YouTube2Text

Ask Claude These 5 Questions, You'll NEVER Waste Time Again — Transcript

by Austin Marchese · 5,074 words · 741 segments · language en · Watch on YouTube

Full transcript

  1. 0:00I just interviewed Ryan O'Yang, an AI
  2. 0:01engineer at Whop, a startup worth over
  3. 0:03$5 billion, and he said something that
  4. 0:05changed how I work.
  5. 0:06>> We've helped millions of people use AI,
  6. 0:08and the fastest builders have all
  7. 0:10answered these five questions.
  8. 0:11>> So I had Ryan ask me these five
  9. 0:13questions, and it turns out if you use
  10. 0:14Claude to answer these simple questions,
  11. 0:16you will build 10 times faster. So in
  12. 0:18this video, I'll break down the exact
  13. 0:20questions he asked me, how they change
  14. 0:22what I do, and then the exact prompts
  15. 0:24you can run to answer these questions
  16. 0:25yourself. And at the end of the video,
  17. 0:27I'll share what Ryan said is the most
  18. 0:28common problem people hit after they
  19. 0:30answer these questions.
  20. 0:31>> I'm Ryan O'Yang, and this is the age of
  21. 0:34AI.
  22. 0:34>> So the first question, what's your
  23. 0:36context and what are your desired
  24. 0:37outcomes? Claude has PhD-level general
  25. 0:39knowledge, but it knows nothing about
  26. 0:41your specific situation, your goals,
  27. 0:43your limitations, your restrictions, and
  28. 0:45what winning specifically looks like for
  29. 0:47you. So I asked Ryan how Whop, which is
  30. 0:48one of the fastest growing startups on
  31. 0:50the planet, answers these questions
  32. 0:51internally so that they can build
  33. 0:53faster.
  34. 0:53>> You can think of Whop as a business that
  35. 0:55helps
  36. 0:56you run your business. If I give Claude
  37. 1:00the correct context about what Whop is
  38. 1:02and where I want Whop to expand,
  39. 1:05it's it helps me get crystal clear on
  40. 1:08the steps I need to like get there. So
  41. 1:10for example, we
  42. 1:13have all these different products at
  43. 1:15Whop, and we saw that more and more
  44. 1:17agents were trying to use our products.
  45. 1:19And by giving Claude that context, it
  46. 1:21can suggest things like, oh, what it
  47. 1:23sounds like you need to do is build a
  48. 1:25really easy to use CLI.
  49. 1:28>> So I want to break down what he just
  50. 1:29described. The context provided to
  51. 1:30Claude didn't just make the answer
  52. 1:32better. It pointed Claude at the right
  53. 1:34problem to look at. And in Ryan's case,
  54. 1:36it was about making something that's
  55. 1:37easy for AI agents to interact with. And
  56. 1:40getting this answer right sets the
  57. 1:41foundation for everything that we're
  58. 1:42going to cover in this video. In the age
  59. 1:44of AI, we need to build things that
  60. 1:45solve real problems because that's how
  61. 1:47you provide value. The fastest builders
  62. 1:49don't do the most things. They make the
  63. 1:51biggest impact. And Ryan, he helps
  64. 1:53millions of people use Claude plugins,
  65. 1:55so his context is different from mine
  66. 1:57and yours. And that's totally fine. The
  67. 1:59point isn't to copy context. It's to
  68. 2:01write down your own. So to establish
  69. 2:03your context, to make sure you're
  70. 2:04actually building the right things, what
  71. 2:05exactly should go into it?
  72. 2:07>> The first thing that you want to do is
  73. 2:09tell Claude about your current situation
  74. 2:11as accurately as possible. So
  75. 2:14what does your business do? Who does it
  76. 2:16target? Where are you located? How much
  77. 2:18effort are you putting into your
  78. 2:19business?
  79. 2:20>> This context informs the direction. And
  80. 2:22notice the one thing that seems
  81. 2:24unnecessary. Where are you located? It
  82. 2:26might feel irrelevant, right? But your
  83. 2:27location, your skills, and how much time
  84. 2:29you have, this all influences what the
  85. 2:31right answer is for your situation. And
  86. 2:33so the second half of this first
  87. 2:35question is your desired outcome. The
  88. 2:36goal itself is usually easy, right? I
  89. 2:38want more customers. I want to save
  90. 2:39time. But the hard part and the most
  91. 2:41effective builders make it concrete. And
  92. 2:43the second you give Claude the correct
  93. 2:44context, it can better turn that
  94. 2:46abstract goal into one number and one
  95. 2:49deadline. On screen is an exact prompt
  96. 2:50that will force you to provide context
  97. 2:52and then drive you to a desired outcome
  98. 2:54based on these goals. And when
  99. 2:55describing your desired outcome, make
  100. 2:57sure to level it up. So in Ryan's case,
  101. 2:59it wasn't make an AI agent tool. It was
  102. 3:01help our customers interact with our
  103. 3:03tools using AI. If you go a level lower,
  104. 3:06a task gets optimized as an isolated
  105. 3:08job. But when you layer it up, Claude
  106. 3:10can make more strategic decisions. So
  107. 3:12when I was working with Ryan to answer
  108. 3:13this question myself and I provided the
  109. 3:15context around buildpartner.ai, my tool,
  110. 3:17we use the WAP CLI, which you can
  111. 3:18install by running this command. But
  112. 3:20this effectively allows you to use WAP
  113. 3:22tools directly in Claude. And then once
  114. 3:24it's set up, because I have WAP pulling
  115. 3:25my website analytics, it can see all of
  116. 3:27my website visitors. And because WAP has
  117. 3:30industry data from top performing
  118. 3:31websites on its platform, it didn't just
  119. 3:33hand me a number. It was comparing my
  120. 3:35current product to real benchmarks. And
  121. 3:38so that's how I can pull the desired
  122. 3:39outcome not out of thin air, but based
  123. 3:41on real numbers. And when doing this and
  124. 3:43trying to answer this question on what
  125. 3:44my context and desired outcome was, it
  126. 3:46became clear that even though I wanted
  127. 3:49more monthly recurring revenue, the
  128. 3:50problem I needed to solve was improving
  129. 3:52my conversion rate. So once we provided
  130. 3:54Claude with the right context to make a
  131. 3:56clear desired outcome, what should you
  132. 3:58actually build to get there? Brings us
  133. 3:59to the second question. What's the most
  134. 4:01important thing to build? And this is
  135. 4:03about defining the MVP. So this question
  136. 4:05actually has two parts. The first is
  137. 4:06find what only you can do. And then
  138. 4:08second, find the minimum steps to one
  139. 4:11real result. So to help us answer this,
  140. 4:13I actually asked Ryan how they think
  141. 4:14about it at Whop. And listen to how he
  142. 4:16describes your unique ability or your
  143. 4:18secret sauce. And then at the end of the
  144. 4:19clip, I'll describe how you can identify
  145. 4:21your own secret sauce.
  146. 4:22>> What we spend our time doing is
  147. 4:24fruitlessly prioritizing
  148. 4:27what
  149. 4:28is actually unique to our business. Like
  150. 4:31what can Whop do that no one else can
  151. 4:34do? And that's usually like the most
  152. 4:36most most most high leverage thing to
  153. 4:38focus on. And for us, it's really about
  154. 4:41our ecosystem. The millisecond that you
  155. 4:43get a payment on Whop, you can spend it
  156. 4:45on a Whop ad or you can spend it on your
  157. 4:47Whop card to maybe pay rent for your
  158. 4:50office or pay a contractor or, you know,
  159. 4:53buy something that helps you grow your
  160. 4:54business even faster. I think like the
  161. 4:56web of things that you can do after you
  162. 4:58make money on Whop is incredibly
  163. 5:00important and is part of our secret
  164. 5:01sauce.
  165. 5:02>> So Whop's secret sauce is this
  166. 5:03ecosystem. But how do you find your
  167. 5:05secret sauce? So this is a bit abstract,
  168. 5:07but your sauce is your unique ability.
  169. 5:09That is the hard thing that you can do
  170. 5:11that others can't.
  171. 5:12>> Your location, your skills, how much
  172. 5:15time you have to spend on the business,
  173. 5:17like these are all things that make you
  174. 5:18unique. For any operator, for any
  175. 5:20executive, for any just person who wants
  176. 5:23to be more efficient, you should just
  177. 5:24ask yourself questions of like, what do
  178. 5:26I like to do? What am I good at? And
  179. 5:29that you usually get to your sauce
  180. 5:31pretty quickly.
  181. 5:31>> A good way to think this is actually
  182. 5:33there's a perfect example in Silicon
  183. 5:34Valley. In the pilot episode, Richard's
  184. 5:36Pied Piper is a forgettable music
  185. 5:38copyright app. Then Hooli and Peter
  186. 5:41Gregory spot the compression algorithm
  187. 5:43hitting inside it, and then a bidding
  188. 5:45war starts. The value was actually in
  189. 5:47the one thing that only Richard could
  190. 5:48build. That was his unique sauce. Now
  191. 5:50once you've identified this unique sauce
  192. 5:52for your situation, there's actually two
  193. 5:54ways to implement this depending on what
  194. 5:56you're working on. And so the first is
  195. 5:57streamline and automate what isn't your
  196. 5:59unique ability. And this allows you to
  197. 6:00focus your human energy on what makes
  198. 6:03you unique, in turn helping you save
  199. 6:05time on the other things. And the second
  200. 6:07is double down on what makes you unique
  201. 6:09and figure out ways to enhance that
  202. 6:10unique ability to in turn increasing the
  203. 6:12value you provide. For either one, this
  204. 6:14unique ability can serve as a measuring
  205. 6:16stick. So right, are you working on your
  206. 6:18unique ability? If no, how can I speed
  207. 6:20up these tasks? If yes, how can I double
  208. 6:23down on this? This is how you build
  209. 6:24faster, not by doing more, but by
  210. 6:26pointing your time at the things that
  211. 6:28deliver value faster. Now whichever
  212. 6:30bucket you fall into, your focus needs
  213. 6:32to be on creating your minimum viable
  214. 6:33product. And so I asked Ryan how he
  215. 6:35thinks about creating an MVP and he told
  216. 6:37me about a car detailing business he
  217. 6:38started using WAP's own tools.
  218. 6:41>> Let's just start really dumb, right?
  219. 6:42Like what do I need to do to get my
  220. 6:45first customer? Well, I have to talk to
  221. 6:47someone who has a dirty car. Simplest,
  222. 6:50simplest, simplest, I'm going to put
  223. 6:51some flyers on people's cars. And then
  224. 6:53how am I going to deliver the service?
  225. 6:55Well,
  226. 6:56I guess I have to buy a pressure washer.
  227. 6:58I guess I have to buy soap. I guess I
  228. 6:59need to buy like a wash mitt and one of
  229. 7:02those sponge thingies.
  230. 7:04Boom, now I have a to-do list of like
  231. 7:06things I need to buy from Home Depot and
  232. 7:09what I have to print and what I have to
  233. 7:11do to get my first customer. And I would
  234. 7:13say that is literally the MVP of a car
  235. 7:15wash business.
  236. 7:16>> So with this in mind, you can ask
  237. 7:17yourself, what do I actually need at the
  238. 7:19bare minimum to get one result? That's
  239. 7:21the core of MVP thinking. Start with the
  240. 7:24result you want, work backwards and cut
  241. 7:26everything that isn't necessary to get
  242. 7:28there. And I do want to call out that
  243. 7:29this isn't just for new businesses or
  244. 7:31new products. MVP thinking works on
  245. 7:33everything you're already doing.
  246. 7:34>> If your metric is saving time,
  247. 7:37well, then you ask yourself questions
  248. 7:39about like where am I wasting time?
  249. 7:40>> This is just a general process that you
  250. 7:43can constantly do no matter what state
  251. 7:45you're in.
  252. 7:45>> Exactly. Like we help people run their
  253. 7:47business, but you could just as easily,
  254. 7:49you know, run this for um
  255. 7:53your job, your process, even like your
  256. 7:55health, you know. I think this applies
  257. 7:57for almost any goal that you have.
  258. 7:59>> So that's Ryan describing it, but here's
  259. 8:00how you can put it into practice. First,
  260. 8:02take the metric you chose in question
  261. 8:04one and list the actions that could move
  262. 8:06it. Then narrow that list down to the
  263. 8:08minimum steps needed to test one
  264. 8:10high-leverage outcome. Now, I'll dive
  265. 8:12into this more in the next question Ryan
  266. 8:14asked me, but by cutting scope on
  267. 8:16anything you're building, you're
  268. 8:17focusing on the high-signal components
  269. 8:19that make this valuable. And honestly,
  270. 8:21the hard part isn't doing this. It's the
  271. 8:23strategy of knowing which move is worth
  272. 8:25your time. So how do I think about this
  273. 8:27strategically? Well, when looking at my
  274. 8:28tool, buildpartner.ai, I have my whole
  275. 8:30product hooked up in Wapp, so it sees
  276. 8:32all the website traffic. I can then use
  277. 8:34Wapp's economic intelligence tool to
  278. 8:36help me determine high-leverage tasks I
  279. 8:38can complete for my business. And
  280. 8:40because Wapp has the data on what's
  281. 8:42working across businesses on its
  282. 8:44platform, it can point at the highest
  283. 8:46leverage move better than almost anybody
  284. 8:48because it has access to this unique
  285. 8:50data. So that way, it'll recommend
  286. 8:51something concrete like fixing a landing
  287. 8:53page or starting ads. Then I'll ask to
  288. 8:55create a list of minimum steps to test
  289. 8:57it. Now, if you're starting a business
  290. 8:58or you're scaling a business, if you
  291. 9:00don't have Wapp configured, you can use
  292. 9:01the first link below to sign up, which I
  293. 9:03am affiliated with transparently. But
  294. 9:05either way, here is a prompt to help you
  295. 9:07strategically answer this question about
  296. 9:09what is the most important thing to
  297. 9:10build. So at this point, you've provided
  298. 9:12context and your desired outcome to
  299. 9:14Clone, and you know the one thing to
  300. 9:15build for your MVP. Now, those first two
  301. 9:17questions a lot of people do, but this
  302. 9:19next question 99% of people skip, but it
  303. 9:22is more important than the first two.
  304. 9:23Question three, what are you not
  305. 9:25building? So this is the biggest problem
  306. 9:27with building with AI. It's not knowing
  307. 9:28what to build, it's knowing what not to
  308. 9:30build. So if you found what you can do
  309. 9:32uniquely, Ryan's argument is that you
  310. 9:34should spend your time on nothing else.
  311. 9:36>> Once you have your goal and once you
  312. 9:38know what your sauce is, I would say the
  313. 9:41most important question to ask is what
  314. 9:43should you not build? And this when I
  315. 9:46say like like ruthless focus on the
  316. 9:48things that are important,
  317. 9:50it it sort of automatically means you
  318. 9:52shouldn't be focusing on things that are
  319. 9:54not important.
  320. 9:55>> And he is right, right? These other
  321. 9:56things can just be so distracting
  322. 9:58because building is so easy now that
  323. 10:00everything can feel like it's worth
  324. 10:02building. And here's him talking about
  325. 10:03this building addiction loop.
  326. 10:04>> Get addicted to building everything, you
  327. 10:08you
  328. 10:08Everything is shiny. You know, I as
  329. 10:10programmer, I would love to build my own
  330. 10:12database and like, oh yeah, dude. Like
  331. 10:14look at my cool database console and
  332. 10:16it's got all these great colors and like
  333. 10:18I'm like benchmarking performance. Like
  334. 10:20oh I can save like,
  335. 10:21you know, a bunch of milliseconds here
  336. 10:24on this database query, but like who
  337. 10:25cares? Like it it doesn't matter. Um
  338. 10:29and I should know that it doesn't matter
  339. 10:31honestly at the beginning because it
  340. 10:33stops me from going down these rabbit
  341. 10:35holes that um honestly caused me to lose
  342. 10:38sight of the original focus and the goal
  343. 10:40that I had.
  344. 10:40>> And the keyword he says here that I want
  345. 10:42you to focus on is but who cares? It
  346. 10:45doesn't matter. Oh, you build a tool
  347. 10:46that's 80% as good as a tool we already
  348. 10:48use. Great. It doesn't matter. Oh, you
  349. 10:50build something that doesn't actually
  350. 10:51provide value. Great. Oh, you build
  351. 10:53something that's really cool looking,
  352. 10:54but it doesn't save me time or help me
  353. 10:56make more money. Great. It doesn't
  354. 10:57matter. And so one real benefit here of
  355. 10:59what you are not building is it makes it
  356. 11:01super clear what you shouldn't do. Your
  357. 11:04job as the manager of Claude is to
  358. 11:06connect the pieces, not to build all of
  359. 11:08the pieces. And if you ignore that,
  360. 11:10you'll end up with an AI tool graveyard
  361. 11:11where you'll have 10 20 half used tools,
  362. 11:14none of them actually doing anything.
  363. 11:15And so for Build Partner, there is just
  364. 11:16a bunch of things that I've elected that
  365. 11:18I will not build. Do I want to vibe code
  366. 11:20a payment platform that has to figure
  367. 11:22out how to accept payments in 195
  368. 11:25countries? No. Instead, I just tell
  369. 11:27Claude use WAP to set up payment for
  370. 11:29this product and it just does it and now
  371. 11:30I can get payment without building
  372. 11:32anything. Do I want to figure out how to
  373. 11:33run ads across TikTok, Instagram,
  374. 11:35YouTube, Facebook, and Google Search?
  375. 11:37No, I'm not an ads expert. So I'll just
  376. 11:39use WAP CLI from Claude and tell it I
  377. 11:41want to configure ads for these
  378. 11:43platforms and it just does it. I don't
  379. 11:45have to think about setting up accounts
  380. 11:46on every channel, optimizing CPMs. I'm
  381. 11:48leaning on WAP, the tool to do it. And
  382. 11:50so the point is if you want to build
  383. 11:51faster, you can't focus on things that
  384. 11:53don't matter and that you shouldn't
  385. 11:55build yourself. Now, there is one word
  386. 11:57of caution here when exploring the use
  387. 11:59other tools framework. And here's Ryan
  388. 12:01addressing this exact issue that you may
  389. 12:03run into once you get tool trigger
  390. 12:05happy.
  391. 12:05>> I don't want to manage 10 tools. My
  392. 12:07agent doesn't even want to manage 10
  393. 12:08tools. It's about maximizing value per
  394. 12:10tool
  395. 12:12versus maximizing the number of tools.
  396. 12:14>> So the goal is to look for tools with
  397. 12:15robust functionality to cut down on what
  398. 12:18you need to build. Obviously here we're
  399. 12:20talking about Wapp, but I chose them
  400. 12:22because I use them for payments, help me
  401. 12:23run ads, business intelligence, and my
  402. 12:25website analytics. I do not want to
  403. 12:27build any of those, and I wanted a
  404. 12:28single tool that does all of these
  405. 12:30things. So to help you answer this
  406. 12:32question, run this prompt, which will
  407. 12:33look at what you're building and what
  408. 12:34you spent your time on in the past, and
  409. 12:36then help you identify what you should
  410. 12:37not be building. Once you've done that,
  411. 12:39run this prompt, which will help you
  412. 12:40identify any external tools that can
  413. 12:42help you with what you're trying to do.
  414. 12:44At this point, we've answered three of
  415. 12:45the five questions. Now before we get to
  416. 12:47the last two, and then a word of caution
  417. 12:49from Ryan about what you need to do
  418. 12:51after answering these five questions.
  419. 12:53You guys all know the drill. This is our
  420. 12:55anti-slop agreement. Everything in this
  421. 12:57video is built by humans, for humans. So
  422. 12:59all that I ask as part of this agreement
  423. 13:01is you subscribe to help this content
  424. 13:02reach more people. And if you already
  425. 13:04subscribed, make sure to turn on post
  426. 13:05notifications. That really helps as
  427. 13:06well. And as a thank you to everyone, I
  428. 13:08appreciate all of you. I give away a
  429. 13:10Claude Max subscription every video. So
  430. 13:11this video's winner is T Carson 5045,
  431. 13:14who's building agents for narrow AEC
  432. 13:17pain points. And to enter this video's
  433. 13:19giveaway, comment below with what you're
  434. 13:20building or what you're deciding not to
  435. 13:22build. And a fun fact here that I want
  436. 13:24to share with all of you. Ryan and I
  437. 13:25actually built our last startup
  438. 13:26together, IYK, which raised over $30
  439. 13:28million. So for everyone building,
  440. 13:30double down on your human connections
  441. 13:32because you never know where your career
  442. 13:34is going to go. But anyway, we've gone
  443. 13:36through three of the five questions.
  444. 13:38Question four is what are your building
  445. 13:39hot zones? Hot zones are critical parts
  446. 13:42of your work that you need to get right.
  447. 13:43For everything else, just let AI carry
  448. 13:45the torch and you monitor the outcome.
  449. 13:47For me, within Build Partner, we have a
  450. 13:49custom plugin that helps people build
  451. 13:51faster with Claude. And this goes out to
  452. 13:52over 5,000 people when I update it. So
  453. 13:54this is a hot zone to say the least. So
  454. 13:57I asked Ryan, what do the most
  455. 13:58successful businesses on Wapp do when
  456. 14:00thinking about this?
  457. 14:01>> All the successful business owners that
  458. 14:03we see on Wapp are really dialed on
  459. 14:06making sure that um they're a they're
  460. 14:09really working with their AI to spend a
  461. 14:11lot of time on the things that like
  462. 14:14really are the most important to their
  463. 14:15business.
  464. 14:16>> But from there, he does add one really
  465. 14:18important caution this. You can't just
  466. 14:20let AI run free. These hot zones are
  467. 14:22areas of what you're building to be
  468. 14:24extra focused on. On screen you can see
  469. 14:26a prompt that will help you identify
  470. 14:27what your hot zones are. And then what I
  471. 14:29like to do is I I add in my claw.md
  472. 14:31anytime I'm working with a hot zone,
  473. 14:33flag it with me and make sure I approve.
  474. 14:35And so again, for Bill partner, these
  475. 14:36hot zones and how I think about it is
  476. 14:38the plugin as well as our WAP payment
  477. 14:40flow. I want to make sure that the
  478. 14:41payments is working properly. But this
  479. 14:43is just part of it, right? How do you
  480. 14:45actually work in the hot zones
  481. 14:46themselves? A term that you've likely
  482. 14:48heard is eval, which is short for
  483. 14:50evaluations. And this is the process of
  484. 14:52testing your changes to make sure it's
  485. 14:54getting better, not worse. The fastest
  486. 14:56builders know when to measure twice and
  487. 14:58cut once. And in hot zones, it makes
  488. 15:00sense to establish evaluation criteria
  489. 15:02when needed. Evals are definitely a
  490. 15:04large topic and I might have to make a
  491. 15:06full video on this, but it's important
  492. 15:07that you have a strong foundational
  493. 15:09understanding of this process. So what
  494. 15:11should you do today to implement your
  495. 15:12first eval? This is what I call building
  496. 15:14a minimum viable eval and here's Ryan
  497. 15:16talking about how to do this.
  498. 15:18>> Claw's really good at making evals
  499. 15:20nowadays. Um
  500. 15:22everyone's is going to look different,
  501. 15:24but um it's generally going to be
  502. 15:28your input, which is the thing that
  503. 15:31you're changing. In my case, it's the
  504. 15:32skill file. There's going to be some
  505. 15:34program that claw helps you write, which
  506. 15:37like runs all of these call it tests
  507. 15:39that does the squats for you. And then
  508. 15:41at the end, you get a report, which is
  509. 15:43how your input did against the suite of
  510. 15:48tests that you ran against it.
  511. 15:50>> So from that, think of it like a simple
  512. 15:51equation. You have your input and your
  513. 15:53output. This is the whole grading
  514. 15:54mechanism. And then you have what you're
  515. 15:56changing. This is the variable that
  516. 15:58you're adjusting to see what the output
  517. 16:00becomes. And you may be thinking this is
  518. 16:02just for technical work, but it isn't.
  519. 16:03In our conversation, we covered two
  520. 16:05kinds of evals. The first is a standard
  521. 16:07eval, which uses a programmatic approach
  522. 16:09to evaluate quickly through testing
  523. 16:11cycles. This is typically when you hear
  524. 16:13about people talking about evals, they
  525. 16:14run loops and they automatically improve
  526. 16:16themselves. In order for that to work,
  527. 16:18it has to be extremely programmatic. The
  528. 16:20second slow evals are for non-technical
  529. 16:22work, and this involves more human
  530. 16:23judgment for molding the evaluation
  531. 16:25criteria. So to put a real-world example
  532. 16:27of this, I did a slow eval for a skill
  533. 16:29that I built for an engineering firm
  534. 16:31that I worked with. Let's call the skill
  535. 16:32/generate report. And so we would use
  536. 16:34the skill to generate the report. We
  537. 16:36would take that output, we would get
  538. 16:38feedback from the engineering firm that
  539. 16:39we worked with, and then we would adjust
  540. 16:40the skill until the output was approved
  541. 16:42by them, and then we would store that
  542. 16:44approved version. Then the next time we
  543. 16:45would run the report and they would
  544. 16:47provide feedback, we would change the
  545. 16:49skill, but we would rerun it on all the
  546. 16:51past approved reports to make sure that
  547. 16:53nothing breaks and it doesn't regress
  548. 16:54for what those previous edge cases were.
  549. 16:56And so you can imagine over time that we
  550. 16:58build a test suite for these reports and
  551. 17:00it can run this evaluation criteria over
  552. 17:02time to get better and better. And this
  553. 17:03process isn't instantaneously, but this
  554. 17:05slow eval process can really make a
  555. 17:07difference. Now to identify any specific
  556. 17:09hot zones that are worth building a
  557. 17:11minimum viable eval for, use this prompt
  558. 17:13which will help you get started. Now the
  559. 17:14last question that Ryan asked me is what
  560. 17:16I think is the most underrated. Question
  561. 17:19five, what are your bottlenecks? And you
  562. 17:21may be wondering, I am not sure where my
  563. 17:23bottlenecks are. And so Ryan thinks
  564. 17:24about it as something that you identify
  565. 17:26as you run towards your goals.
  566. 17:28>> Let's say you are trying to scale your
  567. 17:30revenue
  568. 17:32and you get your conversion rate to a
  569. 17:33good place,
  570. 17:35your bottleneck is how many people
  571. 17:36actually land on your website in the
  572. 17:37first place. You're converting all of
  573. 17:39them, but like no one ever lands there,
  574. 17:40right? Or you could have the opposite
  575. 17:42problem. Oh, so many people land on your
  576. 17:43website, but none of them check out.
  577. 17:45>> What?
  578. 17:46>> Like your bottleneck is your conversion
  579. 17:47rate.
  580. 17:48>> And what he discussed here is why you do
  581. 17:49questions one through four first. That's
  582. 17:51so that we know we're moving in the
  583. 17:52right direction so that these
  584. 17:54bottlenecks are relevant to our desired
  585. 17:56outcome. But the biggest transformations
  586. 17:57are when you find the real bottleneck
  587. 17:59and then suddenly one fix can just
  588. 18:01change everything. So how do you
  589. 18:02actually spot your bottleneck? Often
  590. 18:04it's the things that you actually avoid.
  591. 18:06>> Everything that people don't enjoy doing
  592. 18:08eventually becomes a bottleneck. And
  593. 18:10I'll give you a good example, like a lot
  594. 18:12of business owners, they start out being
  595. 18:14employees. Maybe you started off as a
  596. 18:17website designer
  597. 18:19and you've made this killer website and
  598. 18:20you just love making websites, and you
  599. 18:22keep improving your website. Well,
  600. 18:25you're eventually going to be
  601. 18:26bottlenecked by the quality of your
  602. 18:28product,
  603. 18:29or maybe how you drive people to your
  604. 18:31website cuz you're not a marketer,
  605. 18:32right? But you just because you love
  606. 18:34improving your website so much, it's
  607. 18:35really hard for you to switch
  608. 18:37and like focus on these hard problems
  609. 18:39that are uncomfortable, you're
  610. 18:41unfamiliar with them.
  611. 18:42>> So a lot of the times you can identify
  612. 18:43bottlenecks through what you don't want
  613. 18:45to do or what you're not good at or
  614. 18:47looking at what you haven't spent any
  615. 18:48time on. That's the avoidance test.
  616. 18:50Procrastination is wisdom. Use this as a
  617. 18:53signal for where your bottlenecks may be
  618. 18:55located. Here's a prompt to help you
  619. 18:56identify bottlenecks and quantifiably
  620. 18:58identify what the Wp CLI can solve for
  621. 19:01you. And then if you have the Wp CLI
  622. 19:02installed, here's how you can use that
  623. 19:04to help streamline these bottlenecks.
  624. 19:06Now at this point you may have two
  625. 19:07questions. How do you know if you're
  626. 19:09working on the right bottleneck? That's
  627. 19:11where you want to think about effort
  628. 19:12versus return. If you feel like you're
  629. 19:14putting in way more energy than you're
  630. 19:15getting benefit out of, you likely
  631. 19:17aren't focused on a bottleneck. You're
  632. 19:19focused on something else. Then how do
  633. 19:21you know if you've solved a bottleneck?
  634. 19:23The best way to do this is use
  635. 19:24benchmarks. If you look at something
  636. 19:25like your conversion rate and it's some
  637. 19:27percentage below industry standard,
  638. 19:28that's a bottleneck you can still
  639. 19:30improve. But in order to do that, that
  640. 19:32requires knowing what the standard is,
  641. 19:33which is exactly where I'll use Wp's
  642. 19:35data to help me get that answer. So here
  643. 19:36are two prompts to help you. First, find
  644. 19:38your bottlenecks by what you avoid, and
  645. 19:40second, use the Wp CLI to put an actual
  646. 19:42number to this bottleneck. So after
  647. 19:44these five questions, what's the part
  648. 19:45Ryan said almost everyone gets wrong?
  649. 19:48And this is the most common problem. You
  650. 19:49only ask these questions once. The
  651. 19:51answers that you provide to all five of
  652. 19:53these questions, the reality is it has a
  653. 19:54shelf life. What worked on day one isn't
  654. 19:57enough on day 100. And you couldn't have
  655. 19:59known the right answers when you started
  656. 20:01because things change. So you have to
  657. 20:03schedule periodic reviews. Here's how Wp
  658. 20:05does it and the cadence I'd copy.
  659. 20:06>> Every quarter we find out what our
  660. 20:10biggest desired outcome is. What what
  661. 20:12are we trying to improve about the
  662. 20:14business of Wp, right? Wp is a business.
  663. 20:16We happen to help people build
  664. 20:18businesses, but we are also a business
  665. 20:19and there's different things to focus
  666. 20:21on. And then week over week we might
  667. 20:23think about, okay, what is the most
  668. 20:25important thing that's going to move us
  669. 20:26to that outcome?
  670. 20:27>> Whenever you're working on something,
  671. 20:28you can bucket your work into two
  672. 20:30categories. The first is execution and
  673. 20:32then strategy. Execution is for when you
  674. 20:34know exactly what's needed and you want
  675. 20:35to revisit this weekly if not daily.
  676. 20:37This is your list of to-dos and for me,
  677. 20:39I literally just run this on an Apple
  678. 20:41Notes every day. I have my items that I
  679. 20:43have to do. Question four and five that
  680. 20:44we covered in this video, that's where
  681. 20:46this falls. What eval are you proving?
  682. 20:48What metric did it move? What are you
  683. 20:49working on today to push you toward the
  684. 20:51desired outcome? In my eyes, execution's
  685. 20:53actually really easy. [music] The second
  686. 20:55bucket is harder, strategy. This is
  687. 20:57where you're planning what's needed. But
  688. 20:59if you do this every day, then you get
  689. 21:00stuck in strategy and you don't actually
  690. 21:02build anything. So realistically, you
  691. 21:04want to revisit this monthly or
  692. 21:05quarterly. And questions one and three
  693. 21:07are strategic. That's what's my sauce?
  694. 21:10What am I not building? The goal is to
  695. 21:11find a Goldilocks zone. Execute quickly
  696. 21:13so you avoid analysis paralysis, but
  697. 21:16make sure it's actually in the right
  698. 21:17direction. And if you skip re-asking
  699. 21:19yourself these questions, here's what
  700. 21:21can happen.
  701. 21:22>> If people don't stop and ask themselves
  702. 21:24questions about like what's happening,
  703. 21:26what they want to do, what's giving them
  704. 21:28joy, what they're good at, they would
  705. 21:29never have discovered these other
  706. 21:31opportunities and they'd be maybe like
  707. 21:34incorrectly focused on that one thing
  708. 21:36that they identified maybe like many
  709. 21:38months or even years ago.
  710. 21:40>> So what you need to do today is put a
  711. 21:41recurring event on your calendar, weekly
  712. 21:43for execution, monthly or quarterly for
  713. 21:45strategy, and don't overthink it. So ask
  714. 21:47yourself these five questions to make
  715. 21:49sure you're building the right things.
  716. 21:51First, what is your context and desired
  717. 21:53outcome? Second, what's the most
  718. 21:55important thing to build? Third, what
  719. 21:56are you not building? Fourth, what are
  720. 21:58your hot zones? And fifth, what are your
  721. 22:00bottlenecks? And then you need to ask
  722. 22:02them again on a specific schedule. Swab
  723. 22:04helped me put this all into practice,
  724. 22:06right? The data behind my numbers, the
  725. 22:07CLI that pulls them in, and the tools
  726. 22:09that mean I don't have to build
  727. 22:10payments, a website, or analytics
  728. 22:12myself. So if you want to try WAP, click
  729. 22:14the first link in the description and
  730. 22:16shout out my boy Ryan. I love having
  731. 22:17friends who are smarter than me building
  732. 22:19awesome things so that I can work
  733. 22:20together with them. Now one thing that I
  734. 22:22did discuss here is knowing the
  735. 22:23importance of when to use the right
  736. 22:25tools. WAP is one of them. But if you
  737. 22:27want a breakdown of eight free tools
  738. 22:29everyone watching this video should use,
  739. 22:31go check out this video. If you like
  740. 22:33this, I guarantee you'll love that. I'll
  741. 22:34see you over there. Peace.

About this transcript

This page contains the full transcript of Ask Claude These 5 Questions, You'll NEVER Waste Time Again by Austin Marchese, generated from the public captions YouTube serves with the video. The transcript has 5,074 words across 741 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.