Transcriptly

How to Build a Product that Hits PMF on Day 1 | Granola, Christopher Pedregal

EO16:01Added Aug 31, 2026

Meet Christopher Pedregal, the Co-founder and CEO of Granola.Granola is an AI note taking service that is being called "one of the most beloved AI products i...

Watch on YouTube →
Contributed by 刘嘉琪

Transcript

Transcript format

  1. Intro

  2. 00:00We had product market fit and we had it
  3. 00:01on day one of launch. I think the first
  4. 00:03lesson is like talking to users is the
  5. 00:05lifeblood of the built-in product. This
  6. 00:07happens all the time. You're building
  7. 00:09something, you're like, "This is great."
  8. 00:10And then you put it in front of users
  9. 00:12and then they don't understand it. For
  10. 00:13example, there's this amazing video on
  11. 00:16YouTube, a product designer doing a user
  12. 00:18test and you see the user grab the
  13. 00:20square and put it into the square hole.
  14. 00:22The designer's like, "Yes." And then you
  15. 00:24see the user grab the circle and put it
  16. 00:26in the square hole and the user's like,
  17. 00:27"What?" And then you just see like
  18. 00:29basically completely misunderstanding
  19. 00:31the product in front of them. If it
  20. 00:32takes a month to to get feedback on it,
  21. 00:34it almost not worth doing because the
  22. 00:37thinking you had when you made the
  23. 00:38original decisions, you don't even
  24. 00:40remember. So I think the short cycle
  25. 00:42there is super important. What I don't
  26. 00:43believe is that you should talk to users
  27. 00:45and then just do whatever they ask you
  28. 00:46to do, right? Cuz users are going to say
  29. 00:49a whole lot of things and sometimes
  30. 00:51they'll contradict. Sometimes what users
  31. 00:53think they want and what they actually
  32. 00:54want or what they actually need are
  33. 00:56different. I think if you just make a
  34. 00:57prioritized list of user needs and build
  35. 00:59those, that doesn't work very well.
  36. 01:11Hi, I'm Chris Pedrick. I'm the
  37. 01:14co-founder and CEO of Granola. Granola
  38. 01:16is an AI notepad for meetings. Uh, think
  39. 01:19of it like Apple Notes except it will
  40. 01:21listen in the background and when your
  41. 01:23meeting ends, it'll take whatever crappy
  42. 01:25or raw notes you've written and flesh
  43. 01:27them out into great notes. We launched
  44. 01:28Granola a year and a half ago. We were
  45. 01:31four people at the time. We're now 35
  46. 01:34people. We raised our series B at a $250
  47. 01:38million valuation and we're growing
  48. 01:41super quickly.

  49. Build the User's Best Friend

  50. 01:46I really love building tools that let
  51. 01:48people do things. This idea of oh, if I
  52. 01:50make this product great and people use
  53. 01:52it, it means they get value out of it
  54. 01:53and their lives are perhaps a little bit
  55. 01:55better as a result. Like that's
  56. 01:56something that I've always uh found
  57. 01:57really motivating. I was born in '86 and
  58. 02:00Google like the search engine launched
  59. 02:02in '98. So I was 12 years old when
  60. 02:04Google launched. I was just fascinated.
  61. 02:06I lived in a small town in the middle of
  62. 02:09the woods in upstate New York. So the
  63. 02:11internet was like this window onto the
  64. 02:13world. I kind of looked at California
  65. 02:15and Silicon Valley and Stanford and
  66. 02:17Google and all those company. I looked
  67. 02:19at them as like a you know up on an
  68. 02:20altar. How do they do this? So when I
  69. 02:23went to university, I kind of studied I
  70. 02:24want to understand how do computers
  71. 02:26work? How does the internet work? How do
  72. 02:27you build things for the internet? So I
  73. 02:28studed computer science. Like there are
  74. 02:29these very key moments that happened to
  75. 02:31me in college where I was like oh
  76. 02:32interesting that that speaks to me. And
  77. 02:34like one I was um in a security class
  78. 02:37and I learned that hey cryptography is
  79. 02:39basically magic but most almost all
  80. 02:41security vulnerabilities come from
  81. 02:43people doing the wrong thing or systems
  82. 02:45not being designed in a way that makes
  83. 02:46sense to people and I was like oh
  84. 02:48interesting even something as technical
  85. 02:49as security the human interaction
  86. 02:51element is huge and then later in my
  87. 02:53journey I discovered this field called
  88. 02:55human computer interaction and it's
  89. 02:57basically the study of how do people use
  90. 03:00technology and how do you design
  91. 03:01technology for people like if you want
  92. 03:03to build a really really fantastic
  93. 03:05product, you really really need to
  94. 03:07understand your user in that moment.
  95. 03:09This is something I talk about a lot
  96. 03:10internally at Granola that Granola as a
  97. 03:13product, the product should feel like it
  98. 03:14has soul. When we interact with a
  99. 03:17product, like if I pick up this mug and
  100. 03:19I hold it and I interact with it, it's a
  101. 03:21little bit like interacting with a
  102. 03:22person. What I mean by building a
  103. 03:24product with soul is that that whole
  104. 03:27product should feel coherent,
  105. 03:29consistent, like it's coming from the
  106. 03:30same place. It's coming from the same
  107. 03:33set of values. And I think it's very
  108. 03:36easy to tell if a product has that and
  109. 03:39it's very hard to build. The reason it's
  110. 03:41hard to build is because the bigger the
  111. 03:43more complex a product gets, the more
  112. 03:45people are involved in the making of the
  113. 03:47product. And when lots of people are
  114. 03:49involved in making it, they're all going
  115. 03:51to have different values, different
  116. 03:52viewpoints for different ideas. If
  117. 03:54they're not aligned in a very deep way,
  118. 03:57every screen of the app is going to feel
  119. 03:58very different. You know, there's that
  120. 04:00that famous thing which is if you look
  121. 04:02at a product and you can tell the
  122. 04:03organizational structure of the company
  123. 04:05that built it by the design of the
  124. 04:07product and that's really really bad.
  125. 04:09Take your your best friend. Imagine your
  126. 04:11best friend, right? You can kind of
  127. 04:12imagine what your best friend is going
  128. 04:15to do in most circumstances. If you make
  129. 04:17fun of them or you make a joke or you
  130. 04:19praise them, you're kind of going to
  131. 04:20know if they're going to be graceful or
  132. 04:22blush or make fun of you back or start
  133. 04:24crying. You kind of understand what they
  134. 04:26stand for, what they like, what their
  135. 04:27values are. We build these mental
  136. 04:29pictures of people. If you start talking
  137. 04:31to your best friend and all of a sudden
  138. 04:32they start acting like your principal,
  139. 04:34you know, or your boss, it's like you're
  140. 04:36you don't really know what to do with
  141. 04:37that. And we do the same thing with
  142. 04:38products. What I mean is like a product
  143. 04:40with soul is the exact opposite. It's
  144. 04:42not about the company structure. It's
  145. 04:44coming from a much deeper place in terms
  146. 04:46of of values and personality.
  147. 04:50I'm Taylor Rottwell, the founder and CEO
  148. 04:52of Laravel. Our mission is to help
  149. 04:55developers build, deploy, and monitor
  150. 04:57applications consistently and
  151. 04:58beautifully.
  152. 05:00Starting projects is great, but more
  153. 05:02importantly, you need to ship them.
  154. 05:03Great ideas mean nothing until they
  155. 05:05ship.
  156. 05:07That's exactly why I built Laravel, to
  157. 05:09bring your entire journey from idea to
  158. 05:11production into one seamless flow.
  159. 05:14It's more than a PHP framework. It's a
  160. 05:16complete connected ecosystem built for
  161. 05:18speed and confidence. We provide
  162. 05:21elegant, opinionated solutions for
  163. 05:23everything your modern web app needs.
  164. 05:25And you can deploy in under a minute
  165. 05:27with Laravel cloud straight from Git.
  166. 05:29Zero config needed. Laravel now powers
  167. 05:33nearly 600,000 live websites, helping
  168. 05:35developers worldwide ship their ideas
  169. 05:37faster.
  170. 05:39Build with elegance. Deploy with
  171. 05:41confidence. Monitor with precision.
  172. 05:44Stop configuring. Start shipping. Visit
  173. 05:46laravel.com today.

  174. The 50% Rule: Cut your product in half before Day 1

  175. 05:54My whole philosophy is that there's a
  176. 05:56explore and exploit. I think a huge
  177. 05:58advantage for us when we started
  178. 06:00building Granola is that we didn't
  179. 06:02launch it for a year. And what that
  180. 06:04meant is that during that year as we
  181. 06:06were exploring things and trying new
  182. 06:07things, we had a lot more freedom to
  183. 06:10change the product drastically. I'd say
  184. 06:12for the first 9 months, 10 months, we
  185. 06:15were mostly adding things, adding new
  186. 06:17features, trying to improve them, adding
  187. 06:19new like screens, what have you. And
  188. 06:21then at one point, we kind of realized,
  189. 06:23okay, here's the thing that's going to
  190. 06:25be most important, most useful to users.
  191. 06:27And we went and we cut about 50% of what
  192. 06:29we had built. That would have been
  193. 06:31really hard to do if we had been
  194. 06:33launched and we had lots of users. Like,
  195. 06:35it'd be very hard for us today in one
  196. 06:37fell swoop cut half of the product. I
  197. 06:39think our users would really yell at us.
  198. 06:41That's a big part of the the way we
  199. 06:42build. So for any product that you're
  200. 06:44building, first you have to go through
  201. 06:46this explore phase where you're figuring
  202. 06:47out what's the shape of the solution.
  203. 06:50And there you want to try as many
  204. 06:51different things as possible. And once
  205. 06:53you kind of know what the shape is, then
  206. 06:55you go into a different mode. We call it
  207. 06:57exploit internally where you just
  208. 06:59polish, right? You take the rough shape
  209. 07:00and you polish, polish, polish, polish.
  210. 07:02And here's where the details matter. But
  211. 07:04you need to make sure you're polishing
  212. 07:05the right shape because if you're
  213. 07:06polishing the wrong shape, then all of
  214. 07:08that is throwing. The question is kind
  215. 07:09of like do we hit something that's
  216. 07:11valuable and now it's time to polish or
  217. 07:13do we should we still be looking for a
  218. 07:14new shape, right? The answer is like it
  219. 07:17is really hard to know if you have the
  220. 07:21right shape of a product. It's really
  221. 07:23hard. When we launched Granola, I didn't
  222. 07:26think we had it at that point. I was
  223. 07:28like, I think this is good enough where
  224. 07:29we'll learn more by giving it to more
  225. 07:31people. That's why we decided to launch
  226. 07:33when we did cuz we said like, are we
  227. 07:34going to learn faster by continuing to
  228. 07:36do this? are going to learn faster by
  229. 07:38launching it and getting feedback from
  230. 07:39lots of users. And up until that point,
  231. 07:41it was really clear to me that we would
  232. 07:43learn faster by just onboarding two,
  233. 07:45three people every day and getting
  234. 07:46feedback. And at some point once we
  235. 07:48started doing that and we started
  236. 07:49hearing the same things over and over
  237. 07:51and we kind of understood those users,
  238. 07:52we said, "Okay, now let's give it to
  239. 07:53lots of people because maybe there are
  240. 07:55people who want to use it for use case
  241. 07:57we've never thought of." And the fact is
  242. 07:59we had product market fit on day one and
  243. 08:01we didn't realize it. And one of my
  244. 08:02biggest mistakes was not noticing it. It
  245. 08:04took me six months to realize that we
  246. 08:06had product market fit and we had it on
  247. 08:07day one of launch. The first lesson is
  248. 08:10like it is really hard to know when you
  249. 08:11have it. I heard the story that the
  250. 08:13Facebook team they had this other idea
  251. 08:15which is kind of like Dropbox for music
  252. 08:18some kind of file sharing thing and they
  253. 08:20worked on it I think for the first 6
  254. 08:21months or 9 months after they launched
  255. 08:23Facebook because they were like oh I
  256. 08:25don't know if there's a there there you
  257. 08:26know we might want to work on this other
  258. 08:27thing more and that's Facebook right
  259. 08:29that's like the generational company of
  260. 08:31the decade and they still didn't know
  261. 08:33necessarily if they were on to something
  262. 08:35huge so what I got wrong I think a lot
  263. 08:37of early founders get wrong a lot of
  264. 08:39people will sit down and say I'm going
  265. 08:41to build a product here's the painoint
  266. 08:42I'm going to solve. I want to build a
  267. 08:43great solution and it's all about can I
  268. 08:45execute on that. Have you ever played
  269. 08:46those video games where you land on a
  270. 08:49new level and there's like a little map
  271. 08:51in the corner and it's all gray and then
  272. 08:53you need to kind of go and walk around
  273. 08:55and then the map kind of unblur. But
  274. 08:57when you start off basically you have no
  275. 08:59idea of what the terrain around you
  276. 09:00looks like. I think that's the right
  277. 09:02metaphor. The right solution is
  278. 09:04unknowable until you go out and you try
  279. 09:07it and it gets contact with the world.
  280. 09:09You can't sit and design the perfect
  281. 09:10thing. You need to like put stuff out
  282. 09:12there, probe the system, and see how the
  283. 09:14system probes back.
  284. 09:20I very much believe that talking to
  285. 09:22users is the lifeblood of building
  286. 09:24product. I think if you're not
  287. 09:26constantly talking to users, not
  288. 09:27constantly getting feedback, you're just
  289. 09:28not going to build something really

  290. Systemize the Feedback Loop

  291. 09:29good. You need to systematize your
  292. 09:32ability to talk to users. If the
  293. 09:34activation energy for you to go and talk
  294. 09:37to your user is high, then you're just
  295. 09:40not going to do it very often. And if
  296. 09:42you can lower that activation energy
  297. 09:44necessary to go talk to a user, and if
  298. 09:47you can do that for your whole company,
  299. 09:48great things are going to happen. So
  300. 09:50this was 2013. I started edtech company,
  301. 09:54an AI tutor for high school kids called
  302. 09:56Socratic in New York City. And I uh I
  303. 09:59ran that for 5 years. And then Google
  304. 10:02acquired that company. It was really
  305. 10:04hard for us to actually go and talk to
  306. 10:06high school students. I remember we'd
  307. 10:07know that there were a few colleges and
  308. 10:09a few high schools around our office in
  309. 10:11New York and I would sometimes go after
  310. 10:14school and be like, "Hey, do you want to
  311. 10:16answer some questions?" And it felt
  312. 10:17super creepy. Eventually, it took us a
  313. 10:19while to get there, but we figured out
  314. 10:21this system where on every Tuesday and
  315. 10:23Thursday, we would have I think it was
  316. 10:26like five to eight high school students
  317. 10:28come into the Socratic office and spend
  318. 10:30the afternoon. We didn't have anything
  319. 10:31to talk to them about. that's fine. They
  320. 10:33would just sit there, do their homework,
  321. 10:34and leave, and they would get paid for
  322. 10:35it, and they were super happy. But if we
  323. 10:37were working on a new feature or if we
  324. 10:39were working on messaging or whatever it
  325. 10:41was, they would be there, and we could
  326. 10:43show them whatever we were doing. And
  327. 10:45the moment we had that, our ability to
  328. 10:48improve our product and make it better
  329. 10:50sped up dramatically. A big difference
  330. 10:52at Granola versus Socratic is that um we
  331. 10:54can talk to users remotely, like over
  332. 10:56Zoom. And that actually makes a lot of
  333. 10:57sense because our product's meant to be
  334. 10:59used during meetings, right? So we'll do
  335. 11:01a lot of video calls with with users. So
  336. 11:03we we usually have standing user
  337. 11:04interviews 4 days a week that anybody at
  338. 11:06the company can join and whoever is
  339. 11:08working on stuff can ask for
  340. 11:09information. So that's something that
  341. 11:11I've taken away from the Socratic
  342. 11:13experience and have kind of carried with
  343. 11:14me since. The second thing is in my

  344. Stop Collecting Redundant Feedback

  345. 11:17opinion like the secret with user
  346. 11:18interviews is you don't need to hear the
  347. 11:20same thing from 10 people. If I put a
  348. 11:22prototype in front of you and you say
  349. 11:25this button's super confusing and I look
  350. 11:27at it and I'm like oh I totally
  351. 11:28understand why you're confused. I don't
  352. 11:30need 10 other people to tell me that. I
  353. 11:31should go and I should change that
  354. 11:32button immediately so that next time I
  355. 11:35show it to somebody I learn what the
  356. 11:37next problem is. And I think this is
  357. 11:39something that especially in big
  358. 11:40companies that's unheard of. It's very
  359. 11:42easy to delude yourself. Be very very

  360. Be Skeptical, Critical, and Honest

  361. 11:45skeptical and critical and honest about
  362. 11:49if what you've built is actually
  363. 11:50something people want and they're
  364. 11:52actually going to use. So like in a user
  365. 11:54interview, never ever ask someone if
  366. 11:56they would use something. Or you can ask
  367. 11:58it but then completely ignore the
  368. 11:59answer. If you're putting them in front
  369. 12:00of a prototype, ask them like what would
  370. 12:02you do next? And then if they say
  371. 12:04something, say, "Is that true?" Or like,
  372. 12:05"Don't you think you'd be tired?" Or,
  373. 12:07you know, like really probe on it from a
  374. 12:09negative perspective because that'll I
  375. 12:10think get you to the reality faster.
  376. 12:13That's just human nature. In a user
  377. 12:14interview, you're going to try to say
  378. 12:15nice things. So I categorically ignore
  379. 12:17all positive things that people say and

  380. Use Conservative Metrics

  381. 12:20use really conservative
  382. 12:23metrics of engagement. Internally, when
  383. 12:26we say a user, it's only a user who's
  384. 12:28done at least one meeting, new meeting
  385. 12:31on this day, and that meeting has to
  386. 12:33have over 5 minutes of transcription.
  387. 12:35Otherwise, we don't count you as a user.
  388. 12:37So, if you use Granola yesterday, we
  389. 12:39don't count you as a user today. If you
  390. 12:41open up Granola and looked at 10
  391. 12:42meetings, but you didn't do a new
  392. 12:44meeting, we don't count you as a user.
  393. 12:45And we basically from the get-go, we've
  394. 12:47always had very conservative metrics for
  395. 12:49what's activity, so that we kind of we
  396. 12:52don't kid ourselves. It's so so easy to
  397. 12:54come up with excuses or say like, "Oh,
  398. 12:56you know, it's actually because of this
  399. 12:58other thing that we're going to fix.
  400. 12:59It'll be fine. Like, the product's
  401. 13:00actually good enough." What I don't
  402. 13:01believe is that you should talk to users
  403. 13:03and then just do whatever they ask you
  404. 13:04to do. Because users are going to say a
  405. 13:06whole lot of things and sometimes
  406. 13:08they'll contradict. Sometimes what users
  407. 13:11think they want and what they actually
  408. 13:12want or what they actually need are
  409. 13:14different. So, my philosophy is to talk
  410. 13:18to users so you can really get their
  411. 13:20context. you can hold that in your

  412. Trust Your Intuition Over Feedback

  413. 13:21brain, but then use your intuition and
  414. 13:25your vision of what the product should
  415. 13:26do. So, from the get- go, I think we've
  416. 13:29always known that we wanted Granola to
  417. 13:31be a very simple, minimal design product
  418. 13:34where it feels nice to spend time in.
  419. 13:37It's not distracting. It's not vying for
  420. 13:39your attention. That's been rooted on
  421. 13:41what we wanted in a product, right? Cuz
  422. 13:43we use Granola and we've always wanted
  423. 13:44to use Granola. But as we try to build
  424. 13:46that out, we have hundreds and hundreds
  425. 13:48of users kind of in our heads that we
  426. 13:50can kind of close our eyes and be like,
  427. 13:51"Oh, what would Nancy think of this?" We
  428. 13:53can kind of visualize the response
  429. 13:54pretty well. And I think that's super
  430. 13:55important. I think if you just make a
  431. 13:57prioritized list of user needs and build
  432. 13:59those, that doesn't work very well. Like
  433. 14:01whenever we've done that, users never
  434. 14:03ended up using those features very much.
  435. 14:05Like this happens all the time where
  436. 14:07you're building something, you're like,
  437. 14:08"This is great." And then you put it in
  438. 14:09front of users and then they don't
  439. 14:11understand it, they hate it, they don't
  440. 14:13get it. There's this amazing video on
  441. 14:16YouTube, a product designer doing a user
  442. 14:18test, and you see the user grab the
  443. 14:19square and put it into the square hole.
  444. 14:22The designer's like, "Yes." And then you
  445. 14:23see the user grab the circle and put it
  446. 14:25in the square hole, and the user's like,
  447. 14:27"What?" And then you just see like
  448. 14:28basically completely misunderstanding
  449. 14:30the product in front of them. I think
  450. 14:32the only way you can build an intuition

  451. Shorten Feedback Loops

  452. 14:35is by putting stuff in front of people
  453. 14:37and getting feedback on it quickly. If
  454. 14:39it takes a month to to get feedback on
  455. 14:41it, it almost not worth doing because
  456. 14:44the thinking you had when you made the
  457. 14:45original decisions, you don't even
  458. 14:47remember. So I think the short cycle
  459. 14:49there is super important. When we
  460. 14:50started building Granola, we took a very
  461. 14:53very iterative approach. Like we built
  462. 14:56the bare minimum thing and then we tried
  463. 14:58to get people to use it and see all the
  464. 15:00reasons why it didn't work and then we
  465. 15:01try to fix those and then we try to
  466. 15:03change it. And I think your intuitions
  467. 15:04do get better but you can never stop
  468. 15:07doing it. That's the other It's like
  469. 15:08their intuitions get better and then you
  470. 15:10stop talking to users and your
  471. 15:11confidence level still stays high. So
  472. 15:12you're like, "Oh, I understand users. I
  473. 15:14know what I can what what I need to
  474. 15:15build." And then when you start user
  475. 15:16testing again, you realize you're
  476. 15:18totally off pieced. And I think two
  477. 15:20different paths you can take when you're
  478. 15:22building products on top of AI. You can
  479. 15:23build a product that will basically
  480. 15:24replace the human or you can try to
  481. 15:27build a product that's going to augment
  482. 15:29or enhance what the human does. That's
  483. 15:32basically where we're going with
  484. 15:33Granola. So we're starting with meaning
  485. 15:34notes. We're going to help you with all
  486. 15:36the work you do after a meeting. Whether
  487. 15:38that is writing follow-up emails,
  488. 15:39whether that's drafting memo, we're
  489. 15:41going to help you prepare for your
  490. 15:42previous meetings. We're going to help
  491. 15:43you do analysis across all your
  492. 15:45meetings. We're going to help you do
  493. 15:47more and more and more and hopefully
  494. 15:48make make your life a little bit better
  495. 15:50every day.