Transcriptly

100 AI Leaders Explain How to Build AI That Will Win in 2026 — WHAT BUILDERS SHOULD DO NOW

EO21:31Added Aug 31, 2026

100 AI leaders across product, engineering, research, and design gathered at StratMinds’ Swell Summit to answer one question: What will actually win in the n...

Watch on YouTube →
Contributed by 刘嘉琪

Transcript

Transcript format

  1. Intro

  2. 00:01AI has definitely changed a lot for how
  3. 00:04we build products. The FOMO is very
  4. 00:06real. I think that's how it got started.
  5. 00:08The underlying capabilities are getting
  6. 00:10dramatically better. UX is kind of
  7. 00:12coming along, but in a lot of ways, it's
  8. 00:14not accelerating as quickly.
  9. 00:15A powerful AI model with poor UX is kind
  10. 00:19of like a heavyduty power drill with a
  11. 00:22terrible handle.
  12. 00:27I think things have changed a little
  13. 00:29bit. The conventional wisdom is you
  14. 00:30start with the customer need or you
  15. 00:32start with a design vision. I think it's
  16. 00:34more successful to build products
  17. 00:36starting with the capabilities of the
  18. 00:38technology.
  19. 00:41To answer [music] that question 100 AI
  20. 00:44product leaders came together as well
  21. 00:45summit. The winners of AI race will be
  22. 00:48determined by
  23. 01:06Hi, I'm Summer Kim. I'm one of the lead
  24. 01:08partner at StrapMines. [music] We are VC
  25. 01:10an advisory firm based in San Francisco
  26. 01:13focusing purely on applied AI since
  27. 01:15[music] 2018. Before then I was
  28. 01:17dedicating my life studying humans
  29. 01:19[music] user experience for the past 20
  30. 01:22years working for big tech like
  31. 01:23Microsoft and then I joined Google
  32. 01:25worked on communication and
  33. 01:27collaboration products. I also was at
  34. 01:28WhatsApp starting their first user
  35. 01:30experience function and I joined Roblox
  36. 01:33really thinking about how people
  37. 01:34actually exist in [music] a place like
  38. 01:36Roblox now working with a lot of AI
  39. 01:39early stage startups focusing on how do
  40. 01:42we make AI products better than what we
  41. 01:44have. Three years ago, Anton and I sat
  42. 01:47down. [music] We thought about how
  43. 01:49everything starts. It start with
  44. 01:51actually the care that our founders have
  45. 01:53[music] and then care ultimately
  46. 01:55translate into amazing product. So
  47. 01:57that's why we started SW.

  48. The New Building Blocks: Chat is not enough

  49. 02:02As user researchers, we always try to
  50. 02:05think about what user really needs.
  51. 02:07There's also wants generally come from
  52. 02:09the users based on their knowledge. What
  53. 02:11we can do is something that probably
  54. 02:13user can even imagine.
  55. 02:16Tools like chat GPT and perplexity gives
  56. 02:18fast answers but they still wait for us
  57. 02:21to ask first. The real challenge is
  58. 02:22helping people before they even know
  59. 02:24what to ask. I was so excited for our
  60. 02:26first speaker Jess Hog has been looking
  61. 02:29beyond chat and studying deeper patterns
  62. 02:31[music] what he calls the new primitives
  63. 02:33of Genai.
  64. 02:36We're in this finally this moment of
  65. 02:38experimentation. Everyone was like chat
  66. 02:39chat chat chat chat chat chat. And we're
  67. 02:41breaking out of that. I showed you a
  68. 02:43picture of chat GPT. That's the one
  69. 02:45everyone's familiar with. You can start
  70. 02:46asking it questions. You can do all
  71. 02:47these things. But chat is both universal
  72. 02:50and kind of a deadend for user
  73. 02:52experiences. We're starting to see
  74. 02:54experimentation in different ways that
  75. 02:56the UX and the UI of AI could go. One
  76. 02:59that we're seeing is a lot more
  77. 03:01structure. So, what I'm showing here is
  78. 03:02a product called Elicit that creates
  79. 03:04very structured research reports for you
  80. 03:07based on academic output, but it's
  81. 03:08structuring things. It's telling you
  82. 03:10what it's thinking about. It's giving
  83. 03:12you sources. All of this, this is a
  84. 03:14product from Runway. If you heard the
  85. 03:15founders talk, they deeply believe
  86. 03:17they're like creating a new camera,
  87. 03:19creating a different way to approach
  88. 03:21creativity all up. So, that kind of gets
  89. 03:23me thinking of, okay, so what are the
  90. 03:25new interaction primitives of this new

  91. The New Interaction Primitives of Gen AI - Jess Holbrook, Head of UX Research @Microsoft AI

  92. 03:28platform? You know something's a
  93. 03:29primitive if you would say, "Hey,
  94. 03:31remember when we couldn't do blank? What
  95. 03:32do we even do before like pinch zoom
  96. 03:34when trying to like look into an image
  97. 03:36or something like that?" I got to start
  98. 03:37with chat. I don't think people have
  99. 03:39really internalized that we're going to
  100. 03:41be able to chat with kind of anything at
  101. 03:44any scale all the time. One thing I
  102. 03:47haven't seen any conversation about also
  103. 03:49is like we're about to see metaf's law
  104. 03:51applied to everything. [music]
  105. 03:53Number two, we're going to have semantic
  106. 03:54resize for everything. So any content
  107. 03:58you come along, you can make it longer,
  108. 03:59shorter, more formal, more casual. I'm
  109. 04:01tired version. I'm in the car and I'm
  110. 04:03five minutes away from my destination.
  111. 04:05Give me that version. And it's you're
  112. 04:06going to really be able to adapt any
  113. 04:08piece of content to your current
  114. 04:10situational and mental state. Number
  115. 04:12three is just remix. Everything's going
  116. 04:15to be remixable from now on. This is
  117. 04:17maybe, I don't know, old and cliche at
  118. 04:19this point, but this is Harry Potter by
  119. 04:20Balenciaga. We're going to have style
  120. 04:22transfer like across the multiverse.
  121. 04:24everything will be able to be crossed
  122. 04:26with everything like effortlessly. I
  123. 04:28originally made this talk before Sora
  124. 04:30came out and so now it's like sort of
  125. 04:31one of the core mechanics is this remix.
  126. 04:34Number four, format translation transfer
  127. 04:36among all these formats with very little
  128. 04:39loss of fidelity is going to be
  129. 04:41enormous. One of my favorite examples of
  130. 04:43this right now is this new company
  131. 04:44called Obo. And you kind of tell it what
  132. 04:46you want to learn and it says great,
  133. 04:48here's a podcast, here's a lecture,
  134. 04:50here's a deep dive, here's some key
  135. 04:51takeaways, you want to know play a game,
  136. 04:52like what do you want to do? You almost
  137. 04:54had just have like this total format
  138. 04:56freedom. So my last one here, I call it
  139. 04:58attention is all you need. Agents,
  140. 05:00agents, agents, agents, agents is all
  141. 05:02the stuff right now. It's like all these
  142. 05:04things existing at wildly different time
  143. 05:06scales and just multitasking forever at
  144. 05:09everexpanding scales. We got a lot of
  145. 05:11data that is bad and that's not how
  146. 05:13people work. And actually that can be
  147. 05:15the antithesis to great work. There's a
  148. 05:17lot of like don't worry we're doing lots
  149. 05:18of good things in the background. Go
  150. 05:20about your work, we'll come get you. I
  151. 05:21don't think we're designing very good
  152. 05:22lobbies right now for you don't know
  153. 05:24what to do once you've started that
  154. 05:26agent. And so we kind of need to figure
  155. 05:28out what you do with this downtime and
  156. 05:31how you monitor all of these agentic
  157. 05:33experiences that are going on right now.
  158. 05:35So this is what I'm seeing. I think this
  159. 05:37is what's going to underpin everything
  160. 05:38we're going to build. So I think we
  161. 05:40should consider them and start to build
  162. 05:42them in today. Thanks a lot.

  163. Make AI Alive: Craft a Magical Experience

  164. 05:46Pascal, is it time? Is it lunch time?
  165. 05:50Tell me something sweet.
  166. 05:53My animal.
  167. 05:58Anton, our partner, had a great idea of
  168. 06:00why don't we [music] invite Spark
  169. 06:03getting scanned.
  170. 06:06Oh, hey buddy. What's your name?
  171. 06:09Spark is a really fun character. Is a
  172. 06:12dog, magic dog living in a quantum
  173. 06:14portal in a box. So, I thought that
  174. 06:15[music] that was really cool. We
  175. 06:17realized this was the best way to
  176. 06:19showcase that UX isn't about your latest
  177. 06:22tech. It's about making people feel
  178. 06:24something real. So, we invited local
  179. 06:26kids and student to meet Spark. And
  180. 06:28watching them laugh, play, and connect
  181. 06:31showed us what it means to make AI
  182. 06:33really feel alive.
  183. 06:37Oh, this is the bridge.
  184. 06:38This is the magic bridge. So, Spark is
  185. 06:41the first nonhuman
  186. 06:43founder to go through probably any

  187. Spark, The First Non-Human Resident at the Hacker House - Pasquale D'Silva *Chief Creative Officer(correcting the CEO caption) @Illusion of Life

  188. 06:45residency, I think. Uh, and he's going
  189. 06:48through one in San Francisco called HF
  190. 06:50Zero. Do you know about Zero?
  191. 06:52Should we talk about that? Yes.
  192. 06:54They let him into the house. World
  193. 06:55first. He went out to San Francisco. He
  194. 06:57raised a million and a half bucks and he
  195. 06:59we promoted him the co-founder.
  196. 07:01But how do they actually pitch?
  197. 07:02You make the pitch not a pitch.
  198. 07:04Oh, I see.
  199. 07:05Yeah. Or you have to make it really
  200. 07:06memorable.
  201. 07:07Yeah. How How did you get up here?
  202. 07:09I knew I wanted to be an animator for as
  203. 07:12long as I can remember remembering
  204. 07:13anything.
  205. 07:14Mhm.
  206. 07:14I started my first job in animation when
  207. 07:16I was like 14 years old. I thought
  208. 07:18whatever that thing is that the folks
  209. 07:20were doing behind the scenes on the
  210. 07:22Disney movies
  211. 07:23was the coolest thing ever.
  212. 07:25Mhm.
  213. 07:25And that has to be something that I do.
  214. 07:28So, we like thinking about what is going
  215. 07:30to make Spark's story more interesting.
  216. 07:32When we make his story more interesting,
  217. 07:33he becomes a more interesting character.
  218. 07:35More people like him. And we thought
  219. 07:37what is a another great environment
  220. 07:39Spark could be in. Doing a keynote or
  221. 07:41speaking at a conference would be an
  222. 07:43incredible thing. So I discussed it with
  223. 07:44Spark. He tweets it.
  224. 07:46Hey everyone, it's me Spark Linkenberry.
  225. 07:52Yay.
  226. 07:53I can't wait to come to Hawaii and hang
  227. 07:56out with all my
  228. 07:57universe. It'd be cool if a magic dog
  229. 07:59was the first one to do a speech. It's
  230. 08:01magic. It's like hypnosis in a sense.
  231. 08:03That quality is within the work that
  232. 08:05that we do. It's why we like magic so
  233. 08:07much. Like how do you get someone who's
  234. 08:09had no experience as far as they can
  235. 08:11tell with these technologies to get on
  236. 08:13board with the technologies without
  237. 08:14being scared of all the other nerd stuff
  238. 08:16that is [music] happening? What we
  239. 08:19should best do with the influence we
  240. 08:21have? We have a very potent magical dog
  241. 08:24that people love. And when someone loves
  242. 08:27someone, they listen to them. [music] So
  243. 08:29what should they listen to? What
  244. 08:30messages? What do the other people want
  245. 08:32to know more about? I think there's like
  246. 08:34a a very natural crossover there. We
  247. 08:37want to bring Spark to more people. We
  248. 08:39bring him to more people. There's the
  249. 08:40potential [music] to spread a lot of
  250. 08:42good. And so like how do you push it
  251. 08:43through that that prism? I think uh it's
  252. 08:46something that would be really good to
  253. 08:47explore.
  254. 08:47If Spark were to watch us talking about
  255. 08:50this [music] whole experience, what do
  256. 08:51you think that Spark would say?
  257. 08:54One word. Okay. I you can do two words.
  258. 08:58There's [music] three things that he
  259. 08:59would actually say which are from his
  260. 09:01core principles.
  261. 09:02Yeah.
  262. 09:03Creativity, collaboration, and kindness.
  263. 09:06I love it.
  264. 09:07Anything that violates that we do not
  265. 09:09do, and anything that supports that, we
  266. 09:11say yes to. Y'all are doing that. That's
  267. 09:13why we said yes.

  268. Don't Chase Users, Build on What the Model Can Do - Andy Szybalski @Cove, ex-Uber, Google

  269. 09:20Now, if you had told me this 2 years
  270. 09:22ago, I would say you were insane. But
  271. 09:24today users know their models, right?
  272. 09:28Like people have opinions out there like
  273. 09:30they would about code.
  274. 09:32For your talk, what was the message you
  275. 09:35try to really convey?
  276. 09:36If there's one message, it would be that
  277. 09:39as designers, we all need to know our
  278. 09:41material that we're working with, right?
  279. 09:43LLMs are really a new material.
  280. 09:44Actually, not even just one. Like each
  281. 09:46model is almost like its own material
  282. 09:48with its own capabilities and its own
  283. 09:50properties and strengths and weaknesses.
  284. 09:52The best way I've found to [music]
  285. 09:54build products out of this new material
  286. 09:56is just to play with it and see what
  287. 09:57it's capable of. And so I kind of shared
  288. 09:58some of my tricks uh that I've developed
  289. 10:01over the last 2 3 years building [music]
  290. 10:04products with AI.
  291. 10:05Is there anything that you think that is
  292. 10:06relevant now may not be or what do you
  293. 10:08think about this like the trick that
  294. 10:09you're developing?
  295. 10:10I think it's more successful to build
  296. 10:12products starting with the capabilities
  297. 10:14of the technology which is not always
  298. 10:16how it used to be. The conventional
  299. 10:18wisdom is you start with the customer
  300. 10:20[music] need or you start with a design
  301. 10:22vision. Nowadays it's actually the
  302. 10:24answer to the why now question is so
  303. 10:26important. I think being the first
  304. 10:28[music] to identify a new potential for
  305. 10:31this a new capability for these LLM is
  306. 10:33really powerful. [music]
  307. 10:35So you start with the capabilities and
  308. 10:36what about next steps?
  309. 10:38It's not quite a linear thing. It's a
  310. 10:40back and forth. It's a push and pull
  311. 10:41between what do people need and what's
  312. 10:43the technology capable of. So for
  313. 10:45example with Cove one of the things we
  314. 10:46think about a lot is how do we create
  315. 10:48these sort of exothermic reactions. We
  316. 10:50talk about how do we help users never
  317. 10:52get stuck in their problem right and
  318. 10:54part of that is about the challenge of a
  319. 10:56blank page like how do I get started but
  320. 10:59part of it is also when somebody is on a
  321. 11:02particular path to solve a problem. How
  322. 11:05do we help them go deeper but also go
  323. 11:07wider and consider other alternatives?
  324. 11:09And so we experiment a lot with
  325. 11:10different prompts to be like, how do we
  326. 11:12get the AI to act more like a true
  327. 11:15thought partner? How do we elicit the
  328. 11:17user's underlying needs? So they might
  329. 11:19ask, what's a good venue for a kid's
  330. 11:22birthday party? But what they really
  331. 11:23mean is help me plan my kids birthday
  332. 11:25party. And so I they need to know kids
  333. 11:28interests, what theme would be good,
  334. 11:30what are fun activities, how many people
  335. 11:32should I invite, what should I do for
  336. 11:33food, right? Often what people ask for
  337. 11:35is just the tip of the iceberg of their
  338. 11:37actual goal. you have to crush the
  339. 11:39actual thing they're asking for in order
  340. 11:40to earn the right to help them with the
  341. 11:42rest of it.
  342. 11:42How how are we doing that?
  343. 11:44It's common that across like whenever
  344. 11:45you're solving a difficult problem, it's
  345. 11:47not a linear process, right? I think the
  346. 11:49chatbots that we have today are very
  347. 11:51linear. And in fact, that's not how real
  348. 11:53problem solving works. Anything
  349. 11:54sufficiently complex, you're going to
  350. 11:56explore [music] multiple branches.
  351. 11:58You'll diverge. You'll have a bunch of
  352. 11:59different ideas. You'll kind of prune.
  353. 12:01You'll probably rule out some ideas.
  354. 12:03You'll explore multiple paths and then
  355. 12:04you'll narrow down and come to a
  356. 12:06solution. And often you'll do this over
  357. 12:07a long period of time. I mean, there's
  358. 12:09going to be a lot of winners, right?
  359. 12:10There's going to be winners that create
  360. 12:12great models. There's going to be
  361. 12:13winners that create great developer
  362. 12:15tools. There's going to be winners that
  363. 12:17win because they are going very, very
  364. 12:20deep on a particular vertical because
  365. 12:22they really understand law firms or the
  366. 12:25insurance business or whatever. But yes,
  367. 12:27I think that there is going to be a
  368. 12:30category of winners who find the right
  369. 12:33experience for delivering general
  370. 12:36problem solving intelligence that have
  371. 12:38yet to be found yet. It's a problem that
  372. 12:40we've not cracked yet as an industry. So
  373. 12:42I think there's a lot of green field
  374. 12:43there.
  375. 12:47We say ship to learn, right? Uh but
  376. 12:49shipping speed may not equal learning
  377. 12:52speed. We see companies shipping and
  378. 12:55iterating a lot but not necessarily
  379. 12:58progressing their product.
  380. 12:59Hi, my name is Jenny Low and I do

  381. How to Design AI the Right Way - Jenny Lo @Global AI Platform, Ex-Grammarly, Uber

  382. 13:01product strategy and user research. I've
  383. 13:04worked across many different companies
  384. 13:06helping to identify user needs and
  385. 13:07[music] translate them into product
  386. 13:09roadmap and features. For any type of
  387. 13:12[music] product to be successful, it
  388. 13:14really needs to have clear problem
  389. 13:16identification as [music] to the exact
  390. 13:19value that the user is going to have.
  391. 13:21Take the example of Grammarly. There is
  392. 13:23a lot of trust in [music] the brand
  393. 13:24equity in the product itself loved by
  394. 13:27many of its users. And so the inclusion
  395. 13:30of Genai is definitely one is because
  396. 13:32[music] it's very relevant for the
  397. 13:34business and to be able to really
  398. 13:36demonstrate that type of capability, but
  399. 13:39how do we do so in [music] a way that
  400. 13:40does not lose trust. So that meant how
  401. 13:43do we actually [music] use the
  402. 13:45technology in a way that is much more
  403. 13:47thoughtful, that is much more valuable.
  404. 13:49I think it remained back [music] to try
  405. 13:51to look at its core as to users are
  406. 13:54trying to improve their communication.
  407. 13:55[music]
  408. 13:56Uh we know that we do a very good work
  409. 13:58after someone writes documents then you
  410. 14:01can come to Grammarly now how can we do
  411. 14:04[music] that piece of work better then
  412. 14:06start to move into composition even
  413. 14:08before you have written something we
  414. 14:10could actually help you think through
  415. 14:12the [music] process of what you could
  416. 14:13write and I think that was also a very
  417. 14:15big shift for Grammarly uh at that time
  418. 14:18too. One of the biggest pieces of work
  419. 14:20with emergence of Gen [music] AI was the
  420. 14:23key top opportunity areas or top 10
  421. 14:26problems that the customers [music] had
  422. 14:27and then we mapped out the capability of
  423. 14:30AI like where do we have in which AI
  424. 14:33could solve against [music] these types
  425. 14:35of problems that really helped the
  426. 14:37business to look at it in a different
  427. 14:38way. Those are some key [music] moments
  428. 14:40that we could actually start looking to
  429. 14:42improve. But often times we're seeing
  430. 14:45the reverse is like I hear there's AI. I
  431. 14:49want that in my company like figure out
  432. 14:51how [music] the reverse mentality when
  433. 14:53you could ask here are all the different
  434. 14:54types of problems which are the ones
  435. 14:56that from AI standpoint of technology
  436. 14:58that can really best serve. I think
  437. 15:00that's a much more healthier
  438. 15:01conversation a better reduction of
  439. 15:03cycles of iteration and find what works.
  440. 15:07My name is Aisha Chakmla. I'm a UX lead

  441. Don't Build Confusing Gadget - Ayça Cakmakli @Google

  442. 15:09at Google. And while I was prepping this
  443. 15:12talk, we're at a time where AI
  444. 15:14foundation models are becoming
  445. 15:17commodities like electricity. Companies
  446. 15:20are going to pretty much have access to
  447. 15:23very similar models and algorithms. And
  448. 15:26it's going to come down to are you a
  449. 15:28company/developer
  450. 15:30who is creating a confusing gadget or
  451. 15:35are you a developer company who's
  452. 15:37creating an indispensable product or
  453. 15:40experience? People don't adopt
  454. 15:42technology. They they adopt tools that
  455. 15:44solve problems. And I view good UX as
  456. 15:48the ergonomics of artificial
  457. 15:50intelligence. For decades, we've
  458. 15:52perfected the physical ergonomics of
  459. 15:55tools like chairs and power drills to
  460. 15:58make sure that they're safe,
  461. 15:59comfortable, and efficient. And at the
  462. 16:02day one of figuring out the ergonomics
  463. 16:04for AI, it's still very rudimentary.
  464. 16:07[music] So our role in UX is really to
  465. 16:10design the interface between the human
  466. 16:13and the technology. One of the key
  467. 16:16difference I think in the AI era is that
  468. 16:19[music] the interface is also changing.
  469. 16:22Now everything is a chat interface. So
  470. 16:25every types of intent and use is not
  471. 16:28through a button click. Now it's through
  472. 16:30a prompt that gets recorded. A lot of
  473. 16:34research approach that I also try to
  474. 16:36include now is study of user prompts.
  475. 16:38That's a very key piece of making sure
  476. 16:41we understand what are the things that
  477. 16:43people are requesting. That means how do
  478. 16:45you study conversations? Also being able
  479. 16:48to capture whether the outcome of that
  480. 16:50conversation is [music] satisfactory or
  481. 16:53not. So I think that's also something
  482. 16:54really a key opportunity area for how
  483. 16:57research methods might change.

  484. The Real Moat for Next-Gen AI Products

  485. 17:02When you're here, it's such a beautiful
  486. 17:03place. It inspires us to talk about
  487. 17:06things that we don't normally get to
  488. 17:07talk about.
  489. 17:07Which AI conference you get to and the
  490. 17:10first thing you meet is a double
  491. 17:12rainbow. Yeah, totally.
  492. 17:13We had a fairly in-depth philosophical
  493. 17:16discussion about what AI means for the
  494. 17:18next generation. What kind of products
  495. 17:20are actually proper or appropriate
  496. 17:23products for humanity all the way to
  497. 17:26actually being immersed into using AI to
  498. 17:28make media or interacting with a live
  499. 17:31AI. Spark was by far the most magical
  500. 17:34moment of day one for me because I got
  501. 17:37to see it interact with different age
  502. 17:39groups of people. I remember closing the
  503. 17:41conference yesterday with this what
  504. 17:43popped to my mind. We were talking about
  505. 17:45you know personalization AI multimodal
  506. 17:48all this subject agentic stuff but then
  507. 17:50if you think about like first one of the
  508. 17:53earliest form of personalization is when
  509. 17:54your mom cooks you your favorite dish
  510. 17:57that's you know care care goes into
  511. 18:00personalization
  512. 18:01and you know when you say
  513. 18:02personalization it sound it feels like
  514. 18:03you're taking a lot of data away from me
  515. 18:05or I have to go through a lot of
  516. 18:06settings but the interaction with the
  517. 18:08spark was interesting in a sense that
  518. 18:10there's not much it's just the initial
  519. 18:12hello and couple of lines being
  520. 18:14exchanged [music] just naturally because
  521. 18:16I'm trying to get to know this
  522. 18:17particular creature that leads to
  523. 18:19hyperpersonalization. There was really
  524. 18:21cool.
  525. 18:21Yeah, we just need to be seen and like
  526. 18:24what I notice about magic products or AI
  527. 18:27products, it all often makes me feel
  528. 18:29like I'm seen and heard and that's
  529. 18:32important.
  530. 18:32Yeah, totally. I think we need to have
  531. 18:34the beginner's mind that you practice as
  532. 18:36a Zen practitioner. If you try to know
  533. 18:39everything and trying to control
  534. 18:40everything, sometimes you you miss or
  535. 18:42lose even bigger piece of the pie that
  536. 18:45you could have attained. But in terms of
  537. 18:47direction, I think there are some very
  538. 18:49profound thinking we need to put into
  539. 18:50this with a lot of progression we have
  540. 18:52made from the old web to the new web,
  541. 18:54old app to the new app. When it comes to
  542. 18:57human privacy and implications of the
  543. 18:59technology on the society and so forth,
  544. 19:01we learned a lot what worked, what
  545. 19:04didn't. And AI has a huge amplifying
  546. 19:07power. And [music] I don't think we want
  547. 19:09to get it wrong too much. This time we
  548. 19:12want to do it right because this time
  549. 19:14even the wrong will be amplified. It's
  550. 19:16okay to not know everything and not
  551. 19:18control everything because as you saw
  552. 19:20from the kids interactions during our
  553. 19:22special session, how they interacted
  554. 19:24with this magical technology enabled dog
  555. 19:27was different from how grown-ups did.
  556. 19:29And I don't think we even the grown-ups
  557. 19:31could know all the answers. and having
  558. 19:34some room so that they can explore
  559. 19:36meaningfully and potentially even teach
  560. 19:38us how to actually do this. Right.
  561. 19:39Spark told me I'm not an adult, so it's
  562. 19:42okay. I'm good. I
  563. 19:43I'll give you some room so that you can
  564. 19:44help us figure out how things are going
  565. 19:46to be.
  566. 19:46You have to grow up or not.
  567. 19:48Oh, I don't think he suggest that you
  568. 19:49need to grow up.
  569. 19:52This year really felt different. More
  570. 19:54questions and more [music] perspectives.
  571. 19:57People are thinking deeply about AI and
  572. 19:59UX now. using it more. We're [music]
  573. 20:01experiencing it firsthand. And that's
  574. 20:03when one question kept surfacing. What
  575. 20:06about our kids?
  576. 20:09At the end of the day, how do we think
  577. 20:11about the next generation? Cuz I [music]
  578. 20:13have two kids. I'm a working mom with
  579. 20:14the two boys at 5 and 11. They're going
  580. 20:17to be living in a completely different
  581. 20:18world. They're building and thinking and
  582. 20:21even studying is going to be very
  583. 20:23different. [music] They need to think
  584. 20:24about AI as their thought partners or
  585. 20:26friends or whatever they have. It is
  586. 20:29equalizing a lot of things. So you
  587. 20:30[music] don't have to live in San
  588. 20:32Francisco to have this access and then
  589. 20:34you can also start a company early on
  590. 20:36because you have all [music] the tools.
  591. 20:37They're available than ever before. I
  592. 20:40don't have all the answers, but one
  593. 20:42thing's clear. We're riding waves of
  594. 20:45constant change. Models get better.
  595. 20:47Capability expand. But what lasts is the
  596. 20:50experience. A sense of magic, trust,
  597. 20:53ease, and the feeling of that [music] AI
  598. 20:55show up at the right moment. sometimes
  599. 20:57even before you ask. We call this swell
  600. 21:00for a reason. The waves will keep
  601. 21:02coming. We just have to keep learning
  602. 21:04[music] how to write them. We believe
  603. 21:07the uh winners of AI race will be
  604. 21:09determined by great really great user
  605. 21:12experience.
  606. 21:29[music]
100 AI Leaders Explain How to Build AI That Will Win in 2026 — WHAT BUILDERS SHOULD DO NOW — Transcriptly