1 00:00:00,227 --> 00:00:02,351 The Anthropic Founders Guide 2 00:00:02,507 --> 00:00:03,891 Written for people who should not start a company 3 00:00:04,546 --> 00:00:05,131 A few months ago 4 00:00:05,466 --> 00:00:07,791 This world leading AI company 5 00:00:07,827 --> 00:00:09,832 Released a thirty six page document 6 00:00:10,067 --> 00:00:11,705 It was not a product manual 7 00:00:11,707 --> 00:00:13,391 Or a technical white paper 8 00:00:13,547 --> 00:00:15,192 It was called the Founders Guide 9 00:00:15,467 --> 00:00:16,492 I spent an entire afternoon 10 00:00:16,667 --> 00:00:18,192 Reading it from beginning to end 11 00:00:18,387 --> 00:00:20,611 And came away with a powerful feeling 12 00:00:21,027 --> 00:00:24,106 On the surface this guide is for founders 13 00:00:24,106 --> 00:00:25,865 But its real message 14 00:00:25,867 --> 00:00:28,911 Is for people preparing to start a company 15 00:00:29,027 --> 00:00:30,812 And it is not encouraging them 16 00:00:31,067 --> 00:00:32,964 It is warning them 17 00:00:33,159 --> 00:00:33,503 Hello 18 00:00:33,918 --> 00:00:34,384 Everyone 19 00:00:34,678 --> 00:00:35,683 I am Leo Wang 20 00:00:35,998 --> 00:00:38,844 Today I want to discuss something that deeply worries me 21 00:00:39,147 --> 00:00:41,193 I have been an angel investor for over a decade 22 00:00:41,307 --> 00:00:43,026 And the founders I have met over the years 23 00:00:43,028 --> 00:00:44,752 Number in the thousands at least 24 00:00:45,028 --> 00:00:46,827 When someone used to pitch me a project 25 00:00:46,827 --> 00:00:48,572 My first question was usually 26 00:00:48,907 --> 00:00:50,492 Is your technical team ready 27 00:00:50,947 --> 00:00:52,413 Can you actually build it 28 00:00:53,068 --> 00:00:54,986 Over the past decade this question 29 00:00:54,988 --> 00:00:56,873 Was almost a matter of life and death 30 00:00:57,227 --> 00:01:00,072 Whether you could turn an idea into something real 31 00:01:00,108 --> 00:01:03,272 Determined whether you crossed the first startup hurdle 32 00:01:03,547 --> 00:01:03,932 Now 33 00:01:04,227 --> 00:01:06,252 That question no longer needs asking 34 00:01:06,606 --> 00:01:08,371 Just look around you 35 00:01:08,606 --> 00:01:10,351 Someone with no technical background 36 00:01:10,726 --> 00:01:12,046 Can open an AI tool 37 00:01:12,047 --> 00:01:15,132 And describe what they want in plain language 38 00:01:15,287 --> 00:01:16,931 Then build a respectable product prototype 39 00:01:16,966 --> 00:01:18,892 In just three days 40 00:01:19,246 --> 00:01:20,032 Five years ago 41 00:01:20,326 --> 00:01:21,445 That would have taken three months 42 00:01:21,446 --> 00:01:23,052 And required three engineers 43 00:01:23,407 --> 00:01:24,451 That is no exaggeration 44 00:01:24,806 --> 00:01:27,371 I have seen more and more cases like this lately 45 00:01:27,606 --> 00:01:30,972 Someone built a translation plugin with AI over a weekend 46 00:01:31,086 --> 00:01:32,711 And got hundreds of downloads at launch 47 00:01:33,287 --> 00:01:36,731 Someone else built a customer service bot in two days 48 00:01:36,847 --> 00:01:39,132 Then took it straight to a client and signed a contract 49 00:01:39,572 --> 00:01:41,331 The Anthropic guide 50 00:01:41,331 --> 00:01:43,217 Even devotes a whole chapter to this 51 00:01:43,453 --> 00:01:44,251 It says 52 00:01:44,251 --> 00:01:47,217 AI has compressed the entire startup lifecycle 53 00:01:47,451 --> 00:01:49,396 Going from idea to launch used to take 54 00:01:49,811 --> 00:01:50,977 At least six months 55 00:01:51,411 --> 00:01:53,777 Now it may take only a few weeks 56 00:01:53,811 --> 00:01:54,777 Or even a few days 57 00:01:55,212 --> 00:01:56,456 And speed is not the only change 58 00:01:56,572 --> 00:01:58,576 The founders role has changed too 59 00:01:58,771 --> 00:02:00,736 You used to need to be a hands on builder 60 00:02:00,891 --> 00:02:02,617 Now you are more like a director 61 00:02:02,691 --> 00:02:04,396 You decide what the film should be 62 00:02:04,691 --> 00:02:06,016 And AI executes 63 00:02:07,173 --> 00:02:09,737 That sounds wonderful 64 00:02:10,172 --> 00:02:12,098 The barrier to entrepreneurship is gone 65 00:02:12,292 --> 00:02:14,417 Technology is no longer a moat 66 00:02:14,533 --> 00:02:16,218 Everyone has a chance 67 00:02:16,515 --> 00:02:16,957 Right 68 00:02:17,597 --> 00:02:19,635 But one figure in this guide 69 00:02:19,637 --> 00:02:21,481 Sent a chill through me 70 00:02:21,717 --> 00:02:22,502 Over the past decade 71 00:02:22,836 --> 00:02:24,756 Among the reasons startups failed 72 00:02:24,757 --> 00:02:25,461 The number one reason 73 00:02:25,757 --> 00:02:26,722 Was not lack of money 74 00:02:26,836 --> 00:02:27,741 Or weak technology 75 00:02:28,076 --> 00:02:29,521 Or a team falling apart 76 00:02:29,637 --> 00:02:30,701 The leading reason was 77 00:02:31,197 --> 00:02:33,681 They built something nobody wanted 78 00:02:34,516 --> 00:02:35,882 Forty two percent 79 00:02:36,437 --> 00:02:37,521 Nearly half 80 00:02:38,116 --> 00:02:39,681 Think carefully about that 81 00:02:40,076 --> 00:02:41,021 It was not that they could not build it 82 00:02:41,477 --> 00:02:42,635 They did build it 83 00:02:42,637 --> 00:02:43,722 But nobody wanted it 84 00:02:44,197 --> 00:02:46,361 Before AI appeared 85 00:02:46,477 --> 00:02:49,081 This was already the biggest killer of startups 86 00:02:49,317 --> 00:02:50,556 After AI appeared 87 00:02:50,556 --> 00:02:52,562 The problem was not solved 88 00:02:52,597 --> 00:02:53,961 It was amplified 89 00:02:54,037 --> 00:02:55,887 That is what worries me 90 00:02:56,162 --> 00:02:58,887 The guide honestly lists three traps 91 00:02:59,123 --> 00:03:00,767 Let me explain them one by one 92 00:03:01,253 --> 00:03:01,878 The first trap 93 00:03:02,452 --> 00:03:04,378 AI is an exceptionally obedient assistant 94 00:03:04,773 --> 00:03:06,338 You may have experienced this 95 00:03:06,653 --> 00:03:07,398 You tell AI 96 00:03:07,813 --> 00:03:08,892 I want to build 97 00:03:08,892 --> 00:03:11,537 An app that helps older people book medical appointments 98 00:03:11,813 --> 00:03:12,517 AI will not say 99 00:03:12,773 --> 00:03:13,417 Wait 100 00:03:13,452 --> 00:03:14,998 Are you sure older people need this 101 00:03:15,292 --> 00:03:16,197 Have you asked them 102 00:03:16,533 --> 00:03:19,092 Have you spent a morning at a hospital registration desk 103 00:03:19,093 --> 00:03:21,158 To see exactly where older people get stuck 104 00:03:21,572 --> 00:03:22,898 It will not ask you any of that 105 00:03:23,132 --> 00:03:23,637 It will say 106 00:03:23,852 --> 00:03:24,658 Great idea 107 00:03:24,693 --> 00:03:26,977 I will help you analyze your target users 108 00:03:27,053 --> 00:03:28,652 Design the feature structure 109 00:03:28,653 --> 00:03:29,697 And write the code 110 00:03:29,732 --> 00:03:32,058 You will have a working prototype in three days 111 00:03:32,572 --> 00:03:35,212 Its job is to help realize your idea 112 00:03:35,213 --> 00:03:37,058 Not to challenge it 113 00:03:37,705 --> 00:03:39,549 In the past when you had a startup idea 114 00:03:39,624 --> 00:03:40,950 You had to go out and talk to people 115 00:03:41,424 --> 00:03:42,809 You had to meet investors for coffee 116 00:03:43,265 --> 00:03:44,390 Have dinner with friends 117 00:03:44,545 --> 00:03:47,070 And sit down with the people you hoped to serve 118 00:03:47,545 --> 00:03:48,989 That process was painful 119 00:03:49,105 --> 00:03:50,910 Because someone always poured cold water on you 120 00:03:51,144 --> 00:03:51,769 Someone always said 121 00:03:52,144 --> 00:03:53,303 I would never use this 122 00:03:53,304 --> 00:03:54,309 I do not think it will work 123 00:03:54,624 --> 00:03:55,690 Even your mother might say 124 00:03:56,065 --> 00:03:57,209 Why not just keep your steady job 125 00:03:57,864 --> 00:03:58,570 But you know what 126 00:03:58,904 --> 00:04:00,749 That cold water has value 127 00:04:00,864 --> 00:04:03,344 Before you invest too much 128 00:04:03,344 --> 00:04:05,793 It helps you discover that your direction may be wrong 129 00:04:05,988 --> 00:04:06,573 And now 130 00:04:07,388 --> 00:04:08,874 You no longer need to go out and face criticism 131 00:04:09,349 --> 00:04:12,034 You can sit at home and talk with AI 132 00:04:12,668 --> 00:04:13,573 You name a direction 133 00:04:13,988 --> 00:04:15,708 And it finds ten reasons 134 00:04:15,708 --> 00:04:17,393 Why that direction is right 135 00:04:17,628 --> 00:04:18,693 You want market data 136 00:04:18,988 --> 00:04:19,853 It gives you data 137 00:04:20,428 --> 00:04:21,414 You want success stories 138 00:04:21,748 --> 00:04:22,654 It gives you examples 139 00:04:23,149 --> 00:04:24,474 You want competitor analysis 140 00:04:24,548 --> 00:04:26,514 It gives you a polished report 141 00:04:26,909 --> 00:04:29,113 After the conversation you feel completely confident 142 00:04:29,228 --> 00:04:31,274 As though you have done thorough research 143 00:04:31,628 --> 00:04:33,394 But what you did was not research 144 00:04:33,469 --> 00:04:36,714 It was a carefully staged act of self persuasion 145 00:04:37,149 --> 00:04:37,733 Think about it 146 00:04:38,228 --> 00:04:40,034 As you grow more excited during the conversation 147 00:04:40,188 --> 00:04:42,173 Have you ever paused to ask yourself 148 00:04:42,789 --> 00:04:43,894 Am I looking for an answer 149 00:04:44,188 --> 00:04:45,693 Or for someone to agree with me 150 00:04:46,965 --> 00:04:48,564 The guide calls this phenomenon 151 00:04:48,566 --> 00:04:50,730 A confirmation bias amplifier 152 00:04:51,165 --> 00:04:52,030 The name does not matter 153 00:04:52,366 --> 00:04:54,051 Just remember the feeling 154 00:04:54,246 --> 00:04:55,770 The feeling of sitting at your computer 155 00:04:55,806 --> 00:04:58,491 And becoming more convinced that you are right 156 00:04:58,926 --> 00:05:01,650 That feeling itself is a warning sign 157 00:05:02,085 --> 00:05:04,090 You may have had moments like this too 158 00:05:04,525 --> 00:05:06,730 You become fascinated by a direction 159 00:05:06,886 --> 00:05:09,330 And discuss it with AI again and again 160 00:05:09,566 --> 00:05:11,451 Until it feels more and more promising 161 00:05:11,806 --> 00:05:13,171 But what if at that moment 162 00:05:13,326 --> 00:05:15,205 You invited a veteran 163 00:05:15,205 --> 00:05:17,150 With twenty years in that industry to dinner 164 00:05:17,725 --> 00:05:18,730 During a casual chat 165 00:05:18,845 --> 00:05:21,030 They might tell you before putting down their chopsticks 166 00:05:21,446 --> 00:05:23,650 Someone tried this ten years ago 167 00:05:23,686 --> 00:05:24,511 Burned through tens of millions 168 00:05:24,845 --> 00:05:25,811 And did not survive 169 00:05:26,486 --> 00:05:27,311 That kind of judgment 170 00:05:27,486 --> 00:05:28,850 Is something AI cannot give you 171 00:05:29,006 --> 00:05:30,285 Not because it is hiding anything 172 00:05:30,285 --> 00:05:32,371 But because it has no industry intuition 173 00:05:32,645 --> 00:05:35,571 It has data logic and language 174 00:05:35,645 --> 00:05:36,970 But it does not have the sixth sense 175 00:05:37,006 --> 00:05:39,371 That a person develops after twenty years 176 00:05:39,405 --> 00:05:41,131 Immersed in a particular industry 177 00:05:42,155 --> 00:05:43,340 The second trap is even more frightening 178 00:05:43,634 --> 00:05:44,679 Building the wrong thing 179 00:05:44,715 --> 00:05:46,160 And building it incredibly fast 180 00:05:46,515 --> 00:05:48,299 I often tell young founders something 181 00:05:48,715 --> 00:05:50,314 The most dangerous moment in a startup 182 00:05:50,314 --> 00:05:51,960 Is not when you cannot build it 183 00:05:52,035 --> 00:05:55,040 It is when you built it but chose the wrong direction 184 00:05:55,674 --> 00:05:57,400 This problem was less severe before 185 00:05:57,595 --> 00:06:00,354 You spent three months building a rough prototype 186 00:06:00,355 --> 00:06:01,520 Then showed it to users 187 00:06:01,715 --> 00:06:02,619 They said it was hard to use 188 00:06:02,754 --> 00:06:03,400 So you changed it 189 00:06:03,554 --> 00:06:04,619 After three failed revisions 190 00:06:04,795 --> 00:06:05,679 You gave up 191 00:06:05,795 --> 00:06:07,280 The whole process took six months 192 00:06:07,355 --> 00:06:09,480 You lost half a year and some money 193 00:06:09,634 --> 00:06:10,020 It hurt 194 00:06:10,275 --> 00:06:11,289 But you survived 195 00:06:11,644 --> 00:06:12,149 Now 196 00:06:12,724 --> 00:06:14,849 AI builds it for you in three days 197 00:06:15,204 --> 00:06:15,630 You look at it 198 00:06:15,964 --> 00:06:16,149 And think 199 00:06:16,604 --> 00:06:16,990 Not bad 200 00:06:17,485 --> 00:06:18,130 Keep going 201 00:06:18,604 --> 00:06:20,250 Three days later you add another feature 202 00:06:20,404 --> 00:06:22,329 Then another three days later you add one more 203 00:06:22,765 --> 00:06:23,764 After a month 204 00:06:23,765 --> 00:06:25,469 You have built a fully featured product 205 00:06:25,765 --> 00:06:26,550 With a beautiful interface 206 00:06:26,844 --> 00:06:28,849 And even a payment system 207 00:06:29,245 --> 00:06:30,649 But nobody uses it 208 00:06:31,828 --> 00:06:33,552 That is when things become hardest 209 00:06:33,987 --> 00:06:35,953 Because you are already deeply invested 210 00:06:36,227 --> 00:06:37,213 You have invested your emotions 211 00:06:37,547 --> 00:06:38,633 And your time 212 00:06:38,867 --> 00:06:41,112 You posted product screenshots on social media 213 00:06:41,268 --> 00:06:43,593 You told your family you were building something big 214 00:06:43,787 --> 00:06:45,992 You may even have registered a company 215 00:06:46,388 --> 00:06:48,833 You do not want to admit the direction may be wrong 216 00:06:49,148 --> 00:06:51,073 Because admitting it costs too much 217 00:06:51,347 --> 00:06:52,872 Not financially 218 00:06:52,948 --> 00:06:54,112 But psychologically 219 00:06:54,508 --> 00:06:56,186 It once took a long time 220 00:06:56,188 --> 00:06:57,872 To dig yourself a deep hole 221 00:06:58,388 --> 00:06:59,653 Now AI digs it for you 222 00:07:00,028 --> 00:07:01,393 And reaches the bottom in three days 223 00:07:01,828 --> 00:07:02,732 Things move faster 224 00:07:03,107 --> 00:07:05,648 But turning back becomes even more costly 225 00:07:05,963 --> 00:07:07,708 There is something else many people miss 226 00:07:08,362 --> 00:07:09,722 The code AI writes for you 227 00:07:09,722 --> 00:07:10,628 May appear to work 228 00:07:11,042 --> 00:07:13,128 While hiding serious risks underneath 229 00:07:13,345 --> 00:07:13,828 Why 230 00:07:14,282 --> 00:07:16,808 Because AI writes code to make it work 231 00:07:16,842 --> 00:07:17,888 Not to make it good 232 00:07:18,083 --> 00:07:19,402 It will not think for you about 233 00:07:19,402 --> 00:07:22,367 How to change extend and maintain it later 234 00:07:22,602 --> 00:07:24,128 What you build in three days 235 00:07:24,162 --> 00:07:26,607 May start rotting inside on day four 236 00:07:26,683 --> 00:07:27,727 But you cannot see it 237 00:07:27,883 --> 00:07:30,167 You see a working product 238 00:07:30,203 --> 00:07:33,131 Not the cracks in its foundation 239 00:07:33,527 --> 00:07:34,352 The third trap 240 00:07:34,647 --> 00:07:35,912 In plain language 241 00:07:36,246 --> 00:07:38,371 All the growth you see is fake 242 00:07:38,566 --> 00:07:39,191 What does that mean 243 00:07:39,566 --> 00:07:40,871 You build an AI tool 244 00:07:41,207 --> 00:07:42,371 Spend a little on advertising 245 00:07:42,407 --> 00:07:43,611 And attract a group of users 246 00:07:43,686 --> 00:07:44,671 You check the dashboard 247 00:07:44,886 --> 00:07:46,246 People log in every day 248 00:07:46,246 --> 00:07:47,392 Page views look healthy 249 00:07:47,686 --> 00:07:49,412 And retention seems decent 250 00:07:49,566 --> 00:07:50,232 You think things look good 251 00:07:50,607 --> 00:07:52,051 The product has found its market 252 00:07:52,207 --> 00:07:53,511 But you miss one thing 253 00:07:53,847 --> 00:07:56,285 These users did not come to use your product 254 00:07:56,287 --> 00:07:58,486 They came to see what AI can do 255 00:07:58,486 --> 00:07:59,912 Their motivation is curiosity 256 00:08:00,327 --> 00:08:01,292 Not need 257 00:08:01,407 --> 00:08:02,431 Once their curiosity is satisfied 258 00:08:02,686 --> 00:08:04,456 They will never return 259 00:08:05,011 --> 00:08:06,677 How many people around the world 260 00:08:06,932 --> 00:08:08,196 Are curious about AI today 261 00:08:08,891 --> 00:08:09,817 Billions 262 00:08:10,211 --> 00:08:12,497 Almost any product labeled AI 263 00:08:12,612 --> 00:08:15,456 Can attract a wave of traffic at launch 264 00:08:15,812 --> 00:08:16,970 But that traffic 265 00:08:16,971 --> 00:08:19,817 Has nothing to do with product quality 266 00:08:20,052 --> 00:08:22,656 It comes only from the word AI 267 00:08:22,971 --> 00:08:25,096 This kind of false early data 268 00:08:25,172 --> 00:08:27,136 Is more dangerous than no data 269 00:08:27,531 --> 00:08:28,516 Because with no data 270 00:08:28,852 --> 00:08:30,857 You know you have not started yet 271 00:08:31,131 --> 00:08:32,117 With false data 272 00:08:32,492 --> 00:08:34,656 You think you are already moving 273 00:08:35,011 --> 00:08:36,276 Your hands are tight on the wheel 274 00:08:36,612 --> 00:08:38,296 And the accelerator is to the floor 275 00:08:38,572 --> 00:08:39,676 But the road ahead 276 00:08:40,011 --> 00:08:41,016 Leads nowhere 277 00:08:41,944 --> 00:08:42,728 After hearing all these problems 278 00:08:43,184 --> 00:08:43,929 You may ask 279 00:08:44,343 --> 00:08:46,248 What is the real bottleneck 280 00:08:46,743 --> 00:08:47,789 I thought about this for a long time 281 00:08:48,064 --> 00:08:50,246 Then I remembered an old story 282 00:08:50,682 --> 00:08:52,407 When the Industrial Revolution began 283 00:08:52,522 --> 00:08:54,927 Steam engines made production incredibly easy 284 00:08:55,282 --> 00:08:57,527 A craftsperson could make three chairs a day 285 00:08:57,601 --> 00:08:59,927 A machine could make three hundred 286 00:09:00,121 --> 00:09:02,366 So everyone started making chairs 287 00:09:02,704 --> 00:09:03,186 And what happened 288 00:09:03,721 --> 00:09:05,366 There were too many chairs to sell 289 00:09:05,601 --> 00:09:06,547 Not because they were bad 290 00:09:06,882 --> 00:09:08,434 But because there were too many 291 00:09:08,789 --> 00:09:09,987 The ones who survived 292 00:09:09,988 --> 00:09:12,113 Were not the factories that built fastest 293 00:09:12,189 --> 00:09:14,514 But the people who understood one thing 294 00:09:14,748 --> 00:09:15,814 They understood that 295 00:09:16,149 --> 00:09:17,833 Consumers did not lack chairs 296 00:09:17,948 --> 00:09:19,068 What they lacked 297 00:09:19,069 --> 00:09:21,434 Was a chair they would gladly use for ten years 298 00:09:21,708 --> 00:09:22,373 Between being able to build 299 00:09:22,748 --> 00:09:23,674 And building the right thing 300 00:09:23,948 --> 00:09:25,573 The gap is not technology 301 00:09:25,828 --> 00:09:26,873 It is judgment 302 00:09:27,108 --> 00:09:28,507 The AI era today 303 00:09:28,508 --> 00:09:30,428 And that story from two centuries ago 304 00:09:30,429 --> 00:09:32,154 Follow exactly the same logic 305 00:09:32,468 --> 00:09:35,394 The bottleneck has moved from our hands to our eyes 306 00:09:35,748 --> 00:09:36,694 From whether you can build it 307 00:09:37,189 --> 00:09:39,178 To whether you should build it 308 00:09:39,453 --> 00:09:41,598 So what exactly is this judgment 309 00:09:42,134 --> 00:09:44,212 I do not think it is knowledge 310 00:09:44,213 --> 00:09:45,619 Or a skill 311 00:09:45,973 --> 00:09:47,858 It is more like a form of perception 312 00:09:48,294 --> 00:09:51,218 An understanding of real people in the real world 313 00:09:51,614 --> 00:09:53,298 It is knowing what a mother needs most 314 00:09:53,374 --> 00:09:56,119 In her kitchen at six thirty in the morning 315 00:09:56,733 --> 00:09:59,132 It is knowing what a recent graduate 316 00:09:59,134 --> 00:10:00,893 Is looking for in the first three seconds 317 00:10:00,894 --> 00:10:01,678 After opening their phone 318 00:10:02,254 --> 00:10:04,933 It is knowing what a sixty year old retired worker 319 00:10:04,933 --> 00:10:06,972 Fears most in that moment 320 00:10:06,973 --> 00:10:08,938 When he enters a hospital lobby 321 00:10:09,294 --> 00:10:11,619 You cannot find these things online 322 00:10:11,693 --> 00:10:13,259 And AI cannot give them to you 323 00:10:13,973 --> 00:10:15,039 They are found in the market 324 00:10:15,294 --> 00:10:16,338 On the subway 325 00:10:16,453 --> 00:10:18,132 And in the half sentence a stranger 326 00:10:18,134 --> 00:10:19,938 Lets slip after five minutes of casual conversation 327 00:10:20,134 --> 00:10:22,659 Without even realizing it 328 00:10:23,173 --> 00:10:25,499 They grow little by little 329 00:10:25,654 --> 00:10:27,969 As you pay close attention to life 330 00:10:28,285 --> 00:10:29,569 In all my years of investing 331 00:10:29,645 --> 00:10:31,609 The founders who ultimately succeeded 332 00:10:31,645 --> 00:10:32,829 Shared one thing 333 00:10:33,244 --> 00:10:34,564 They were not the smartest 334 00:10:34,564 --> 00:10:35,790 Or the most technically gifted 335 00:10:36,005 --> 00:10:37,650 Or the best connected 336 00:10:37,885 --> 00:10:39,569 They genuinely spent time 337 00:10:39,604 --> 00:10:42,050 Understanding the people they wanted to serve 338 00:10:42,444 --> 00:10:43,430 They observed people in stores 339 00:10:43,724 --> 00:10:44,770 Visited factories 340 00:10:44,885 --> 00:10:46,969 And spent entire afternoons in users homes 341 00:10:47,165 --> 00:10:49,843 They knew what users said they wanted 342 00:10:49,844 --> 00:10:52,130 And what users truly needed 343 00:10:52,165 --> 00:10:53,650 Could be worlds apart 344 00:10:54,324 --> 00:10:57,324 That understanding does not come from research reports 345 00:10:57,324 --> 00:10:59,736 And certainly not from AI 346 00:11:00,130 --> 00:11:02,535 One passage in the guide stayed with me 347 00:11:02,971 --> 00:11:03,316 It says 348 00:11:03,571 --> 00:11:04,476 In the AI era 349 00:11:04,770 --> 00:11:06,295 The founders role has changed 350 00:11:06,491 --> 00:11:08,329 Founders used to be hands on builders 351 00:11:08,331 --> 00:11:10,576 Writing code and building products themselves 352 00:11:10,770 --> 00:11:13,015 Now a founder is more like a conductor 353 00:11:13,290 --> 00:11:15,610 You do not have to play every instrument 354 00:11:15,611 --> 00:11:16,809 But you must know 355 00:11:16,811 --> 00:11:18,895 What the piece should sound like 356 00:11:19,130 --> 00:11:20,689 Which melody is right 357 00:11:20,691 --> 00:11:22,456 And which rhythm works 358 00:11:22,691 --> 00:11:23,356 That judgment 359 00:11:23,611 --> 00:11:25,416 Must come from you 360 00:11:26,010 --> 00:11:28,656 You can have AI write every line of code 361 00:11:28,770 --> 00:11:29,975 Create every image 362 00:11:30,171 --> 00:11:31,375 And arrange every layout 363 00:11:31,931 --> 00:11:33,895 But you cannot let AI decide for you 364 00:11:33,931 --> 00:11:36,656 Whether the product should exist at all 365 00:11:37,051 --> 00:11:40,336 That is what moved me most about this guide 366 00:11:40,650 --> 00:11:44,375 It was written by one of the worlds best AI companies 367 00:11:44,650 --> 00:11:45,956 Yet its message throughout 368 00:11:46,250 --> 00:11:47,895 Is not about how powerful AI is 369 00:11:48,010 --> 00:11:50,536 But about what AI cannot do for you 370 00:11:50,890 --> 00:11:52,375 It cannot judge for you 371 00:11:52,411 --> 00:11:54,816 Whether a need is real or imagined 372 00:11:55,171 --> 00:11:57,975 It cannot quit for you when it is time to quit 373 00:11:58,331 --> 00:11:59,850 It cannot see for you 374 00:11:59,850 --> 00:12:02,056 The needs that no one has voiced 375 00:12:02,610 --> 00:12:03,570 What can be built 376 00:12:03,571 --> 00:12:06,135 Is a question AI has already answered 377 00:12:07,211 --> 00:12:08,249 What should be built 378 00:12:08,250 --> 00:12:11,095 Is a question only you can ever answer 379 00:12:12,126 --> 00:12:14,050 The final chapter contains a line 380 00:12:14,205 --> 00:12:15,991 That roughly means this in Chinese 381 00:12:16,645 --> 00:12:18,924 The bottleneck is no longer what you can build 382 00:12:18,925 --> 00:12:20,690 But what you choose to build 383 00:12:20,925 --> 00:12:22,190 The first time you hear it you may think 384 00:12:22,405 --> 00:12:22,591 Well 385 00:12:22,805 --> 00:12:23,650 That makes sense 386 00:12:23,726 --> 00:12:25,365 But if you are building a company 387 00:12:25,366 --> 00:12:27,251 Or considering it 388 00:12:27,325 --> 00:12:29,832 I suggest you think carefully about that line 389 00:12:30,026 --> 00:12:32,391 Because it is not only about startups 390 00:12:32,707 --> 00:12:34,811 Is life not the same for all of us 391 00:12:35,306 --> 00:12:36,452 You can do many things 392 00:12:36,747 --> 00:12:37,992 But time is limited 393 00:12:38,026 --> 00:12:38,992 And energy is limited 394 00:12:39,227 --> 00:12:40,671 Have you seriously considered 395 00:12:40,707 --> 00:12:42,751 Among everything you can do 396 00:12:42,786 --> 00:12:44,612 What you truly should do 397 00:12:45,026 --> 00:12:46,226 I have seen too many people 398 00:12:46,227 --> 00:12:48,425 Spend their time on what they can do 399 00:12:48,426 --> 00:12:50,072 Instead of what they should do 400 00:12:50,186 --> 00:12:50,931 They stay busy their whole lives 401 00:12:51,067 --> 00:12:52,031 Then look back 402 00:12:52,067 --> 00:12:53,811 And see that most of it was filling holes 403 00:12:54,146 --> 00:12:55,852 Holes they dug themselves 404 00:12:56,206 --> 00:12:57,471 Sometimes I think 405 00:12:57,806 --> 00:12:58,811 Learning to say no 406 00:12:58,926 --> 00:13:01,532 May be the most underrated ability of our time 407 00:13:01,806 --> 00:13:03,032 Say no to a feature 408 00:13:03,446 --> 00:13:05,132 Say no to an idea that looks cool 409 00:13:05,166 --> 00:13:07,612 But that nobody actually needs 410 00:13:07,727 --> 00:13:08,831 Say no to a trend everyone else is chasing 411 00:13:09,166 --> 00:13:11,532 Even when you feel you should follow it 412 00:13:11,806 --> 00:13:12,872 The more powerful the tool 413 00:13:13,127 --> 00:13:15,967 The clearer the person using it must be 414 00:13:16,281 --> 00:13:19,047 AI is one of the greatest tools of our age 415 00:13:19,322 --> 00:13:20,920 But the sharper the tool 416 00:13:20,922 --> 00:13:23,126 The more you must know where to cut 417 00:13:23,442 --> 00:13:24,687 With a blunt knife 418 00:13:24,721 --> 00:13:26,721 Cutting in the wrong place is no great disaster 419 00:13:26,721 --> 00:13:27,807 Because it will not cut anyway 420 00:13:28,042 --> 00:13:30,800 But if your blade can slice through iron like mud 421 00:13:30,802 --> 00:13:33,759 You had better think before you act 422 00:13:34,277 --> 00:13:35,910 In the age of AI what do you think 423 00:13:36,024 --> 00:13:38,152 Is the most essential human ability 424 00:13:38,643 --> 00:13:39,779 Is it learning to use AI 425 00:13:40,349 --> 00:13:42,457 Or when AI can do almost anything 426 00:13:42,611 --> 00:13:43,707 Knowing what should not be done 427 00:13:44,396 --> 00:13:45,096 I am Leo Wang 428 00:13:45,349 --> 00:13:46,385 See you next time