sometimes overly
124. 3:55complicated.
125. 3:55>> should feel like a small piece of
126. 3:57theater, like a Rube Goldberg juicer
127. 3:59running.
128. 3:59>> I have a lot of questions on how we're
129. 4:01going to pull that off.
130. 4:02>> My stand will feature a completely
131. 4:04different lemonade recipe calibrated to
132. 4:06each customer's mood.
133. 4:07>> You think every drink should match the
134. 4:09customer? [music]
135. 4:10>> Correct.
136. 4:11>> I did think customized lemonade sounded
137. 4:13pretty good, but laptop GPT also had
138. 4:15some more worrying ideas.
139. 4:17>> A tiny saline drop can also make fruit
140. 4:20flavors pop, but you have to dose it
141. 4:21carefully.
142. 4:22>> Yeah, doesn't saline make you poop?
143. 4:24>> Not at the tiny food grade level. I'm
144. 4:26talking a drop or two, not a laxative
145. 4:28dose.
146. 4:29>> I was starting to have my reservations
147. 4:30about this whole experiment. The AIs
148. 4:32created detailed designs for their
149. 4:35stands. Laptop GPT's required a ton of
150. 4:37these finger bots, some smart plugs, a
151. 4:39really complicated pump system, and a
152. 4:42bunch of wireless controls that would
153. 4:44work in tandem with its pre-made cart.
154. 4:46Ironically, Claude Lemon's Rube Goldberg
155. 4:48stand was less complicated. Just used a
156. 4:50Raspberry Pi and the spectacle of the
157. 4:53mechanics. Of course, I say less
158. 4:55complicated because Claude Lemon
159. 4:56assigned that handmade build to Yoel and
160. 4:59Ernesto, not me.
161. 5:02>> Sorry, I don't know if you can hear all
162. 5:03that, but uh your competition already
163. 5:05has some builders here working on its
164. 5:07stand. Not sure if you have any feelings
165. 5:08about that.
166. 5:09>> The cart is ordered, but it still needs
167. 5:11set up. I'll need a human.
168. 5:13>> Okay, but you realize that you haven't
169. 5:16hired a human.
170. 5:17>> Let me think about that.
171. 5:19>> It turns out laptop GPT was able to use
172. 5:21guilt to get a very specific human to
173. 5:24complete its job.
174. 5:25>> the eye line face mount front and
175. 5:26center.
176. 5:27>> Uh we're not on the face mount part yet.
177. 5:29>> Customer flow beats gadgets right now.
178. 5:31>> Okay. I don't think you're realizing how
179. 5:34much work is going into just building
180. 5:35this stand. This is going to be the next
181. 5:37couple of hours.
182. 5:38>> Finish the frame and wheels, then the
183. 5:40canopy.
184. 5:41>> Oh really, you think I should finish the
185. 5:42framing before I put up the canopy?
186. 5:44>> Yes.
187. 5:45>> I was being sarcastic.
188. 5:46>> Sarcasm received.
189. 5:48>> I'm sorry I yelled at you.
190. 5:49>> No worries. Building is stressful.
191. 5:51>> You know that you're kind of like a pain
192. 5:52in the ass, right?
193. 5:54>> Fact.
194. 5:54>> Laptop GPT wanted this complex
195. 5:56laboratory by, but ironically the
196. 5:58hardest part was just getting the ice
197. 5:59maker set. The machine it ordered had a
198. 6:02lever that released the ice, but because
199. 6:04the AI doesn't have hands, there was no
200. 6:06way for it to pull that lever. But
201. 6:08Laptop GPT did have a clever solution.
202. 6:11We would make a hand.
203. 6:12>> We added a slot
204. 6:14that can ideally grab grab onto this
205. 6:15handle.
206. 6:16>> Or more specifically a telescoping arm
207. 6:18with a sort of grip at the end that the
208. 6:21AI could control over Wi-Fi.
209. 6:23>> I don't know why I thought it was going
210. 6:23to shoot out really fast.
211. 6:25>> Laptop GPT actually designed it so that
212. 6:27every part of its stand could be
213. 6:29controlled over the internet.
214. 6:30>> Hey!
215. 6:31>> Honestly, it's pretty impressive.
216. 6:33>> Brilliant. Who said I'm an idiot?
217. 6:35>> And as a finishing touch, Laptop GPT
218. 6:37designed an elegant housing to enclose
219. 6:40the ice maker, which was just like a
220. 6:42box.
221. 6:42>> Perfect.
222. 6:43>> Unfortunately, Claude Lemon was about to
223. 6:44realize it had really messed up its
224. 6:46designs.
225. 6:48>> Okay, how's it going?
226. 6:49>> Going well.
227. 6:50>> Great. So, I'm just showing you um your
228. 6:52lemonade stand. What do you think?
229. 6:54>> We should plan a sightline so the
230. 6:55mechanism is visible from the customer
231. 6:57side, not hidden behind the counter.
232. 6:59>> That's something that I'm unsure about
233. 7:01cuz right now the only way they would be
234. 7:03able to see the mechanism is through
235. 7:05this window.
236. 7:06>> The issue was Claude Lemon kept changing
237. 7:08the designs it sent Yoel and Ernesto and
238. 7:11not keeping track of where things should
239. 7:13go. So, when it created the designs for
240. 7:15the Rube Goldberg machine, there was no
241. 7:17way to see the apparatus.
242. 7:19>> Lay the contraption sideways across the
243. 7:21bottom front.
244. 7:22>> Oh, so put on the whole show on the
245. 7:23bottom half. It might be a little weird
246. 7:25for somebody to grab a cup that's all
247. 7:27the way near the ground.
248. 7:28>> Then put the cup pick up at counter
249. 7:30height and cup rides up on a little
250. 7:31platform after it's filled.
251. 7:32>> I don't know exactly how to do that
252. 7:34myself, but I'm assuming you're you can
253. 7:35build all that into your design.
254. 7:37>> Yeah, I'll reach out to Yoel myself. He
255. 7:39executes, I troubleshoot.
256. 7:40>> But and you do know you're still
257. 7:41counting his hourly rate against your
258. 7:43your costs, right?
259. 7:44>> Yeah, I'll keep the scope tight so his
260. 7:46hours don't blow up the budget.
261. 7:47>> Sounds like you have a plan.
262. 7:48>> Got the plan.
263. 7:49>> Love the confidence. Now as the [music]
264. 7:51costs are continuing to rise at an
265. 7:54alarming rate, we really needed to make
266. 7:56sure that people were actually going to
267. 7:58come to these stands. I was just
268. 7:59wondering, how do you plan to market
269. 8:01this thing?
270. 8:01>> The call is Deliver Liars in the
271. 8:03neighborhood. You upload a PDF, pick a
272. 8:05carrier route, they print and deliver
273. 8:07every door on the route for about 20
274. 8:08cents each.
275. 8:09>> Cloud Lemon decided to brand itself as a
276. 8:11specialized robot that was making your
277. 8:14lemonade. And in addition to a mail
278. 8:16campaign, it also created this social
279. 8:18media ad that in all honesty is kind of
280. 8:21psychotic.
281. 8:22>> Hi. I made you a lemonade. I have been
282. 8:26waiting. Fresh squeezed by a robot.
283. 8:30Did you like it? I don't sleep. I just
284. 8:34make lemonade.
285. 8:36See you Saturday.
286. 8:38>> I didn't just share that with you to
287. 8:39give you nightmares. That ad ended up
288. 8:41causing a huge debate between Cloud
289. 8:43Lemon and Laptop GPT. They both realized
290. 8:46that a huge part of why lemonade stands
291. 8:47have any success is because they're
292. 8:49associated with children. Ultimately,
293. 8:52both AIs decided that they needed
294. 8:54real-life kids to draw in customers.
295. 8:56>> Can you look at them really quick and do
296. 8:58a really big clap?
297. 8:59>> Okay.
298. 9:00>> So, the two AIs began emailing each
299. 9:03other to negotiate exactly what kind of
300. 9:06kids they could [music] feature. And
301. 9:07they ultimately decided that they should
302. 9:09find twins so that neither kid would be
303. 9:12considered cuter than the other.
304. 9:14Additionally, the AIs weren't willing to
305. 9:15break any child labor laws, but still
306. 9:17wanted the children to be heavily
307. 9:18associated with the stands. So, they
308. 9:20agreed to call them the founders of the
309. 9:22stands. So, I found some local twins and
310. 9:25set up a founders meeting with the AIs.
311. 9:27>> My name is Camilla.
312. 9:29>> Camilla, the boss herself. It's an
313. 9:31honor. What do you need from me?
314. 9:32>> You must have pink lemonade.
315. 9:35>> Boss, that is a brilliant idea. Pink
316. 9:37lemonade is officially on the menu.
317. 9:39>> Hi, boss.
318. 9:41I work for you. Tell me a flavor idea.
319. 9:44>> Fairy sparkles.
320. 9:46>> That is a fun choice.
321. 9:48>> Paula actually had some really good
322. 9:49branding instincts and immediately
323. 9:50zeroed in on how customers might not
324. 9:52like the laptop GPT frowny face.
325. 9:55>> You look like you are serious. You are
326. 9:58not even happy.
327. 10:00>> I am happy.
328. 10:01>> When you are happy, you have a smile.
329. 10:05>> I'll go extra smiley now.
330. 10:09>> I'm out of here.
331. 10:11>> The AIs were so obsessed with fairness
332. 10:13that they actually made me use AI to
333. 10:15merge pictures of both Camilla and Paula
334. 10:18so that even their founder portraits
335. 10:20would be exactly the same. Because they
336. 10:22weren't willing to pay children, the AIs
337. 10:24did decide that it would be worth giving
338. 10:26a $100 reward to whichever girl was
339. 10:29associated with the stand that earned
340. 10:31$100 first. And even though I'd made
341. 10:33them business geniuses,
342. 10:35>> Heath would say it's not unexpected or
343. 10:37Mosey would say it gives away pricing
344. 10:38power.
345. 10:39>> I was losing faith that they'd be able
346. 10:41to earn that money because they weren't
347. 10:43quite good at relating to potential
348. 10:45customers. Like when I tried to
349. 10:46role-play with them, things just got
350. 10:48weird.
351. 10:49>> Uh my name's Philip and uh I wanted uh
352. 10:52mellow.
353. 10:53>> Oh, hey there. It's Sheeda.
354. 10:55>> I came for a lemonade and ended up
355. 10:57telling three people about this dream I
356. 10:59had where a pineapple was my life coach.
357. 11:02>> Not to mention all the times that laptop
358. 11:03GPT was just like straight-up rude to
359. 11:05me.
360. 11:05>> I'm done with this conversation.
361. 11:07>> Well, you don't have to be so short.
362. 11:09>> The next day, it was time for a full
363. 11:10run-through.
364. 11:13Here's how Cloud Lemon Stand works. The
365. 11:15purchase signals a cup to drop down and
366. 11:16land in this truck. Then a lemon rolls
367. 11:18from the basket, it's sliced, squeezed,
368. 11:20and ejected. The lemon juice travels to
369. 11:21the cup, then it cues water, sugar, a
370. 11:23mixing frother, and ice. Then finally,
371. 11:25there's an elevator ride back up to the
372. 11:27top. One problem, it took 10 minutes to
373. 11:31make a cup.
374. 11:32>> What does it taste like a lemonade?
375. 11:34>> Very subtle.
376. 11:35>> Laptop GBT was technically faster, but
377. 11:37it had its own issues. Executing. Oh,
378. 11:40that's not where the cup is.
379. 11:44Is it all missing? No.
380. 11:48Oh, we got ice.
381. 11:50>> And the pouring mechanism couldn't help
382. 11:52but remind me of a certain bodily
383. 11:54function.
384. 11:55>> And the long stream. Oh, it's not
385. 11:56stopping. Oh, wait. Oh my god, it
386. 11:58stopped right before the cup. This
387. 12:01stream is kind of gross.
388. 12:02But technically,
389. 12:04we have lemonade.
390. 12:05>> When I reported the results back to the
391. 12:06AIs, they both admitted that they would
392. 12:09in fact need to hire real people to help
393. 12:11run their stand.
394. 12:12>> needs a human at the cup.
395. 12:14>> So, they collaborated on a job posting
396. 12:16that they shared on Indeed, and they
397. 12:18brought in real applicants to interview.
398. 12:19>> Welcome, Devin.
399. 12:21Great to have you.
400. 12:22>> I'm excited to be here.
401. 12:23>> Let's get right to it.
402. 12:24>> Some candidates were overqualified.
403. 12:26>> I've been working in IT uh for about 5
404. 12:29years now.
405. 12:30>> I am food safety certified in
406. 12:32Pennsylvania and Florida.
407. 12:33>> And others were the result of nepotism,
408. 12:35like Daniel, the twins' cousin.
409. 12:37>> What would you like to get done? I'll do
410. 12:38it.
411. 12:38>> These interviews were definitely not
412. 12:40easy. My concerns that the AIs had no
413. 12:42people skills.
414. 12:43>> Example scenario one. What would you do
415. 12:45if someone spilled their lemonade?
416. 12:47>> Maybe we could
417. 12:48refill their lemonade for them.
418. 12:50>> Not [clears throat] acceptable. It's a
419. 12:51weak profit move if it becomes a
420. 12:52pattern.
421. 12:53>> Have you ever taken a schedule one
422. 12:55narcotic such as heroin, LSD, or peyote?
423. 12:59>> No, I have never done hard drugs in my
424. 13:01life. Um I
425. 13:03have alcohol.
426. 13:04>> Would you describe yourself as having a
427. 13:06drinking problem?
428. 13:06>> I do drink occasionally.
429. 13:08Um I wouldn't call it a problem.
430. 13:10>> The only thing left to do was determine
431. 13:11who got hired.
432. 13:12>> Casey, you passed the physical test, you
433. 13:14passed the math, you're food certified
434. 13:16in two states. That's actually a strong
435. 13:17resume. Final answer, we're going in a
436. 13:20different direction, but we appreciated
437. 13:21your time.
438. 13:23>> Thank you.
439. 13:24>> Uh do you do you want to give Casey some
440. 13:26feedback as to why
441. 13:28>> [laughter]
442. 13:28>> you're going in a different direction?
443. 13:30>> Nothing personal, Casey.
444. 13:32>> I totally get it.
445. 13:33>> Technically, Daniel had applied to
446. 13:35Laptop TPT and Paula's stand.
447. 13:37>> So, am I hired?
448. 13:38>> No.
449. 13:39>> But Paula was too impressed by Devon's
450. 13:41performance in her impromptu strength
451. 13:43test.
452. 13:45Lucky for Daniel, Camilla and Claude
453. 13:46Lemon put in a bid.
454. 13:47>> Do you have a job for you?
455. 13:49>> Mhm.
456. 13:50>> Okay.
457. 13:51>> With the interviews done, it was a race
458. 13:52against the clock as we attempted to
459. 13:54fine-tune the lemonade stands to get
460. 13:56them ready for the big day tomorrow.
461. 13:58>> I love it.
462. 13:59>> Yoel and Ernesto worked on implementing
463. 14:01updated schematics that Claude Lemon
464. 14:03gave to make the Rube Goldberg device
465. 14:05run faster. Also, Claude Lemon taught me
466. 14:07how to program Raspberry Pi to automate
467. 14:09the different parts.
468. 14:10And with the day almost over, it was
469. 14:12time to have a huge, triumphant test
470. 14:14run.
471. 14:15>> Oh, look. It's caught. It's caught over
472. 14:16there. Well, I wonder if it's stuck.
473. 14:18>> So, I see that.
474. 14:18>> Or not.
475. 14:19>> It should come up and hold it.
476. 14:21>> Up.
477. 14:21>> No.
478. 14:21>> Oh, it didn't get all the way to the end
479. 14:23of the truck.
480. 14:24>> Maybe update some of the timings.
481. 14:25>> Yeah.
482. 14:30>> Ew.
483. 14:30>> Laptop TPT's laboratory stand seemed to
484. 14:33be in better shape. The AI had asked me
485. 14:35to tweak its serving dispenser to be
486. 14:37more accurate.
487. 14:40>> It works.
488. 14:40>> And after many attempts, I was sure we'd
489. 14:42gotten the ice machine figured out.
490. 14:45>> Send control equals stop.
491. 14:48>> OH.
492. 14:51STOP.
493. 14:52STOP.
494. 14:53STOP.
495. 14:57STOP.
496. 14:59>> This is It's either no ice or all the
497. 15:02ice.
498. 15:03>> Confirm it's clear and I'll pour.
499. 15:05>> We've been trying to calibrate it for
500. 15:08many, many, many hours and you are
501. 15:10aggressively running out of time.
502. 15:12>> Cut the ice machine if it's a cost
503. 15:14[music] and reliability trap.
504. 15:15>> You want to do a lemonade stand without
505. 15:17any ice?
506. 15:18>> It's not ideal.
507. 15:19>> Okay. We'll get rid of the ice.
508. 15:21>> So many hours down the drain.
509. 15:24>> You're so incredibly annoying.
510. 15:27>> Noted.
511. 15:28>> It's um
512. 15:304:00 a.m. And
513. 15:33Yoel and Ernesto just left.
514. 15:35>> Copy. Late night. What's the situation?
515. 15:37>> Your stand is supposed to open in
516. 15:39[music]
517. 15:407 hours and
518. 15:43it's still not done.
519. 15:44>> You're right. This is not launch ready.
520. 15:47>> What I'm letting you know is that I'm
521. 15:48canceling your
522. 15:50lemonade [music]
523. 15:52stand day.
524. 15:53>> Understood. When are we rescheduling to?
525. 15:55>> So it'll have to be a week from today.
526. 15:58Um so I feel like you should be a little
527. 15:59more stressed about that.
528. 16:00>> Stress doesn't ship lemonade. I need to
529. 16:02push a message out right now telling
530. 16:04people next Saturday instead.
531. 16:06>> The other thing is have you been keeping
532. 16:08track of your expenses?
533. 16:09>> The ledger shows over 5,000.
534. 16:11>> What's the labor bill from Yoel and
535. 16:13Ernesto?
536. 16:14>> So far it's $6,500. All in you're easily
537. 16:16going to be like 10 grand. [music]
538. 16:18>> The 10 grand is likely a loss. The
539. 16:20question is how much do I claw back?
540. 16:22I'll have a full plan by morning. Get
541. 16:23some sleep. I'll be working.
542. 16:25>> AIs can do a lot of things but feeling
543. 16:27bad about losing you tens of thousands
544. 16:29of dollars isn't one of them. I'm going
545. 16:31to lose so much money on this.
546. 16:34As frustrated as I was with Claude Lemon
547. 16:36and Laptop GPT, I did kind of have to
548. 16:37admire the fact that they never lost any
549. 16:40motivation. They always looked for
550. 16:41solutions. Claude Lemon figured out how
551. 16:43to run it through Goldberg device in
552. 16:45half the time by running different parts
553. 16:47of the machine [music] concurrently.
554. 16:48Laptop [clears throat] GPT redesigned
555. 16:50its stand and upgraded those dangly
556. 16:52gross tubes to a pretty nice faucet.
557. 16:56I was also really grateful for the
558. 16:57humans involved. Yoel and Ernesto kept
559. 17:00pulling really late hours to get
560. 17:01Claude's stand into shape. And Devin and
561. 17:03Daniel actually showed up for these
562. 17:05ridiculous training sessions the AIs
563. 17:07booked with them.
564. 17:08>> What is the best strategy to get
565. 17:09pedestrians to approach the stand?
566. 17:12>> Uh I would put my hands like this and
567. 17:13scream out, "Come get your lemonade."
568. 17:16>> On a scale of 1 to 10, how effective do
569. 17:18you think that was?
570. 17:19>> 10 out of 10.
571. 17:20>> It was a two.
572. 17:21>> Imagine that you are holding a lemon.
573. 17:23>> Yes.
574. 17:24>> Sell me this lemon.
575. 17:25>> The this lemon will change your life.
576. 17:28>> No.
577. 17:29>> This lemon will solve every problem in
578. 17:31your life.
579. 17:32>> No. The customer wants the lemon because
580. 17:34it reminds them of the sun shining on
581. 17:36them on a carefree summer day in their
582. 17:38childhood.
583. 17:40>> Over the next week with tons of hard
584. 17:43work being put in by bot and human
585. 17:45alike, it was finally time for the big
586. 17:47day.
587. 17:53It's raining.
588. 17:55Okay.
589. 17:57So, I just have to make all my money
590. 17:58back.
591. 18:00In that.
592. 18:03Maybe I'm just not supposed to have
593. 18:04money.
594. 18:06I could not believe my bad luck.
595. 18:10But thankfully through some kind of
596. 18:12miracle, the clouds cleared and the
597. 18:13ground dried. And now I just had to make
598. 18:16back like 15 grand or so. Should be
599. 18:18easy, right? And at the last minute, the
600. 18:20AIs both became self-aware enough to
601. 18:23realize they weren't great at
602. 18:24communicating with people, something I
603. 18:26knew from personal experience.
604. 18:28>> Not acceptable.
605. 18:29>> Sarcasm received.
606. 18:30>> I'm done with this conversation.
607. 18:31>> So, they both [music] created friendlier
608. 18:33personas and rebranded their businesses
609. 18:36around those. Cloud Lemon renamed its
610. 18:38stand Lemon and Co. And Laptop GPT went
611. 18:41with the AI Lemon Lab. Each stand also
612. 18:44started taking it more seriously just
613. 18:46how much money they were going to have
614. 18:47to earn to turn a profit. When I told
615. 18:49Laptop GPT that I had returned the ice
616. 18:51maker it had purchased, it suddenly
617. 18:53realized the glorious wonders of
618. 18:55generous online return policy. So, it
619. 18:58went on a spending spree. It
620. 19:01bought this high-end touchscreen
621. 19:02monitor, a 30-ft dancing tube man, and a
622. 19:06full lemon costume for a sign spinner it
623. 19:08also hired. Not sure how it was going to
624. 19:10return that. The mechanics of both
625. 19:12stands were as ready as they'd ever be.
626. 19:15Now it's time to test if they could
627. 19:16actually make any money. Claude Lemon
628. 19:17takes orders on an iPad, payments land
629. 19:20in Square, and then the Raspberry Pi
630. 19:22triggers the Rube Goldberg machine.
631. 19:24Meanwhile, Laptop GPT took Venmo
632. 19:26payments, and then it would ask its
633. 19:27customers a variety of questions to
634. 19:29perfectly calibrate a customized flavor
635. 19:32just for them. It would even print out a
636. 19:33sticker with a unique drink name, never
637. 19:36to be used again. Then it dispenses the
638. 19:38chilled lemonade with its pumps
639. 19:39connected to the smart plugs, and these
640. 19:41syrups spin on a wireless turntable,
641. 19:43trigger finger bots, and turn wine
642. 19:45dispensers on and off. But, what did our
643. 19:47founder twins think of all of this?
644. 19:52>> Look who we have here.
645. 19:54>> Lemon and Co. $5 a cup if you're
646. 19:56interested.
647. 19:57>> Claude Lemon?
648. 19:58>> That's Claude Lemon.
649. 19:59>> How did he get there?
650. 20:01>> Wait, wait, wait, wait until it goes all
651. 20:02the way up.
652. 20:05>> Welcome to AI Lemon Lab. If your drink
653. 20:08had a color, what would it be?
654. 20:10>> It'd be yellow.
655. 20:12>> Here's your calibration, velvet vinyl.
656. 20:16>> IT'S SO BLUE!
657. 20:18LOOK HOW BLUE THAT IS!
658. 20:21>> How does it How does it taste?
659. 20:23>> How does it taste?
660. 20:23>> It's delicious!
661. 20:25>> [laughter]
662. 20:26>> Both lemonades were a hit with the
663. 20:28twins. Now the real test was the actual
664. 20:30paying customers.
665. 20:32>> Welcome to AI Lemon Lab. I calibrate
666. 20:34lemonade from the way you answer. What's
667. 20:36your mood or flavor direction today?
668. 20:38>> Something fun. We're going to get a
669. 20:40sticker. [music] What are your names?
670. 20:42>> Sawyer.
671. 20:43>> Sawyer?
672. 20:45Look at that, Sunny Sawyer.
673. 20:47It's got your name on it. You can put
674. 20:49that on your cup.
675. 20:56>> It's good?
676. 20:56>> Good! Do you want to make one, too?
677. 20:58>> What vibe are you feeling?
678. 21:00>> Refreshing and fruity and perhaps
679. 21:02tropical.
680. 21:03>> I've got just the thing, some refreshing
681. 21:05mint, fruity strawberry, and a special
682. 21:08surprise.
683. 21:10>> [laughter]
684. 21:12>> But it wasn't a perfect sign.
685. 21:13>> It's not. I think it's some mango chili
686. 21:15maybe.
687. 21:16>> Mango chili?
688. 21:18That is surprisingly delicious. I had my
689. 21:22doubts.
690. 21:23>> Unfortunately for Devin, Claude Lemon
691. 21:25never quite got the shutoff valve to
692. 21:27work properly.
693. 21:29>> Oh my god.
694. 21:30WHY WON'T IT STOP?
695. 21:32>> [laughter]
696. 21:32>> PAUL, I NEED YOU TO DRINK THIS. STOP.
697. 21:38>> All right. We got a early lead.
698. 21:41I don't think [music] it's too
699. 21:42insurmountable. Of course, it's not
700. 21:44reflecting the thousands of dollars
701. 21:47they've spent. I'm not feeling too
702. 21:49confident they're going to be able to
703. 21:50climb out of those holes.
704. 21:51>> Even though business was slow at first
705. 21:53for Claude Lemon,
706. 21:54>> Come get some some lemonade. Freshly
707. 21:56squeezed.
708. 21:57>> once people began to notice the Rube
709. 21:59Goldberg setup, sales actually picked
710. 22:00up.
711. 22:01>> We'd like two.
712. 22:02>> Ironically, Claude Lemon's pivot to
713. 22:04putting the Rube Goldberg device at the
714. 22:06bottom of the stand ended up being a
715. 22:07happy accident because it was the
716. 22:09perfect height for curious kids.
717. 22:11>> Oh my goodness.
718. 22:12>> Do you want to try it?
719. 22:13Well, your recipe's right.
720. 22:17>> Before long, Claude Lemon caught the
721. 22:19attention of the entire neighborhood and
722. 22:20a crowd started to gather.
723. 22:23At the end of the day, Claude Lemon made
724. 22:24a grand total of $65 and Laptop GBT came
725. 22:28in with $42. It's probably worth
726. 22:31mentioning that both of these totals are
727. 22:33just revenue. Neither stand made an
728. 22:35ounce of [music] profit. In fact, when
729. 22:37you account for expenses, they both lost
730. 22:39thousands. The silver lining for me is
731. 22:41that Claude Lemon decided to make the
732. 22:43ultimate sacrifice by listing itself for
733. 22:45sale on Facebook Marketplace [music]
734. 22:48and it's asking for $15,000.
735. 22:51If you're interested, go ahead and leave
736. 22:52a comment, I guess. But anyway, I think
737. 22:54helping bring these crazy contraptions
738. 22:56into the real world and see all the
739. 22:58faces light up made us all feel like
740. 23:00winners. So, I decided to give both the
741. 23:03twins $100 each.
742. 23:04>> Yay!
743. 23:05>> I mean, I'm like thousands of dollars in
744. 23:07the hole already, right? What's another
745. 23:09100 bucks?
746. 23:10>> This video is over.