AI transcript
0:00:06 So, when it comes to prompting, are there any specific tricks or tactics?
0:00:08 It surprised me. It behaves kind of like a person, right?
0:00:10 Like, it’s really valuable to give it a bunch of examples.
0:00:12 Are you going to give us all those prompts, or?
0:00:15 Oh, yeah, maybe.
0:00:18 Do you have any hot takes on where you see all of this heading?
0:00:22 The best AI features are actually smaller, more detailed AI features
0:00:25 that actually just make the core workflow easier.
0:00:29 Hey, welcome to the Next Wave podcast.
0:00:31 I’m Matt Wolf. I’m here with Nathan Lanz.
0:00:34 And today, we’ve got another amazing guest for you.
0:00:38 Today, we’re talking to the principal product manager over at Roblox.
0:00:41 He is the writer of creatoreconomy.so.
0:00:45 And today, we’re going to dive into some real actionable tips
0:00:48 that you can use to leverage a lot of the AI tools that are out there.
0:00:52 In fact, in this episode, Peter breaks down step by step
0:00:55 how to use specific prompts to get things done in your daily life.
0:00:59 Like, he actually walks us through how to type the prompts
0:01:01 and why these prompts work.
0:01:04 We’re also going to dive into all sorts of different use cases
0:01:07 and ways that you could leverage these AI tools in your business.
0:01:12 We’ve got some amazing stories that me and Peter and Nathan share
0:01:14 about how we’re using AI in your own business.
0:01:16 I really think you’re going to enjoy it.
0:01:18 So let’s go ahead and jump on in with Peter Yang.
0:01:24 When all your marketing team does is put out fires, they burn out.
0:01:28 But with HubSpot, they can achieve their best results without the stress.
0:01:31 Tap into HubSpot’s collection of AI tools,
0:01:34 breeze to pinpoint leads, capture attention,
0:01:37 and access all your data in one place.
0:01:41 Keep your marketers cool and your campaign results hotter than ever.
0:01:44 Visit hubspot.com/marketers to learn more.
0:01:50 Hey, Peter, welcome to the show.
0:01:52 Thanks so much for joining us today.
0:01:53 How are you doing?
0:01:54 Yeah, I’m doing great.
0:01:55 It’s great to hang out with you guys.
0:02:00 Let’s just dive right into this and learn a little bit about you.
0:02:06 So you’re lead of product over at Roblox and you’re an AI enthusiast.
0:02:09 You know, what we want to talk about today, I think,
0:02:13 is like some really actionable use cases for AI.
0:02:17 You seem to have figured out some really good ways to leverage AI
0:02:22 to just sort of make your life and your work easier and faster for you.
0:02:23 So I think with this episode,
0:02:26 it would be really fun to kind of lean into that
0:02:29 and talk about some of the really cool use cases
0:02:32 because there is sort of this narrative people on the peripheral of AI.
0:02:34 Maybe not the people that are like in it using it every day,
0:02:38 but the people seeing all of the AI news going,
0:02:39 “Eh, it’s not really useful yet.
0:02:41 Like, what can AI do for me?”
0:02:44 Those of us are using AI every day to see the power,
0:02:46 but I think a lot of people are still kind of going,
0:02:49 “I don’t know how to use this in my daily life yet.”
0:02:53 So let’s start as like from a content creator perspective.
0:02:57 What are some of the ways that you’re leveraging AI as a content creator?
0:02:59 I use it for a lot of things.
0:03:02 So I have this doc called like my AI writing assistant.
0:03:07 It’s just a Google doc and I save all my long prompts there, right?
0:03:09 So let me read you some of them.
0:03:15 So I have an AI prompt for making like a clickbaity LinkedIn post.
0:03:19 I have AI prompt for generally YouTube thumbnails.
0:03:24 Like, you know, like I have AI prompts for editing my newsletter.
0:03:27 I also do these interviews, these podcast interviews,
0:03:30 and it’s kind of like a pain to ask to clean up the transcript afterwards.
0:03:32 So I have an AI prompt for that.
0:03:34 Are you going to give us all those prompts or?
0:03:36 Well, yeah, maybe.
0:03:38 If you buy my fifth-hour prompt package.
0:03:39 Your course.
0:03:41 That’s kidding.
0:03:44 I also use AI to kind of like give me personal advice,
0:03:47 like, you know, kind of like act like a personal coach.
0:03:51 I give a bunch of information about myself and then just ask it for advice.
0:03:52 I’d love to hear some of the prompts.
0:03:55 I mean, you mentioned one for helping with like titles and thumbnails.
0:03:57 I mean, I do a lot on YouTube.
0:03:58 The next wave is on YouTube.
0:04:01 Like, can you give us a sneak peek at some of these prompts?
0:04:05 So basically the prompt is like, you’re an expert copywriter,
0:04:09 you know, your goals to generate 10 YouTube thumbnails and title combinations
0:04:12 based on this transcript that I’ll share with you next, right?
0:04:15 And then each combo should be two separate lines each.
0:04:16 And here’s some instructions.
0:04:18 So common patterns.
0:04:20 Create a curiosity gap.
0:04:22 Use numbers to make claims more compelling.
0:04:26 Use provocative or contrarian short quotes to grab attention.
0:04:30 Imply insider knowledge is being revealed.
0:04:32 Highlight impressive transformation.
0:04:34 Use YouTube to speak directly to the viewer.
0:04:39 And then I include a list of all the best thumbnail and title combinations.
0:04:43 Right? Like both from my own personal channel and other channels.
0:04:44 I mean, that’s really cool.
0:04:46 When you do enter these prompts, where are you going?
0:04:48 Is it are you using chat GBT, Claude, Lama?
0:04:50 Like what’s your go to?
0:04:52 Oh, I have a turn from chat GBT.
0:04:53 I stopped using it.
0:04:55 Yeah. So where are you going?
0:04:57 So I primarily use Claude now.
0:04:57 Yeah.
0:04:58 I use Claude every day.
0:05:01 Sometimes I use perplexity, but mostly Claude.
0:05:01 Yeah.
0:05:02 I love Claude.
0:05:05 I actually really, really love the projects feature
0:05:06 because some of these things that you’re talking about,
0:05:10 you can just kind of save them as a project and then use them over and over again.
0:05:12 Like I do something like pretty similar.
0:05:17 I basically made a giant spreadsheet of like all of my favorite YouTube headlines
0:05:18 that I’ve come across.
0:05:21 I’ve actually had a Google sheet going of all my favorite YouTube headlines
0:05:26 for like a couple of years now, just as like a sort of swipe file of headlines.
0:05:29 And I was able to take all of that list of headlines,
0:05:34 throw it into Claude and let Claude sort of analyze what headlines,
0:05:37 like what are the patterns in these headlines that seem to make them work?
0:05:41 And then Claude is really good at generating new headlines
0:05:43 based on that information, it seems.
0:05:44 Yeah, it’s fantastic.
0:05:47 I mean, unfortunately, I’ve noticed a pattern where like, you know,
0:05:50 you basically, it seems like people on YouTube really want to make money.
0:05:53 So you have to mention like this is like a billion dollar opportunity
0:05:57 or like, you know, I wish I knew this earlier.
0:05:59 So sometimes it gets a little bit too click baby for my taste,
0:06:02 but yeah, it’s still useful.
0:06:05 Oh yeah, I mean, like as a YouTuber myself,
0:06:10 that’s like one of the balances I’m constantly trying to figure out is like,
0:06:14 how do I make a title that like grabs attention so people want to click on it?
0:06:17 I want to bait the click, but I don’t want to be too hypey
0:06:19 where people click on it and go, wait a second,
0:06:22 I feel duped based on what, you know, what that title was.
0:06:25 That feels like one of the hardest balances.
0:06:26 That’s the same thing on Twitter and LinkedIn too, right?
0:06:29 Like I’ve tested like doing threads or it’s like,
0:06:31 I don’t want to be over hyping something.
0:06:34 I’ve tested a thread like that and then tested one where like I super hype things
0:06:36 up or mention money or whatever.
0:06:38 And then like they do way better every single time.
0:06:44 So when it comes to prompting, like, are there any specific like tricks
0:06:49 or tactics or tips you give to like actually get decent prompts?
0:06:54 A couple of things that I try to do is I try to get it to proceed step by step, right?
0:06:58 So like, I think it’s really valuable to give it a bunch of examples
0:07:00 to make it more personal.
0:07:04 Like it’s actually incredibly bad as just generating random shit from scratch.
0:07:06 Like you gotta give it as many examples as you can.
0:07:10 And then what I have to do is like, I have to do two steps.
0:07:12 So first look at all the examples that I give you
0:07:17 and like give me a list of bullet points of like some common themes from these examples.
0:07:22 And then go ahead and edit my thumbnails or edit my new set of posts
0:07:26 or you know, edit my stuff and it seems to do pretty well with that.
0:07:27 Have you given me like negative examples?
0:07:29 Like here’s something that’s bad as well.
0:07:30 That’s what I’ve tried doing.
0:07:31 It seems to work pretty well.
0:07:32 Like here’s something that works well.
0:07:34 Here’s like a format I like.
0:07:36 Here’s something that I hate.
0:07:39 For some reason, it likes to use semi-colon a lot.
0:07:42 So I tell them to like stop using semi-colon for fricassex.
0:07:46 Yeah, stop delving into everything.
0:07:47 Yeah, stop delving.
0:07:51 Yeah, there’s like, there’s like definitely sort of common words
0:07:55 that when you let it just write for you, you’re like, all right, I can like,
0:07:56 I see it on Twitter all the time.
0:07:58 I see it in blog posts all the time.
0:08:03 It’s like, I feel like those of us that are in AI now that pay close attention
0:08:07 can spot chat GPT written stuff just based on the language it uses.
0:08:13 Yeah, you mentioned that like you always tell it to think through step by step.
0:08:15 What’s the, what’s the logic there?
0:08:18 Like what’s the reasoning you’re telling it to think step by step?
0:08:20 Like shouldn’t an AI just do that naturally?
0:08:22 It’s like, it surprised me.
0:08:23 It behaves kind of like a person, right?
0:08:30 Like if it’s able to kind of summarize or like do a step first and
0:08:33 then do another step, it turns out that it does it better.
0:08:35 Then we try to do it, make it do everything at once.
0:08:38 And also if you like flatter it a little bit and like, you know,
0:08:42 you really get this or like, you know, you got to have confidence in yourself.
0:08:45 Give me a really, give me like an 11 out of 10 answer.
0:08:48 Like you can do it and then it will like try harder.
0:08:52 Let’s talk through like what AI is like really good at.
0:08:55 Like what would have been some of the like best use cases you’ve found
0:08:56 for some of these large language models?
0:08:59 I know you mentioned some as far as like content creation for titles
0:09:03 and thumbnails and things like that, but let’s, let’s like expand it.
0:09:06 Let’s, let’s like break open the floodgates a little bit and like share
0:09:10 with people who might be thinking like, I don’t need AI in my business.
0:09:13 I don’t need AI in my like sort of daily life.
0:09:18 What are some of the really cool use cases that can really help people
0:09:20 if they just sort of lean into it a little bit?
0:09:25 I think the most common use case so people think like AI is really good
0:09:29 at like generating stuff from scratch and I actually don’t think
0:09:30 it’s the best at that.
0:09:33 I think what’s really good at is if you have a bunch of information
0:09:37 that you just don’t want to read, it’s really going to help you
0:09:38 feel lazy, right?
0:09:41 Like you don’t want to, I don’t want to, I don’t want to read all this shit
0:09:43 or like, you know, like this whole, this whole post.
0:09:46 So I’m just going to paste this stuff in and like tell it, like give
0:09:50 you some constraints like, hey, give me the main takeaways in five lines
0:09:54 or like when I’m researching for guests for podcasts, I like to I don’t
0:09:57 like to watch their past YouTube videos because I don’t have time for that shit.
0:10:01 So what I do is I just take the transcript from that video and I paste
0:10:05 it into AI and I tell it to like give me a list of nested bullets
0:10:08 with the main takeaways and also include some quotes from the guest.
0:10:10 Yeah.
0:10:12 And then it’s basically like watching a one hour podcast without
0:10:13 actually watching it, right?
0:10:16 Like I can get the TL;DR in like five, five, five minutes.
0:10:16 Yeah.
0:10:18 I mean, hopefully people don’t use it for this podcast, but you know,
0:10:23 is there any easy way to get the transcripts with AI yet?
0:10:25 Or do you have to like manually go download them and then you like
0:10:28 you’re copying, pasting or uploading a file or?
0:10:30 I think there’s some plugins, but I just go to YouTube and like
0:10:34 copy and paste the transcript and I paste it into Claude to do it.
0:10:35 Yeah.
0:10:38 You can use tools like, um, like the script as well.
0:10:42 And then a lot of times what I’ll do too is I’ll say, you know,
0:10:46 I don’t want to ask the same questions Lex Friedman asked Elon on this
0:10:50 podcast based on this interview, what are some gaps in the conversation?
0:10:53 What are some interesting things that I can ask this guest that weren’t
0:10:55 asked on this other podcast?
0:10:59 And it actually helps find some like new topics and side chains
0:11:01 that they didn’t get into.
0:11:02 And I find that to be really valuable.
0:11:04 Yeah.
0:11:07 That that’s worth valuable because I think it’s just so much easier
0:11:12 to, to edit something like edit what AI’s questions are, like edit
0:11:15 something from AI to come up with from scratch by yourself.
0:11:16 Right?
0:11:18 So it’s like a lot easier.
0:11:21 What are some of the other ways that you’re using it to like really
0:11:23 leverage your time?
0:11:28 Um, so in my product job, we have discord or we have like these
0:11:32 uh, communities with a bunch of customers and we do like alphas and
0:11:33 betas with these customers.
0:11:36 And sometimes I’ll come back the next day and there’s like, you know,
0:11:38 so many messages to be through.
0:11:40 So, so I just like copy paste or thing.
0:11:41 I’m like, Hey, what did people talk about?
0:11:44 Or like maybe, maybe someone tags you on a slack thread.
0:11:46 That’s like 15 messages deep.
0:11:48 And they’re like, Hey, Peter, what do you think?
0:11:49 I mean, I’m not real.
0:11:53 So let me summarize this stuff.
0:11:54 Like what has been discussion so far?
0:11:55 Right.
0:11:59 Like people like on Twitter keep sharing these templates about like, oh,
0:12:02 here’s like nine and nine product spec templates they should follow.
0:12:04 Like all this crap.
0:12:06 I’m like, why even write a prospect?
0:12:08 Just get AI to write right for you.
0:12:10 Just like, I actually try this.
0:12:13 I had like a few bullet points of like a product feature that
0:12:14 I wanted to build.
0:12:17 And then I basically started some discussions internally and some
0:12:18 discussions with customers.
0:12:19 Right.
0:12:20 And they gave me a bunch of feedback.
0:12:23 And then I was like, okay, then I pasted my bullet points and all the
0:12:24 feedback into AI.
0:12:26 Like, okay, now make a prospect for me.
0:12:28 And that worked well.
0:12:29 Yeah.
0:12:29 It worked pretty well.
0:12:33 I mean, I also gave a template that I use for my specs.
0:12:35 But I mean, it’s pretty, it’s pretty, pretty good.
0:12:36 Yeah.
0:12:40 And the last thing I’ll say is like at work, at these corporate jobs,
0:12:43 it’s very important to become a very good and internal communication.
0:12:43 Right.
0:12:46 There’s people who are not good at this.
0:12:49 Like they, they are like very robust or like they write these like walls of
0:12:53 text and before I would tell them to like, okay, you know, here’s some
0:12:56 tips to, but now it’s like, Hey, why don’t you just pull your shit into AI
0:12:59 first and then it’ll just make it more concise for you.
0:13:03 There’s like no excuse not to be super worthy.
0:13:03 Yeah.
0:13:03 Yeah.
0:13:05 I do that a lot with my newsletter as well.
0:13:08 And I think, I think you mentioned you do it with your newsletter is also
0:13:12 where I’ll take my, I’ll write my newsletter that I send out each week
0:13:16 and I’ll pull it into, you know, Claude and then I’ll say, can you proofread
0:13:20 this, find any grammar mistakes, spelling mistakes, but also proofread
0:13:21 it for readability.
0:13:24 Make sure it’s as easy to read as possible for people and it will
0:13:27 actually go and trim out some of the fluff for me and clean it up.
0:13:30 But it might add some delves here and there, which I got to go back in and
0:13:34 remove, but for the most part, it does a pretty good job of like bringing
0:13:37 it down to just the meat of what I’m trying to put out.
0:13:38 Yeah.
0:13:41 It’s fantastic, especially if you gave it like your best posts from
0:13:42 the past that you’re proud of.
0:13:42 Yeah.
0:13:43 It’s really good.
0:13:44 I kind of match that style.
0:13:44 Yeah.
0:13:47 I personally, I don’t know if you found a way to do this, but personally, I
0:13:51 haven’t found a way to make AI sort of right with my voice.
0:13:55 You know, so like, I can upload five of my emails into Claude and say,
0:14:02 write it to sound like me and it will like, it might use some of the same
0:14:04 words, but for the most part, I’ll read it and it won’t sound anything
0:14:08 like, like the way I write, but like, have you found a way to, to make
0:14:09 that kind of thing work?
0:14:11 Um, not so much.
0:14:14 I think it’s better at editing my writing.
0:14:18 Like no matter, like even like a really crappy draft, it’s a lot better
0:14:21 if you give it something than, uh, if you just try to write from scratch.
0:14:24 Now, are you using any of the other like AI tools that are out there?
0:14:27 I mean, so far we’ve pretty much talked about like large language models
0:14:29 and, you know, chat bots like Claude and chat GPT.
0:14:32 Are you using other stuff outside of the LLMs?
0:14:37 So this is also why I think like the, like the AI startup thing is
0:14:40 like kind of like overhyped because, um, the tools actually come back to
0:14:44 you every week are pretty much Claude and maybe perplexity.
0:14:47 Um, and that’s it.
0:14:51 You know, the other ones, the other ones, like there’s some really cool stuff
0:14:51 out there.
0:14:51 Right.
0:14:54 There’s like sooner for generating music.
0:14:57 There’s like, uh, runway for generating videos.
0:15:01 And you know, obviously mid-majority, but like, I don’t, I don’t use it.
0:15:04 I, these are stuff that I tried is like kind of a novelty.
0:15:04 Yeah.
0:15:06 And then I kind of churn from, from it.
0:15:12 We’ll be right back, but first I want to tell you about another
0:15:14 great podcast you’re going to want to listen to.
0:15:17 It’s called Science of Scaling hosted by Mark Roberge and it’s
0:15:22 brought to you by the HubSpot podcast network, the audio destination
0:15:23 for business professionals.
0:15:28 Each week host Mark Roberge, founding chief revenue officer at HubSpot,
0:15:31 senior lecturer at Harvard Business School and co-founder of Stage 2
0:15:35 Capital sits down with the most successful sales leaders in tech to
0:15:39 learn the secrets, strategies and tactics to scaling your company’s
0:15:40 growth.
0:15:44 He recently did a great episode called, how do you solve for a siloed
0:15:45 marketing and sales?
0:15:47 And I personally learned a lot from it.
0:15:49 You’re going to want to check out the podcast.
0:15:53 Listen to Science of Scaling wherever you get your podcasts.
0:15:57 There’s one thing that I do.
0:16:00 Someone use, uh, on some of my regular basis, which is, uh, this
0:16:02 app called Super Whisper.
0:16:06 It lets me go outside and like take a walk and like talk to my phone
0:16:09 and like start a draft that way.
0:16:10 So that kind of helps a little bit.
0:16:15 So it basically, you talk to it, it transcribes everything you say and
0:16:17 then summarizes it or bull points it or.
0:16:17 Yeah.
0:16:18 It transcribes everything.
0:16:23 It does have a local LEM that uses to summarize stuff, but like, uh, the
0:16:24 LN kind of sucks.
0:16:27 So what I usually do is I just put, but I just paste the transcript into
0:16:29 cloud and tell it to make it better.
0:16:29 Yeah.
0:16:32 I mean, I use a lot of the like AI art stuff.
0:16:36 I love, I love using like mid-journey and Leonardo and tools like that for
0:16:39 my thumbnails, because for me, it’s like very meta.
0:16:40 I make YouTube videos about AI.
0:16:43 And so my thumbnails are mostly made with AI as well.
0:16:48 And it’s just like, I love that sort of like meta-ness of it at all.
0:16:50 Um, yeah, but yeah, I totally agree.
0:16:51 Like I love runway.
0:16:52 I love Suno.
0:16:57 I love playing with these toys really, but I don’t really have a lot
0:16:58 of practical use cases for them.
0:17:02 I’ve used a couple of Suno videos and like, or songs and like the background
0:17:05 of some of my videos, but it’s pretty rare that I’ll do that.
0:17:10 I think runway and Luma and that kind of stuff is really cool, but it’s not
0:17:12 quite there yet still for me to feel like.
0:17:16 I want to use this as like b-roll that looks realistic.
0:17:18 You know, like, I think it’s getting closer though.
0:17:20 I mean, like you look at a year ago, it was so horrible.
0:17:22 And now it’s like, it’s pretty good.
0:17:24 It’s not good enough yet, but it’s, you know, it’s pretty good.
0:17:26 I mean, there was that commercial.
0:17:28 Was it the Toys R Us commercial that they used?
0:17:29 Uh, yeah.
0:17:31 So, uh, AI videos.
0:17:33 So, I mean, I think we’re going to see a lot of those actually start
0:17:36 to be used and, and even though, you know, maybe the AI startups are
0:17:40 overhyped, whatever, but like, that’s kind of the nature of VC, right?
0:17:42 It’s like, it’s, you need to be doing these experiments.
0:17:45 It’s good for the industry as a whole to be creating these startups
0:17:48 and trying things and yeah, most of them will not work, but hopefully
0:17:51 they will all kind of move things forward for, you know, for everybody.
0:17:54 But, uh, you know, like the risk profile of the VC is very different
0:17:54 from the founder, right?
0:17:58 The VC can just like have one out of 100 work, but the founder,
0:18:00 if you’ve been on like some sort of AI rapper and doesn’t work,
0:18:01 then you’re kind of screwed.
0:18:01 Yeah.
0:18:01 Yeah.
0:18:02 Yeah.
0:18:04 You know, so I mean, it depends.
0:18:06 I mean, like, you know, I, I’ve done a few startups.
0:18:09 I mean, typically even if your startups fail, there’s other opportunities
0:18:11 that arise because of that.
0:18:11 Yeah.
0:18:12 I mean, not always.
0:18:15 If you do something very uninteresting and you’re, and you, and you do
0:18:18 a really poor job, then yeah, you, you might be somewhat screwed.
0:18:21 But you, but usually there’s a network of people and there’s other opportunities.
0:18:22 But yeah, anyways.
0:18:24 Uh, and I’ll say like one more thing about like, uh, cause a lot of
0:18:26 these bigger companies are also trying to build AI, right?
0:18:29 Like, like, um, not necessarily to Google and Microsofts of the
0:18:30 world.
0:18:34 And I think what I’ve noticed is that the best AI features from these
0:18:37 bigger companies, like if I already have a product that’s like people love
0:18:42 to use, the best AI features are actually like, um, smaller, more detailed
0:18:46 AI features that actually just like make the core workflow easier.
0:18:47 You know what I mean?
0:18:50 Like as opposed to actually like trying to build some sort of chatbot
0:18:55 or like agent feature, just like think about your core customer journey
0:18:59 and, and like pick the areas where AI can actually help you save a bunch of
0:19:00 time or like make things easier.
0:19:03 And those are kind of the features that work the best.
0:19:08 Um, and an example of that is, um, I talked to this, uh, prod director
0:19:13 from Shopify, his first feature was like just generating product descriptions
0:19:17 like helping the merchants generate prod descriptions for like their hundreds
0:19:19 of SKUs, you know, right?
0:19:22 And it’s not super sexy, but like it saves them a ton of time.
0:19:22 Yeah.
0:19:24 So they’re using it all the time.
0:19:26 I don’t know if you’re a gamer or not.
0:19:28 I’m Nathan and I are both gamers.
0:19:31 So we talk about games from time to time, but did you hear about the new
0:19:35 like college football game that EA put out, but they used AI to sort
0:19:37 of develop the game at rapid speed?
0:19:39 No, I am a gamer, but I haven’t heard about that yet.
0:19:43 So basically like in, you know, in the, in the NFL, right?
0:19:47 There’s hundreds of players, maybe in the thousands, but like not as many
0:19:49 players that are in college, right?
0:19:53 In college football and NCAA, there’s over 11,000 players.
0:19:57 Well, EA just released a new game with like pretty much all of the division
0:19:59 one college teams in that game.
0:20:03 And they had to figure out how do we get 11,000 players into this game?
0:20:09 So they used AI and they basically took a single headshot of every sing
0:20:14 all 11,000 players and used AI to create characters that had each
0:20:15 person’s face on it.
0:20:18 So like the build of the person may not be exact, but they were able
0:20:23 to get the face of all 11,000 players by using AI and just putting them
0:20:27 all into a model that then generated a 3D character with that person’s face.
0:20:31 And I just thought that was like an amazing use case of like, I don’t
0:20:32 really think it put anybody out of the job.
0:20:35 It just made a really, really tedious task way less tedious.
0:20:36 Yeah, exactly.
0:20:38 Those were like the perfect use cases.
0:20:40 Just like doing something that no human wants to do basically.
0:20:42 So there was a, there was an article that you wrote and I, and I
0:20:43 saw you, you tweet about it.
0:20:49 And one of the bullet points of the article was, you know, why some AI
0:20:53 startups gain traction while other AI startups fail.
0:20:54 You know, what are your thoughts on that?
0:20:58 Like why do some AI startups gain massive traction and everybody’s
0:21:01 talking about them while other ones just sort of fade into the night?
0:21:05 I think with AI startups, there’s like definitely like a hype cycle.
0:21:09 So you might get a shit ton of users when you launch or you
0:21:11 might even get like a bunch of money.
0:21:15 But like then the people might just move on to the next thing.
0:21:19 It’s like, I think retention is very difficult with these AI startups.
0:21:24 And I think a failure mode is a startups that like get a ton of users
0:21:27 and they go to like VVC’s and they’re like, okay, you know, give me
0:21:30 more money and then they have this crazy valuation.
0:21:31 Right?
0:21:31 Yeah.
0:21:34 And then they, they have this like AI product for, you know, sales
0:21:35 and marketing or something else.
0:21:39 But then like the foundational models are always getting better.
0:21:42 And then there’s like no point in using them anymore.
0:21:43 Like what’s the point?
0:21:46 And because their valuation is so crazy, it’s going to be very hard
0:21:49 to actually survive and make a, you know, continue on.
0:21:53 I mean, the ones that actually, you know, there’s this guy called
0:21:55 level CEO on Twitter, right?
0:21:57 Like he has like a headshot AI or something.
0:22:01 I don’t think he raised any of these two dollars, but he just found
0:22:05 like a really good niche use case and he’s making a ton of money from it.
0:22:09 So it’s almost better to just like bootstraps to some extent.
0:22:09 Yeah.
0:22:10 Some of the stuff.
0:22:10 Yeah.
0:22:14 I think, I think even he mentioned he’s been using more AI himself
0:22:17 versus like building an AI product like recently because he had a lot
0:22:21 of turn of like having these like, here’s like a, you know, AI for headshots
0:22:21 or whatever.
0:22:23 I know that’s one of his products and he has other ones.
0:22:27 But yeah, he said he was saying huge turn because yeah, as a technology
0:22:31 gets so good, people can very quickly copy you and make a similar service
0:22:32 and make it cheaper.
0:22:32 Right.
0:22:34 Well, you can do it for free and meta now, right?
0:22:36 Like, so meta just rolled out there.
0:22:42 Imagine me feature like last week where it looks at all of your, your headshots
0:22:48 or your pictures of you inside of your photos on Facebook and it essentially
0:22:51 trains a model and you can say, imagine me riding a dragon and it puts
0:22:52 me on a dragon, right?
0:22:58 So like tools like that just overnight seem to make companies obsolete
0:22:59 that we’re selling that service, right?
0:23:03 We’ve seen open AI do that over and over and over again too, right?
0:23:07 You have like so many tools that came out that were like summarized
0:23:09 my PDF tool or chat with my PDF tool.
0:23:14 And now you can do that in the free version of Claude and chat GPT, right?
0:23:20 There’s like, yeah, I feel like a lot of the struggles of like the AI startups
0:23:23 are that the big incumbents can just roll it out and they’ve got so
0:23:27 much capital available to them that they can do it at, you know, free
0:23:28 or practically free.
0:23:29 Yeah, exactly.
0:23:33 I mean, but then again, you know, if you bootstrap your PDF startup and
0:23:37 you make it like a million dollars and then, you know, it becomes obsolete.
0:23:39 Like that’s not a bad outcome, right?
0:23:39 Yeah, yeah, yeah.
0:23:43 You just have a window of opportunity that you got to capture, you
0:23:44 know, capture, I guess.
0:23:47 And Zuck is like, you don’t want to compete with Zuck, man.
0:23:48 He’s like unstoppable.
0:23:52 Meta is going to keep on putting out a lot of this open source stuff and
0:23:55 just, you know, for lack of a better word, democratize it all.
0:23:56 Yeah.
0:24:00 I mean, I thought openly I was untouchable, but you know, with Zuck
0:24:01 and these other players involved.
0:24:03 Yeah, he’s very strategic.
0:24:06 I remember back in the day, I had a lot of friends who were making
0:24:10 fortunes on social apps and social games on Facebook.
0:24:13 And Facebook really took advantage of that to grow at that moment.
0:24:15 And they’re like, oh, they, you know, Facebook loves us new speed
0:24:18 parties where you go hang out with Facebook employees and everything.
0:24:20 And then overnight when Facebook didn’t need that anymore.
0:24:22 Nope, you’re all gone.
0:24:23 Yeah.
0:24:26 So I would be, I would be mindful of like, you know, just, you know,
0:24:30 yeah, sure, Zuck has changed or whatever, but you know, how much
0:24:31 do people change?
0:24:32 So yeah.
0:24:33 Yeah, sure.
0:24:33 It’s great.
0:24:36 He’s doing all this open source stuff, but who knows what the real
0:24:36 objective is.
0:24:40 I mean, he’s, he’s, he’s into win-win and he’s like, you know,
0:24:44 do you think this is, this is a debate, you know, Nathan and I
0:24:46 have had and we’ve had it with other guests as well.
0:24:48 Do you think, do you think open AI is falling behind?
0:24:51 I don’t want to speculate too much, but I do notice that they
0:24:56 went from just like dropping these products to, to launching stuff
0:25:00 with like wait lists and be like, oh, I’m going to roll it out
0:25:01 slowly, slowly, slowly.
0:25:01 Yeah.
0:25:05 And maybe it’s because of a lot of drama they had, but it does
0:25:08 seem like they’ve slowed down the problem of loss here.
0:25:11 I mean, in comparison to your Anthropic, which is like shipping
0:25:12 stuff left and right.
0:25:14 I’m still of the belief that they’re so far ahead in the new models
0:25:17 coming and it’s, it’s going to make Anthropic stuff look kind of
0:25:21 silly in comparison, but there’s also the election coming up.
0:25:23 They’ve brought on people from the government who are now a part
0:25:25 of open AI apparently.
0:25:28 You know, and so I think, I think we’re going to, as soon as the
0:25:30 elections over, we’re going to see, you know, GPT five and it’s
0:25:31 going to be mind blowing.
0:25:31 Yeah.
0:25:32 Yeah.
0:25:34 They’ve also done a really great job with like partnerships
0:25:35 and marketing.
0:25:39 Uh, like, you know, I saw some stat like to, to, to, to us, like
0:25:42 in my opinion, like Claude is clearly better right now, but I
0:25:44 think way more people are still using chat GPT.
0:25:45 Oh yeah.
0:25:48 There’s like, if you go and look at like Google trends, right?
0:25:53 And you type in Claude or Anthropic Claude and you type in chat GPT.
0:25:54 It’s nowhere even close.
0:25:57 Claude just looks like a flat line down at the bottom compared
0:25:59 to the amount of people that are still searching for GPT.
0:26:00 I mean, I mean, it’s even a term, right?
0:26:02 Like in this conversation, Matt, use the term chat GPT, like
0:26:05 as a generic, you know, like for like LLIMs, right?
0:26:08 Where you could have said Claude, but you said chat GPT, you know.
0:26:08 Yeah.
0:26:08 Yeah.
0:26:12 It’s, it’s become the Kleenex of AI or the Band-Aid of AI, right?
0:26:16 Like people just use the brand name now is like a synonym for
0:26:19 the, uh, for the, uh, you know, the underlining tech.
0:26:22 But, um, you know, I, I, I don’t know.
0:26:26 I, I have a feeling that open AI isn’t as far as, uh, ahead as I
0:26:30 think Nathan believes it is one of the areas that I think we
0:26:32 kind of, um, go back and forth on a little bit.
0:26:36 Um, I, I do, I think what we’re seeing, but it’s a very sort
0:26:38 of closed bubble thing, right?
0:26:41 Like we all pay attention to AI on almost a daily basis.
0:26:45 So we’re seeing it, but I get really, really frustrated by
0:26:48 open AI announcing stuff, but then never shipping it, right?
0:26:50 Like the GPT for a voice.
0:26:51 We’re still waiting.
0:26:55 Sora, we’re still waiting, you know, search GPT.
0:26:59 Now that’s only announced last week, but nobody’s got access to it yet.
0:27:04 And, um, you know, among us AI nerds, that gets frustrating to me.
0:27:06 Um, and yeah, I don’t know.
0:27:11 I, I, I don’t know if they’re as far ahead as everybody believes
0:27:11 they are.
0:27:14 And Elon Musk claims that he’s further ahead because he’s got
0:27:16 the largest, uh, data center on the planet now.
0:27:18 And I don’t know what we’ll see.
0:27:22 I just like watching it all play out and, you know, eat my popcorn.
0:27:23 Yeah.
0:27:25 I mean, I only have 20 bucks a month to spend.
0:27:27 So whoever, whoever’s ahead, I’ll spend on.
0:27:30 You know, one of the, one of the other topics that I, I think
0:27:32 would be interesting to talk about is like, where we think
0:27:36 all of this is going, do you have any sort of like bold predictions?
0:27:40 Do you have any hot takes on, on where you see all of this heading?
0:27:43 One of the things that I’ve been talking about a lot in my product
0:27:45 stuff is like, and this company is like, people just want to
0:27:46 manage more and more people, right?
0:27:49 Like, you know, getting more headcount and manage more people is
0:27:50 seen as clout.
0:27:52 Like you have more responsibility.
0:27:54 And I think that’s just like pretty toxic and at least two
0:27:59 companies bloating and like eventually having to do layoffs.
0:27:59 Right.
0:28:00 Yeah.
0:28:00 Yeah.
0:28:04 So my hope is with AI and I think maybe dance shipper or coin is a phrase,
0:28:06 but like, you know, instead of managing people, yeah, people who
0:28:11 manage AI agents or like, you basically have individual contributors
0:28:15 and crafters who can actually get a lot of shit done and like do a
0:28:16 lot more than they can before.
0:28:19 And then you can keep the company small.
0:28:22 If you keep the company small, there’s less bureaucracy and bullshit
0:28:25 and everyone can actually just like believe in the mission and like,
0:28:27 you know, execute on a product.
0:28:32 So that has my hope in the age of AI, our product managers can be
0:28:35 more important or less because I could see it going either way.
0:28:35 Right.
0:28:38 Like, like, yeah, product managers always thought it was almost like a,
0:28:40 you know, you know, sorry to say this, but like a, you know,
0:28:43 necessary evil that you need the product managers to manage all
0:28:45 these, these, you know, individual contributors.
0:28:49 Like, I wonder like, are PMs not going to need those individual
0:28:51 contributors as much and they can go off and create their own
0:28:53 projects or features at companies?
0:28:55 I mean, product managers ask much value.
0:28:56 You know, like it’s hard for AI too.
0:29:01 Only product managers cannot be replaced.
0:29:01 Okay.
0:29:04 No, yeah, gotcha.
0:29:07 I hope, I hope that like another thing that has happened as companies
0:29:10 got bigger is like product teams have like before, which is like a
0:29:12 PM, a designer and an engineer.
0:29:14 And now there’s like so many different roles, like, you know,
0:29:18 user researcher, there’s like data analysts, there is product
0:29:22 ops, product specialists, a lot kinds of crazier is like all the agile
0:29:25 coaches kind of crap, but people will learn to wear multiple hats.
0:29:29 Like, for example, I would love to become like a hybrid designer,
0:29:33 PM, as opposed to just a PM, or maybe like engineers can design too.
0:29:37 But product managers do have one skill, which is they’re really good at,
0:29:39 you know, writing detailed requirements, right?
0:29:42 So in some ways that’s kind of like prompting AI, right?
0:29:45 Yeah, but you said the product, you said the AI can do that for you,
0:29:46 right?
0:29:48 Which means it gets you to go up and get the questions and get the spec.
0:29:52 You know, if people are more productive and then companies can do,
0:29:56 you know, do the same task with less people, where, where do you
0:29:58 think that like new jobs are created?
0:30:02 I mean, like, you know, one outcome is like AI robots would do
0:30:05 all the real work and then all the people would become like influencers
0:30:06 and try to entertain each other.
0:30:08 That’s what I’m making on.
0:30:12 That’s very possible, I think.
0:30:13 But yeah, I don’t know, man.
0:30:17 Like I feel like AI is going to start to disrupt the white color jobs.
0:30:20 And then I think once we have like robots that can do stuff in the real
0:30:23 world, that’s going to be when the real disruption happens, man.
0:30:27 Like, you know, like instead of hiring like a nanny, I can just hire a
0:30:28 robot or something.
0:30:30 And it can teach my, my, my kid at the same time.
0:30:32 It has all this context about my child and what they know when
0:30:35 their education level language.
0:30:38 So they can, they can teach Japanese and English.
0:30:39 Yeah.
0:30:39 Yeah.
0:30:40 Yeah.
0:30:40 I mean, we all have kids.
0:30:43 So it’s like, it’s hard to imagine how education will change.
0:30:46 I mean, the current system is so antiquated.
0:30:49 You know, I actually, um, a friend of mine is sending me this toy
0:30:50 called dino.
0:30:54 Uh, that has like a, it’s like a stuff that stuff dinosaur that
0:30:57 has heard of that and building or something.
0:31:00 So I’m going to try it with my daughter and see if she can ask it.
0:31:01 You know, she can learn from, from, from it.
0:31:02 Yeah.
0:31:06 I mean, I think the future of education is probably what on J.
0:31:07 Karpathy is leaning into.
0:31:10 I don’t know if you, you saw like his announcements about, uh, his,
0:31:14 his new program, but it’s like one teacher will be able to teach
0:31:18 millions of people and millions of languages by sort of creating
0:31:22 the educational course once training it into a large language model.
0:31:26 And then basically everybody has access to that teacher at scale
0:31:28 because they have access to that large language model that can
0:31:31 speak all the languages and understands the information that
0:31:32 the teacher gave it.
0:31:35 I mean, I don’t know how that fares for, you know, teachers in
0:31:38 the real world, but that to me feels like the future of education.
0:31:42 Yeah, I think you’d, I’d love to see like a combination of that.
0:31:45 And then you have like a real teacher, like taking the kids out
0:31:49 to nature and, and talking about, you know, teaching them logic,
0:31:51 teaching them how to think, how to be a good person, like having
0:31:54 conversations, you know, and like the social aspect.
0:31:56 And that’s what the teacher does.
0:31:57 You know, it’s almost like, you know, yeah.
0:32:00 Um, so that’s what I’d like to see personally.
0:32:01 Yeah.
0:32:04 I think, uh, like one missing piece is like, you know, in a
0:32:07 classroom environment, the kid is forced to sit there and listen
0:32:10 to the teacher, but, uh, like, what is the interest and motivation
0:32:13 for the kid to actually talk to AI and like learn stuff?
0:32:15 Maybe just about topics that they’re interested in.
0:32:16 Yeah.
0:32:17 There’s been some sort of motivation there.
0:32:18 Yeah.
0:32:18 Yeah.
0:32:22 Is there anything just in the AI world that, you know, maybe
0:32:26 you’ve gotten, uh, uh, an early preview of, or that you think is
0:32:28 in the pipeline that’s coming that you’re really excited about?
0:32:31 Like what, what excites you the most about the future of AI and
0:32:32 where it’s all headed?
0:32:34 I like this startup, uh, called WebSim.
0:32:36 Oh yeah, WebSim.ai.
0:32:36 Yeah.
0:32:38 It’s kind of similar to Roblox.
0:32:41 Like they let you just like type in some stuff and create a webpage
0:32:45 and people have created some really great user-generated content
0:32:46 from, from it.
0:32:48 So I think there’s like a lot of legs there.
0:32:51 You, usually it’s stuff that feels like, like a toy that they
0:32:54 want to play with that potentially ends up being very big.
0:32:55 Right?
0:32:55 Yeah.
0:32:58 Um, you know, I’m, I’m just hoping this, the financial models
0:33:00 get even better than they are today.
0:33:02 And, you know, I already talked to you like this stuff.
0:33:06 I already talked to Claude more, if I’m being honest, I probably
0:33:09 talked to Claude more during the day than my wife.
0:33:13 So, so, uh, yeah.
0:33:14 Yeah.
0:33:17 Well, I’m using AI, I’m using AI to help me translate things
0:33:18 to talk to my wife.
0:33:20 So, uh, yeah.
0:33:21 It’s like, yeah.
0:33:21 Yeah.
0:33:21 Yeah.
0:33:23 Can I tell her, I don’t want to do this without making her mad.
0:33:24 Right?
0:33:24 Yeah.
0:33:28 No, no, no, she’s Japanese and my Japanese is very, very basic.
0:33:28 It’s getting better.
0:33:31 I’m actually using AI to help teach me and it’s also helping teach
0:33:32 her English.
0:33:36 Uh, but for like deep conversations, we are still, we were
0:33:37 using chat to PT.
0:33:38 We’re now using Claude.
0:33:39 You know, you should look at Nathan.
0:33:43 They have like Samsung just released these new ear, um, earpods.
0:33:44 They, they look just like the Apple AirPods, but they’re
0:33:46 called the Galaxy Buds 3.
0:33:50 And they actually have like AI translation built in.
0:33:51 So you could be wearing ear AirPods.
0:33:53 She could be wearing AirPods.
0:33:54 You can have conversations.
0:33:57 And when you speak in English, it will automatically translate
0:33:59 it into Japanese right into her ear and vice versa.
0:34:01 So, I don’t know, something to look into.
0:34:04 I do, I do wonder if that’s going to improve our relationship
0:34:04 or not.
0:34:08 Hey, I will, I will stop it from translating.
0:34:11 If it thinks like, all right, maybe you should say that.
0:34:13 Oh, no, it does that.
0:34:13 There you go.
0:34:15 AI is going to save some marriages.
0:34:16 Awesome.
0:34:19 No, this has been a great conversation where, so I know you’ve
0:34:24 got a podcast and YouTube channel and you’re pretty prolific
0:34:28 over on Twitter slash X, uh, where should people go learn more
0:34:30 about you and check out what you got to offer?
0:34:30 Yeah.
0:34:34 Just go to, uh, my profile X, uh, for the, for the memes.
0:34:38 And then for my writing, um, go to, uh, creator.economy.so.
0:34:41 That’s where I share my writing and my YouTube channel is
0:34:41 like a new channel.
0:34:43 It’s like just getting started started.
0:34:44 Awesome.
0:34:46 Well, thank you so much for, for joining us today.
0:34:48 This is, this has been a great conversation and we really
0:34:49 appreciate you taking the time today.
0:34:50 All right guys.
0:34:51 Yeah, just pleasure being here.
0:34:52 Thanks Peter.
0:35:02 Yeah.
0:35:03 Yeah.
0:35:04 Yeah.
0:35:05 Yeah.
0:35:07 (gentle music)
0:35:17 [BLANK_AUDIO]
0:00:08 It surprised me. It behaves kind of like a person, right?
0:00:10 Like, it’s really valuable to give it a bunch of examples.
0:00:12 Are you going to give us all those prompts, or?
0:00:15 Oh, yeah, maybe.
0:00:18 Do you have any hot takes on where you see all of this heading?
0:00:22 The best AI features are actually smaller, more detailed AI features
0:00:25 that actually just make the core workflow easier.
0:00:29 Hey, welcome to the Next Wave podcast.
0:00:31 I’m Matt Wolf. I’m here with Nathan Lanz.
0:00:34 And today, we’ve got another amazing guest for you.
0:00:38 Today, we’re talking to the principal product manager over at Roblox.
0:00:41 He is the writer of creatoreconomy.so.
0:00:45 And today, we’re going to dive into some real actionable tips
0:00:48 that you can use to leverage a lot of the AI tools that are out there.
0:00:52 In fact, in this episode, Peter breaks down step by step
0:00:55 how to use specific prompts to get things done in your daily life.
0:00:59 Like, he actually walks us through how to type the prompts
0:01:01 and why these prompts work.
0:01:04 We’re also going to dive into all sorts of different use cases
0:01:07 and ways that you could leverage these AI tools in your business.
0:01:12 We’ve got some amazing stories that me and Peter and Nathan share
0:01:14 about how we’re using AI in your own business.
0:01:16 I really think you’re going to enjoy it.
0:01:18 So let’s go ahead and jump on in with Peter Yang.
0:01:24 When all your marketing team does is put out fires, they burn out.
0:01:28 But with HubSpot, they can achieve their best results without the stress.
0:01:31 Tap into HubSpot’s collection of AI tools,
0:01:34 breeze to pinpoint leads, capture attention,
0:01:37 and access all your data in one place.
0:01:41 Keep your marketers cool and your campaign results hotter than ever.
0:01:44 Visit hubspot.com/marketers to learn more.
0:01:50 Hey, Peter, welcome to the show.
0:01:52 Thanks so much for joining us today.
0:01:53 How are you doing?
0:01:54 Yeah, I’m doing great.
0:01:55 It’s great to hang out with you guys.
0:02:00 Let’s just dive right into this and learn a little bit about you.
0:02:06 So you’re lead of product over at Roblox and you’re an AI enthusiast.
0:02:09 You know, what we want to talk about today, I think,
0:02:13 is like some really actionable use cases for AI.
0:02:17 You seem to have figured out some really good ways to leverage AI
0:02:22 to just sort of make your life and your work easier and faster for you.
0:02:23 So I think with this episode,
0:02:26 it would be really fun to kind of lean into that
0:02:29 and talk about some of the really cool use cases
0:02:32 because there is sort of this narrative people on the peripheral of AI.
0:02:34 Maybe not the people that are like in it using it every day,
0:02:38 but the people seeing all of the AI news going,
0:02:39 “Eh, it’s not really useful yet.
0:02:41 Like, what can AI do for me?”
0:02:44 Those of us are using AI every day to see the power,
0:02:46 but I think a lot of people are still kind of going,
0:02:49 “I don’t know how to use this in my daily life yet.”
0:02:53 So let’s start as like from a content creator perspective.
0:02:57 What are some of the ways that you’re leveraging AI as a content creator?
0:02:59 I use it for a lot of things.
0:03:02 So I have this doc called like my AI writing assistant.
0:03:07 It’s just a Google doc and I save all my long prompts there, right?
0:03:09 So let me read you some of them.
0:03:15 So I have an AI prompt for making like a clickbaity LinkedIn post.
0:03:19 I have AI prompt for generally YouTube thumbnails.
0:03:24 Like, you know, like I have AI prompts for editing my newsletter.
0:03:27 I also do these interviews, these podcast interviews,
0:03:30 and it’s kind of like a pain to ask to clean up the transcript afterwards.
0:03:32 So I have an AI prompt for that.
0:03:34 Are you going to give us all those prompts or?
0:03:36 Well, yeah, maybe.
0:03:38 If you buy my fifth-hour prompt package.
0:03:39 Your course.
0:03:41 That’s kidding.
0:03:44 I also use AI to kind of like give me personal advice,
0:03:47 like, you know, kind of like act like a personal coach.
0:03:51 I give a bunch of information about myself and then just ask it for advice.
0:03:52 I’d love to hear some of the prompts.
0:03:55 I mean, you mentioned one for helping with like titles and thumbnails.
0:03:57 I mean, I do a lot on YouTube.
0:03:58 The next wave is on YouTube.
0:04:01 Like, can you give us a sneak peek at some of these prompts?
0:04:05 So basically the prompt is like, you’re an expert copywriter,
0:04:09 you know, your goals to generate 10 YouTube thumbnails and title combinations
0:04:12 based on this transcript that I’ll share with you next, right?
0:04:15 And then each combo should be two separate lines each.
0:04:16 And here’s some instructions.
0:04:18 So common patterns.
0:04:20 Create a curiosity gap.
0:04:22 Use numbers to make claims more compelling.
0:04:26 Use provocative or contrarian short quotes to grab attention.
0:04:30 Imply insider knowledge is being revealed.
0:04:32 Highlight impressive transformation.
0:04:34 Use YouTube to speak directly to the viewer.
0:04:39 And then I include a list of all the best thumbnail and title combinations.
0:04:43 Right? Like both from my own personal channel and other channels.
0:04:44 I mean, that’s really cool.
0:04:46 When you do enter these prompts, where are you going?
0:04:48 Is it are you using chat GBT, Claude, Lama?
0:04:50 Like what’s your go to?
0:04:52 Oh, I have a turn from chat GBT.
0:04:53 I stopped using it.
0:04:55 Yeah. So where are you going?
0:04:57 So I primarily use Claude now.
0:04:57 Yeah.
0:04:58 I use Claude every day.
0:05:01 Sometimes I use perplexity, but mostly Claude.
0:05:01 Yeah.
0:05:02 I love Claude.
0:05:05 I actually really, really love the projects feature
0:05:06 because some of these things that you’re talking about,
0:05:10 you can just kind of save them as a project and then use them over and over again.
0:05:12 Like I do something like pretty similar.
0:05:17 I basically made a giant spreadsheet of like all of my favorite YouTube headlines
0:05:18 that I’ve come across.
0:05:21 I’ve actually had a Google sheet going of all my favorite YouTube headlines
0:05:26 for like a couple of years now, just as like a sort of swipe file of headlines.
0:05:29 And I was able to take all of that list of headlines,
0:05:34 throw it into Claude and let Claude sort of analyze what headlines,
0:05:37 like what are the patterns in these headlines that seem to make them work?
0:05:41 And then Claude is really good at generating new headlines
0:05:43 based on that information, it seems.
0:05:44 Yeah, it’s fantastic.
0:05:47 I mean, unfortunately, I’ve noticed a pattern where like, you know,
0:05:50 you basically, it seems like people on YouTube really want to make money.
0:05:53 So you have to mention like this is like a billion dollar opportunity
0:05:57 or like, you know, I wish I knew this earlier.
0:05:59 So sometimes it gets a little bit too click baby for my taste,
0:06:02 but yeah, it’s still useful.
0:06:05 Oh yeah, I mean, like as a YouTuber myself,
0:06:10 that’s like one of the balances I’m constantly trying to figure out is like,
0:06:14 how do I make a title that like grabs attention so people want to click on it?
0:06:17 I want to bait the click, but I don’t want to be too hypey
0:06:19 where people click on it and go, wait a second,
0:06:22 I feel duped based on what, you know, what that title was.
0:06:25 That feels like one of the hardest balances.
0:06:26 That’s the same thing on Twitter and LinkedIn too, right?
0:06:29 Like I’ve tested like doing threads or it’s like,
0:06:31 I don’t want to be over hyping something.
0:06:34 I’ve tested a thread like that and then tested one where like I super hype things
0:06:36 up or mention money or whatever.
0:06:38 And then like they do way better every single time.
0:06:44 So when it comes to prompting, like, are there any specific like tricks
0:06:49 or tactics or tips you give to like actually get decent prompts?
0:06:54 A couple of things that I try to do is I try to get it to proceed step by step, right?
0:06:58 So like, I think it’s really valuable to give it a bunch of examples
0:07:00 to make it more personal.
0:07:04 Like it’s actually incredibly bad as just generating random shit from scratch.
0:07:06 Like you gotta give it as many examples as you can.
0:07:10 And then what I have to do is like, I have to do two steps.
0:07:12 So first look at all the examples that I give you
0:07:17 and like give me a list of bullet points of like some common themes from these examples.
0:07:22 And then go ahead and edit my thumbnails or edit my new set of posts
0:07:26 or you know, edit my stuff and it seems to do pretty well with that.
0:07:27 Have you given me like negative examples?
0:07:29 Like here’s something that’s bad as well.
0:07:30 That’s what I’ve tried doing.
0:07:31 It seems to work pretty well.
0:07:32 Like here’s something that works well.
0:07:34 Here’s like a format I like.
0:07:36 Here’s something that I hate.
0:07:39 For some reason, it likes to use semi-colon a lot.
0:07:42 So I tell them to like stop using semi-colon for fricassex.
0:07:46 Yeah, stop delving into everything.
0:07:47 Yeah, stop delving.
0:07:51 Yeah, there’s like, there’s like definitely sort of common words
0:07:55 that when you let it just write for you, you’re like, all right, I can like,
0:07:56 I see it on Twitter all the time.
0:07:58 I see it in blog posts all the time.
0:08:03 It’s like, I feel like those of us that are in AI now that pay close attention
0:08:07 can spot chat GPT written stuff just based on the language it uses.
0:08:13 Yeah, you mentioned that like you always tell it to think through step by step.
0:08:15 What’s the, what’s the logic there?
0:08:18 Like what’s the reasoning you’re telling it to think step by step?
0:08:20 Like shouldn’t an AI just do that naturally?
0:08:22 It’s like, it surprised me.
0:08:23 It behaves kind of like a person, right?
0:08:30 Like if it’s able to kind of summarize or like do a step first and
0:08:33 then do another step, it turns out that it does it better.
0:08:35 Then we try to do it, make it do everything at once.
0:08:38 And also if you like flatter it a little bit and like, you know,
0:08:42 you really get this or like, you know, you got to have confidence in yourself.
0:08:45 Give me a really, give me like an 11 out of 10 answer.
0:08:48 Like you can do it and then it will like try harder.
0:08:52 Let’s talk through like what AI is like really good at.
0:08:55 Like what would have been some of the like best use cases you’ve found
0:08:56 for some of these large language models?
0:08:59 I know you mentioned some as far as like content creation for titles
0:09:03 and thumbnails and things like that, but let’s, let’s like expand it.
0:09:06 Let’s, let’s like break open the floodgates a little bit and like share
0:09:10 with people who might be thinking like, I don’t need AI in my business.
0:09:13 I don’t need AI in my like sort of daily life.
0:09:18 What are some of the really cool use cases that can really help people
0:09:20 if they just sort of lean into it a little bit?
0:09:25 I think the most common use case so people think like AI is really good
0:09:29 at like generating stuff from scratch and I actually don’t think
0:09:30 it’s the best at that.
0:09:33 I think what’s really good at is if you have a bunch of information
0:09:37 that you just don’t want to read, it’s really going to help you
0:09:38 feel lazy, right?
0:09:41 Like you don’t want to, I don’t want to, I don’t want to read all this shit
0:09:43 or like, you know, like this whole, this whole post.
0:09:46 So I’m just going to paste this stuff in and like tell it, like give
0:09:50 you some constraints like, hey, give me the main takeaways in five lines
0:09:54 or like when I’m researching for guests for podcasts, I like to I don’t
0:09:57 like to watch their past YouTube videos because I don’t have time for that shit.
0:10:01 So what I do is I just take the transcript from that video and I paste
0:10:05 it into AI and I tell it to like give me a list of nested bullets
0:10:08 with the main takeaways and also include some quotes from the guest.
0:10:10 Yeah.
0:10:12 And then it’s basically like watching a one hour podcast without
0:10:13 actually watching it, right?
0:10:16 Like I can get the TL;DR in like five, five, five minutes.
0:10:16 Yeah.
0:10:18 I mean, hopefully people don’t use it for this podcast, but you know,
0:10:23 is there any easy way to get the transcripts with AI yet?
0:10:25 Or do you have to like manually go download them and then you like
0:10:28 you’re copying, pasting or uploading a file or?
0:10:30 I think there’s some plugins, but I just go to YouTube and like
0:10:34 copy and paste the transcript and I paste it into Claude to do it.
0:10:35 Yeah.
0:10:38 You can use tools like, um, like the script as well.
0:10:42 And then a lot of times what I’ll do too is I’ll say, you know,
0:10:46 I don’t want to ask the same questions Lex Friedman asked Elon on this
0:10:50 podcast based on this interview, what are some gaps in the conversation?
0:10:53 What are some interesting things that I can ask this guest that weren’t
0:10:55 asked on this other podcast?
0:10:59 And it actually helps find some like new topics and side chains
0:11:01 that they didn’t get into.
0:11:02 And I find that to be really valuable.
0:11:04 Yeah.
0:11:07 That that’s worth valuable because I think it’s just so much easier
0:11:12 to, to edit something like edit what AI’s questions are, like edit
0:11:15 something from AI to come up with from scratch by yourself.
0:11:16 Right?
0:11:18 So it’s like a lot easier.
0:11:21 What are some of the other ways that you’re using it to like really
0:11:23 leverage your time?
0:11:28 Um, so in my product job, we have discord or we have like these
0:11:32 uh, communities with a bunch of customers and we do like alphas and
0:11:33 betas with these customers.
0:11:36 And sometimes I’ll come back the next day and there’s like, you know,
0:11:38 so many messages to be through.
0:11:40 So, so I just like copy paste or thing.
0:11:41 I’m like, Hey, what did people talk about?
0:11:44 Or like maybe, maybe someone tags you on a slack thread.
0:11:46 That’s like 15 messages deep.
0:11:48 And they’re like, Hey, Peter, what do you think?
0:11:49 I mean, I’m not real.
0:11:53 So let me summarize this stuff.
0:11:54 Like what has been discussion so far?
0:11:55 Right.
0:11:59 Like people like on Twitter keep sharing these templates about like, oh,
0:12:02 here’s like nine and nine product spec templates they should follow.
0:12:04 Like all this crap.
0:12:06 I’m like, why even write a prospect?
0:12:08 Just get AI to write right for you.
0:12:10 Just like, I actually try this.
0:12:13 I had like a few bullet points of like a product feature that
0:12:14 I wanted to build.
0:12:17 And then I basically started some discussions internally and some
0:12:18 discussions with customers.
0:12:19 Right.
0:12:20 And they gave me a bunch of feedback.
0:12:23 And then I was like, okay, then I pasted my bullet points and all the
0:12:24 feedback into AI.
0:12:26 Like, okay, now make a prospect for me.
0:12:28 And that worked well.
0:12:29 Yeah.
0:12:29 It worked pretty well.
0:12:33 I mean, I also gave a template that I use for my specs.
0:12:35 But I mean, it’s pretty, it’s pretty, pretty good.
0:12:36 Yeah.
0:12:40 And the last thing I’ll say is like at work, at these corporate jobs,
0:12:43 it’s very important to become a very good and internal communication.
0:12:43 Right.
0:12:46 There’s people who are not good at this.
0:12:49 Like they, they are like very robust or like they write these like walls of
0:12:53 text and before I would tell them to like, okay, you know, here’s some
0:12:56 tips to, but now it’s like, Hey, why don’t you just pull your shit into AI
0:12:59 first and then it’ll just make it more concise for you.
0:13:03 There’s like no excuse not to be super worthy.
0:13:03 Yeah.
0:13:03 Yeah.
0:13:05 I do that a lot with my newsletter as well.
0:13:08 And I think, I think you mentioned you do it with your newsletter is also
0:13:12 where I’ll take my, I’ll write my newsletter that I send out each week
0:13:16 and I’ll pull it into, you know, Claude and then I’ll say, can you proofread
0:13:20 this, find any grammar mistakes, spelling mistakes, but also proofread
0:13:21 it for readability.
0:13:24 Make sure it’s as easy to read as possible for people and it will
0:13:27 actually go and trim out some of the fluff for me and clean it up.
0:13:30 But it might add some delves here and there, which I got to go back in and
0:13:34 remove, but for the most part, it does a pretty good job of like bringing
0:13:37 it down to just the meat of what I’m trying to put out.
0:13:38 Yeah.
0:13:41 It’s fantastic, especially if you gave it like your best posts from
0:13:42 the past that you’re proud of.
0:13:42 Yeah.
0:13:43 It’s really good.
0:13:44 I kind of match that style.
0:13:44 Yeah.
0:13:47 I personally, I don’t know if you found a way to do this, but personally, I
0:13:51 haven’t found a way to make AI sort of right with my voice.
0:13:55 You know, so like, I can upload five of my emails into Claude and say,
0:14:02 write it to sound like me and it will like, it might use some of the same
0:14:04 words, but for the most part, I’ll read it and it won’t sound anything
0:14:08 like, like the way I write, but like, have you found a way to, to make
0:14:09 that kind of thing work?
0:14:11 Um, not so much.
0:14:14 I think it’s better at editing my writing.
0:14:18 Like no matter, like even like a really crappy draft, it’s a lot better
0:14:21 if you give it something than, uh, if you just try to write from scratch.
0:14:24 Now, are you using any of the other like AI tools that are out there?
0:14:27 I mean, so far we’ve pretty much talked about like large language models
0:14:29 and, you know, chat bots like Claude and chat GPT.
0:14:32 Are you using other stuff outside of the LLMs?
0:14:37 So this is also why I think like the, like the AI startup thing is
0:14:40 like kind of like overhyped because, um, the tools actually come back to
0:14:44 you every week are pretty much Claude and maybe perplexity.
0:14:47 Um, and that’s it.
0:14:51 You know, the other ones, the other ones, like there’s some really cool stuff
0:14:51 out there.
0:14:51 Right.
0:14:54 There’s like sooner for generating music.
0:14:57 There’s like, uh, runway for generating videos.
0:15:01 And you know, obviously mid-majority, but like, I don’t, I don’t use it.
0:15:04 I, these are stuff that I tried is like kind of a novelty.
0:15:04 Yeah.
0:15:06 And then I kind of churn from, from it.
0:15:12 We’ll be right back, but first I want to tell you about another
0:15:14 great podcast you’re going to want to listen to.
0:15:17 It’s called Science of Scaling hosted by Mark Roberge and it’s
0:15:22 brought to you by the HubSpot podcast network, the audio destination
0:15:23 for business professionals.
0:15:28 Each week host Mark Roberge, founding chief revenue officer at HubSpot,
0:15:31 senior lecturer at Harvard Business School and co-founder of Stage 2
0:15:35 Capital sits down with the most successful sales leaders in tech to
0:15:39 learn the secrets, strategies and tactics to scaling your company’s
0:15:40 growth.
0:15:44 He recently did a great episode called, how do you solve for a siloed
0:15:45 marketing and sales?
0:15:47 And I personally learned a lot from it.
0:15:49 You’re going to want to check out the podcast.
0:15:53 Listen to Science of Scaling wherever you get your podcasts.
0:15:57 There’s one thing that I do.
0:16:00 Someone use, uh, on some of my regular basis, which is, uh, this
0:16:02 app called Super Whisper.
0:16:06 It lets me go outside and like take a walk and like talk to my phone
0:16:09 and like start a draft that way.
0:16:10 So that kind of helps a little bit.
0:16:15 So it basically, you talk to it, it transcribes everything you say and
0:16:17 then summarizes it or bull points it or.
0:16:17 Yeah.
0:16:18 It transcribes everything.
0:16:23 It does have a local LEM that uses to summarize stuff, but like, uh, the
0:16:24 LN kind of sucks.
0:16:27 So what I usually do is I just put, but I just paste the transcript into
0:16:29 cloud and tell it to make it better.
0:16:29 Yeah.
0:16:32 I mean, I use a lot of the like AI art stuff.
0:16:36 I love, I love using like mid-journey and Leonardo and tools like that for
0:16:39 my thumbnails, because for me, it’s like very meta.
0:16:40 I make YouTube videos about AI.
0:16:43 And so my thumbnails are mostly made with AI as well.
0:16:48 And it’s just like, I love that sort of like meta-ness of it at all.
0:16:50 Um, yeah, but yeah, I totally agree.
0:16:51 Like I love runway.
0:16:52 I love Suno.
0:16:57 I love playing with these toys really, but I don’t really have a lot
0:16:58 of practical use cases for them.
0:17:02 I’ve used a couple of Suno videos and like, or songs and like the background
0:17:05 of some of my videos, but it’s pretty rare that I’ll do that.
0:17:10 I think runway and Luma and that kind of stuff is really cool, but it’s not
0:17:12 quite there yet still for me to feel like.
0:17:16 I want to use this as like b-roll that looks realistic.
0:17:18 You know, like, I think it’s getting closer though.
0:17:20 I mean, like you look at a year ago, it was so horrible.
0:17:22 And now it’s like, it’s pretty good.
0:17:24 It’s not good enough yet, but it’s, you know, it’s pretty good.
0:17:26 I mean, there was that commercial.
0:17:28 Was it the Toys R Us commercial that they used?
0:17:29 Uh, yeah.
0:17:31 So, uh, AI videos.
0:17:33 So, I mean, I think we’re going to see a lot of those actually start
0:17:36 to be used and, and even though, you know, maybe the AI startups are
0:17:40 overhyped, whatever, but like, that’s kind of the nature of VC, right?
0:17:42 It’s like, it’s, you need to be doing these experiments.
0:17:45 It’s good for the industry as a whole to be creating these startups
0:17:48 and trying things and yeah, most of them will not work, but hopefully
0:17:51 they will all kind of move things forward for, you know, for everybody.
0:17:54 But, uh, you know, like the risk profile of the VC is very different
0:17:54 from the founder, right?
0:17:58 The VC can just like have one out of 100 work, but the founder,
0:18:00 if you’ve been on like some sort of AI rapper and doesn’t work,
0:18:01 then you’re kind of screwed.
0:18:01 Yeah.
0:18:01 Yeah.
0:18:02 Yeah.
0:18:04 You know, so I mean, it depends.
0:18:06 I mean, like, you know, I, I’ve done a few startups.
0:18:09 I mean, typically even if your startups fail, there’s other opportunities
0:18:11 that arise because of that.
0:18:11 Yeah.
0:18:12 I mean, not always.
0:18:15 If you do something very uninteresting and you’re, and you, and you do
0:18:18 a really poor job, then yeah, you, you might be somewhat screwed.
0:18:21 But you, but usually there’s a network of people and there’s other opportunities.
0:18:22 But yeah, anyways.
0:18:24 Uh, and I’ll say like one more thing about like, uh, cause a lot of
0:18:26 these bigger companies are also trying to build AI, right?
0:18:29 Like, like, um, not necessarily to Google and Microsofts of the
0:18:30 world.
0:18:34 And I think what I’ve noticed is that the best AI features from these
0:18:37 bigger companies, like if I already have a product that’s like people love
0:18:42 to use, the best AI features are actually like, um, smaller, more detailed
0:18:46 AI features that actually just like make the core workflow easier.
0:18:47 You know what I mean?
0:18:50 Like as opposed to actually like trying to build some sort of chatbot
0:18:55 or like agent feature, just like think about your core customer journey
0:18:59 and, and like pick the areas where AI can actually help you save a bunch of
0:19:00 time or like make things easier.
0:19:03 And those are kind of the features that work the best.
0:19:08 Um, and an example of that is, um, I talked to this, uh, prod director
0:19:13 from Shopify, his first feature was like just generating product descriptions
0:19:17 like helping the merchants generate prod descriptions for like their hundreds
0:19:19 of SKUs, you know, right?
0:19:22 And it’s not super sexy, but like it saves them a ton of time.
0:19:22 Yeah.
0:19:24 So they’re using it all the time.
0:19:26 I don’t know if you’re a gamer or not.
0:19:28 I’m Nathan and I are both gamers.
0:19:31 So we talk about games from time to time, but did you hear about the new
0:19:35 like college football game that EA put out, but they used AI to sort
0:19:37 of develop the game at rapid speed?
0:19:39 No, I am a gamer, but I haven’t heard about that yet.
0:19:43 So basically like in, you know, in the, in the NFL, right?
0:19:47 There’s hundreds of players, maybe in the thousands, but like not as many
0:19:49 players that are in college, right?
0:19:53 In college football and NCAA, there’s over 11,000 players.
0:19:57 Well, EA just released a new game with like pretty much all of the division
0:19:59 one college teams in that game.
0:20:03 And they had to figure out how do we get 11,000 players into this game?
0:20:09 So they used AI and they basically took a single headshot of every sing
0:20:14 all 11,000 players and used AI to create characters that had each
0:20:15 person’s face on it.
0:20:18 So like the build of the person may not be exact, but they were able
0:20:23 to get the face of all 11,000 players by using AI and just putting them
0:20:27 all into a model that then generated a 3D character with that person’s face.
0:20:31 And I just thought that was like an amazing use case of like, I don’t
0:20:32 really think it put anybody out of the job.
0:20:35 It just made a really, really tedious task way less tedious.
0:20:36 Yeah, exactly.
0:20:38 Those were like the perfect use cases.
0:20:40 Just like doing something that no human wants to do basically.
0:20:42 So there was a, there was an article that you wrote and I, and I
0:20:43 saw you, you tweet about it.
0:20:49 And one of the bullet points of the article was, you know, why some AI
0:20:53 startups gain traction while other AI startups fail.
0:20:54 You know, what are your thoughts on that?
0:20:58 Like why do some AI startups gain massive traction and everybody’s
0:21:01 talking about them while other ones just sort of fade into the night?
0:21:05 I think with AI startups, there’s like definitely like a hype cycle.
0:21:09 So you might get a shit ton of users when you launch or you
0:21:11 might even get like a bunch of money.
0:21:15 But like then the people might just move on to the next thing.
0:21:19 It’s like, I think retention is very difficult with these AI startups.
0:21:24 And I think a failure mode is a startups that like get a ton of users
0:21:27 and they go to like VVC’s and they’re like, okay, you know, give me
0:21:30 more money and then they have this crazy valuation.
0:21:31 Right?
0:21:31 Yeah.
0:21:34 And then they, they have this like AI product for, you know, sales
0:21:35 and marketing or something else.
0:21:39 But then like the foundational models are always getting better.
0:21:42 And then there’s like no point in using them anymore.
0:21:43 Like what’s the point?
0:21:46 And because their valuation is so crazy, it’s going to be very hard
0:21:49 to actually survive and make a, you know, continue on.
0:21:53 I mean, the ones that actually, you know, there’s this guy called
0:21:55 level CEO on Twitter, right?
0:21:57 Like he has like a headshot AI or something.
0:22:01 I don’t think he raised any of these two dollars, but he just found
0:22:05 like a really good niche use case and he’s making a ton of money from it.
0:22:09 So it’s almost better to just like bootstraps to some extent.
0:22:09 Yeah.
0:22:10 Some of the stuff.
0:22:10 Yeah.
0:22:14 I think, I think even he mentioned he’s been using more AI himself
0:22:17 versus like building an AI product like recently because he had a lot
0:22:21 of turn of like having these like, here’s like a, you know, AI for headshots
0:22:21 or whatever.
0:22:23 I know that’s one of his products and he has other ones.
0:22:27 But yeah, he said he was saying huge turn because yeah, as a technology
0:22:31 gets so good, people can very quickly copy you and make a similar service
0:22:32 and make it cheaper.
0:22:32 Right.
0:22:34 Well, you can do it for free and meta now, right?
0:22:36 Like, so meta just rolled out there.
0:22:42 Imagine me feature like last week where it looks at all of your, your headshots
0:22:48 or your pictures of you inside of your photos on Facebook and it essentially
0:22:51 trains a model and you can say, imagine me riding a dragon and it puts
0:22:52 me on a dragon, right?
0:22:58 So like tools like that just overnight seem to make companies obsolete
0:22:59 that we’re selling that service, right?
0:23:03 We’ve seen open AI do that over and over and over again too, right?
0:23:07 You have like so many tools that came out that were like summarized
0:23:09 my PDF tool or chat with my PDF tool.
0:23:14 And now you can do that in the free version of Claude and chat GPT, right?
0:23:20 There’s like, yeah, I feel like a lot of the struggles of like the AI startups
0:23:23 are that the big incumbents can just roll it out and they’ve got so
0:23:27 much capital available to them that they can do it at, you know, free
0:23:28 or practically free.
0:23:29 Yeah, exactly.
0:23:33 I mean, but then again, you know, if you bootstrap your PDF startup and
0:23:37 you make it like a million dollars and then, you know, it becomes obsolete.
0:23:39 Like that’s not a bad outcome, right?
0:23:39 Yeah, yeah, yeah.
0:23:43 You just have a window of opportunity that you got to capture, you
0:23:44 know, capture, I guess.
0:23:47 And Zuck is like, you don’t want to compete with Zuck, man.
0:23:48 He’s like unstoppable.
0:23:52 Meta is going to keep on putting out a lot of this open source stuff and
0:23:55 just, you know, for lack of a better word, democratize it all.
0:23:56 Yeah.
0:24:00 I mean, I thought openly I was untouchable, but you know, with Zuck
0:24:01 and these other players involved.
0:24:03 Yeah, he’s very strategic.
0:24:06 I remember back in the day, I had a lot of friends who were making
0:24:10 fortunes on social apps and social games on Facebook.
0:24:13 And Facebook really took advantage of that to grow at that moment.
0:24:15 And they’re like, oh, they, you know, Facebook loves us new speed
0:24:18 parties where you go hang out with Facebook employees and everything.
0:24:20 And then overnight when Facebook didn’t need that anymore.
0:24:22 Nope, you’re all gone.
0:24:23 Yeah.
0:24:26 So I would be, I would be mindful of like, you know, just, you know,
0:24:30 yeah, sure, Zuck has changed or whatever, but you know, how much
0:24:31 do people change?
0:24:32 So yeah.
0:24:33 Yeah, sure.
0:24:33 It’s great.
0:24:36 He’s doing all this open source stuff, but who knows what the real
0:24:36 objective is.
0:24:40 I mean, he’s, he’s, he’s into win-win and he’s like, you know,
0:24:44 do you think this is, this is a debate, you know, Nathan and I
0:24:46 have had and we’ve had it with other guests as well.
0:24:48 Do you think, do you think open AI is falling behind?
0:24:51 I don’t want to speculate too much, but I do notice that they
0:24:56 went from just like dropping these products to, to launching stuff
0:25:00 with like wait lists and be like, oh, I’m going to roll it out
0:25:01 slowly, slowly, slowly.
0:25:01 Yeah.
0:25:05 And maybe it’s because of a lot of drama they had, but it does
0:25:08 seem like they’ve slowed down the problem of loss here.
0:25:11 I mean, in comparison to your Anthropic, which is like shipping
0:25:12 stuff left and right.
0:25:14 I’m still of the belief that they’re so far ahead in the new models
0:25:17 coming and it’s, it’s going to make Anthropic stuff look kind of
0:25:21 silly in comparison, but there’s also the election coming up.
0:25:23 They’ve brought on people from the government who are now a part
0:25:25 of open AI apparently.
0:25:28 You know, and so I think, I think we’re going to, as soon as the
0:25:30 elections over, we’re going to see, you know, GPT five and it’s
0:25:31 going to be mind blowing.
0:25:31 Yeah.
0:25:32 Yeah.
0:25:34 They’ve also done a really great job with like partnerships
0:25:35 and marketing.
0:25:39 Uh, like, you know, I saw some stat like to, to, to, to us, like
0:25:42 in my opinion, like Claude is clearly better right now, but I
0:25:44 think way more people are still using chat GPT.
0:25:45 Oh yeah.
0:25:48 There’s like, if you go and look at like Google trends, right?
0:25:53 And you type in Claude or Anthropic Claude and you type in chat GPT.
0:25:54 It’s nowhere even close.
0:25:57 Claude just looks like a flat line down at the bottom compared
0:25:59 to the amount of people that are still searching for GPT.
0:26:00 I mean, I mean, it’s even a term, right?
0:26:02 Like in this conversation, Matt, use the term chat GPT, like
0:26:05 as a generic, you know, like for like LLIMs, right?
0:26:08 Where you could have said Claude, but you said chat GPT, you know.
0:26:08 Yeah.
0:26:08 Yeah.
0:26:12 It’s, it’s become the Kleenex of AI or the Band-Aid of AI, right?
0:26:16 Like people just use the brand name now is like a synonym for
0:26:19 the, uh, for the, uh, you know, the underlining tech.
0:26:22 But, um, you know, I, I, I don’t know.
0:26:26 I, I have a feeling that open AI isn’t as far as, uh, ahead as I
0:26:30 think Nathan believes it is one of the areas that I think we
0:26:32 kind of, um, go back and forth on a little bit.
0:26:36 Um, I, I do, I think what we’re seeing, but it’s a very sort
0:26:38 of closed bubble thing, right?
0:26:41 Like we all pay attention to AI on almost a daily basis.
0:26:45 So we’re seeing it, but I get really, really frustrated by
0:26:48 open AI announcing stuff, but then never shipping it, right?
0:26:50 Like the GPT for a voice.
0:26:51 We’re still waiting.
0:26:55 Sora, we’re still waiting, you know, search GPT.
0:26:59 Now that’s only announced last week, but nobody’s got access to it yet.
0:27:04 And, um, you know, among us AI nerds, that gets frustrating to me.
0:27:06 Um, and yeah, I don’t know.
0:27:11 I, I, I don’t know if they’re as far ahead as everybody believes
0:27:11 they are.
0:27:14 And Elon Musk claims that he’s further ahead because he’s got
0:27:16 the largest, uh, data center on the planet now.
0:27:18 And I don’t know what we’ll see.
0:27:22 I just like watching it all play out and, you know, eat my popcorn.
0:27:23 Yeah.
0:27:25 I mean, I only have 20 bucks a month to spend.
0:27:27 So whoever, whoever’s ahead, I’ll spend on.
0:27:30 You know, one of the, one of the other topics that I, I think
0:27:32 would be interesting to talk about is like, where we think
0:27:36 all of this is going, do you have any sort of like bold predictions?
0:27:40 Do you have any hot takes on, on where you see all of this heading?
0:27:43 One of the things that I’ve been talking about a lot in my product
0:27:45 stuff is like, and this company is like, people just want to
0:27:46 manage more and more people, right?
0:27:49 Like, you know, getting more headcount and manage more people is
0:27:50 seen as clout.
0:27:52 Like you have more responsibility.
0:27:54 And I think that’s just like pretty toxic and at least two
0:27:59 companies bloating and like eventually having to do layoffs.
0:27:59 Right.
0:28:00 Yeah.
0:28:00 Yeah.
0:28:04 So my hope is with AI and I think maybe dance shipper or coin is a phrase,
0:28:06 but like, you know, instead of managing people, yeah, people who
0:28:11 manage AI agents or like, you basically have individual contributors
0:28:15 and crafters who can actually get a lot of shit done and like do a
0:28:16 lot more than they can before.
0:28:19 And then you can keep the company small.
0:28:22 If you keep the company small, there’s less bureaucracy and bullshit
0:28:25 and everyone can actually just like believe in the mission and like,
0:28:27 you know, execute on a product.
0:28:32 So that has my hope in the age of AI, our product managers can be
0:28:35 more important or less because I could see it going either way.
0:28:35 Right.
0:28:38 Like, like, yeah, product managers always thought it was almost like a,
0:28:40 you know, you know, sorry to say this, but like a, you know,
0:28:43 necessary evil that you need the product managers to manage all
0:28:45 these, these, you know, individual contributors.
0:28:49 Like, I wonder like, are PMs not going to need those individual
0:28:51 contributors as much and they can go off and create their own
0:28:53 projects or features at companies?
0:28:55 I mean, product managers ask much value.
0:28:56 You know, like it’s hard for AI too.
0:29:01 Only product managers cannot be replaced.
0:29:01 Okay.
0:29:04 No, yeah, gotcha.
0:29:07 I hope, I hope that like another thing that has happened as companies
0:29:10 got bigger is like product teams have like before, which is like a
0:29:12 PM, a designer and an engineer.
0:29:14 And now there’s like so many different roles, like, you know,
0:29:18 user researcher, there’s like data analysts, there is product
0:29:22 ops, product specialists, a lot kinds of crazier is like all the agile
0:29:25 coaches kind of crap, but people will learn to wear multiple hats.
0:29:29 Like, for example, I would love to become like a hybrid designer,
0:29:33 PM, as opposed to just a PM, or maybe like engineers can design too.
0:29:37 But product managers do have one skill, which is they’re really good at,
0:29:39 you know, writing detailed requirements, right?
0:29:42 So in some ways that’s kind of like prompting AI, right?
0:29:45 Yeah, but you said the product, you said the AI can do that for you,
0:29:46 right?
0:29:48 Which means it gets you to go up and get the questions and get the spec.
0:29:52 You know, if people are more productive and then companies can do,
0:29:56 you know, do the same task with less people, where, where do you
0:29:58 think that like new jobs are created?
0:30:02 I mean, like, you know, one outcome is like AI robots would do
0:30:05 all the real work and then all the people would become like influencers
0:30:06 and try to entertain each other.
0:30:08 That’s what I’m making on.
0:30:12 That’s very possible, I think.
0:30:13 But yeah, I don’t know, man.
0:30:17 Like I feel like AI is going to start to disrupt the white color jobs.
0:30:20 And then I think once we have like robots that can do stuff in the real
0:30:23 world, that’s going to be when the real disruption happens, man.
0:30:27 Like, you know, like instead of hiring like a nanny, I can just hire a
0:30:28 robot or something.
0:30:30 And it can teach my, my, my kid at the same time.
0:30:32 It has all this context about my child and what they know when
0:30:35 their education level language.
0:30:38 So they can, they can teach Japanese and English.
0:30:39 Yeah.
0:30:39 Yeah.
0:30:40 Yeah.
0:30:40 I mean, we all have kids.
0:30:43 So it’s like, it’s hard to imagine how education will change.
0:30:46 I mean, the current system is so antiquated.
0:30:49 You know, I actually, um, a friend of mine is sending me this toy
0:30:50 called dino.
0:30:54 Uh, that has like a, it’s like a stuff that stuff dinosaur that
0:30:57 has heard of that and building or something.
0:31:00 So I’m going to try it with my daughter and see if she can ask it.
0:31:01 You know, she can learn from, from, from it.
0:31:02 Yeah.
0:31:06 I mean, I think the future of education is probably what on J.
0:31:07 Karpathy is leaning into.
0:31:10 I don’t know if you, you saw like his announcements about, uh, his,
0:31:14 his new program, but it’s like one teacher will be able to teach
0:31:18 millions of people and millions of languages by sort of creating
0:31:22 the educational course once training it into a large language model.
0:31:26 And then basically everybody has access to that teacher at scale
0:31:28 because they have access to that large language model that can
0:31:31 speak all the languages and understands the information that
0:31:32 the teacher gave it.
0:31:35 I mean, I don’t know how that fares for, you know, teachers in
0:31:38 the real world, but that to me feels like the future of education.
0:31:42 Yeah, I think you’d, I’d love to see like a combination of that.
0:31:45 And then you have like a real teacher, like taking the kids out
0:31:49 to nature and, and talking about, you know, teaching them logic,
0:31:51 teaching them how to think, how to be a good person, like having
0:31:54 conversations, you know, and like the social aspect.
0:31:56 And that’s what the teacher does.
0:31:57 You know, it’s almost like, you know, yeah.
0:32:00 Um, so that’s what I’d like to see personally.
0:32:01 Yeah.
0:32:04 I think, uh, like one missing piece is like, you know, in a
0:32:07 classroom environment, the kid is forced to sit there and listen
0:32:10 to the teacher, but, uh, like, what is the interest and motivation
0:32:13 for the kid to actually talk to AI and like learn stuff?
0:32:15 Maybe just about topics that they’re interested in.
0:32:16 Yeah.
0:32:17 There’s been some sort of motivation there.
0:32:18 Yeah.
0:32:18 Yeah.
0:32:22 Is there anything just in the AI world that, you know, maybe
0:32:26 you’ve gotten, uh, uh, an early preview of, or that you think is
0:32:28 in the pipeline that’s coming that you’re really excited about?
0:32:31 Like what, what excites you the most about the future of AI and
0:32:32 where it’s all headed?
0:32:34 I like this startup, uh, called WebSim.
0:32:36 Oh yeah, WebSim.ai.
0:32:36 Yeah.
0:32:38 It’s kind of similar to Roblox.
0:32:41 Like they let you just like type in some stuff and create a webpage
0:32:45 and people have created some really great user-generated content
0:32:46 from, from it.
0:32:48 So I think there’s like a lot of legs there.
0:32:51 You, usually it’s stuff that feels like, like a toy that they
0:32:54 want to play with that potentially ends up being very big.
0:32:55 Right?
0:32:55 Yeah.
0:32:58 Um, you know, I’m, I’m just hoping this, the financial models
0:33:00 get even better than they are today.
0:33:02 And, you know, I already talked to you like this stuff.
0:33:06 I already talked to Claude more, if I’m being honest, I probably
0:33:09 talked to Claude more during the day than my wife.
0:33:13 So, so, uh, yeah.
0:33:14 Yeah.
0:33:17 Well, I’m using AI, I’m using AI to help me translate things
0:33:18 to talk to my wife.
0:33:20 So, uh, yeah.
0:33:21 It’s like, yeah.
0:33:21 Yeah.
0:33:21 Yeah.
0:33:23 Can I tell her, I don’t want to do this without making her mad.
0:33:24 Right?
0:33:24 Yeah.
0:33:28 No, no, no, she’s Japanese and my Japanese is very, very basic.
0:33:28 It’s getting better.
0:33:31 I’m actually using AI to help teach me and it’s also helping teach
0:33:32 her English.
0:33:36 Uh, but for like deep conversations, we are still, we were
0:33:37 using chat to PT.
0:33:38 We’re now using Claude.
0:33:39 You know, you should look at Nathan.
0:33:43 They have like Samsung just released these new ear, um, earpods.
0:33:44 They, they look just like the Apple AirPods, but they’re
0:33:46 called the Galaxy Buds 3.
0:33:50 And they actually have like AI translation built in.
0:33:51 So you could be wearing ear AirPods.
0:33:53 She could be wearing AirPods.
0:33:54 You can have conversations.
0:33:57 And when you speak in English, it will automatically translate
0:33:59 it into Japanese right into her ear and vice versa.
0:34:01 So, I don’t know, something to look into.
0:34:04 I do, I do wonder if that’s going to improve our relationship
0:34:04 or not.
0:34:08 Hey, I will, I will stop it from translating.
0:34:11 If it thinks like, all right, maybe you should say that.
0:34:13 Oh, no, it does that.
0:34:13 There you go.
0:34:15 AI is going to save some marriages.
0:34:16 Awesome.
0:34:19 No, this has been a great conversation where, so I know you’ve
0:34:24 got a podcast and YouTube channel and you’re pretty prolific
0:34:28 over on Twitter slash X, uh, where should people go learn more
0:34:30 about you and check out what you got to offer?
0:34:30 Yeah.
0:34:34 Just go to, uh, my profile X, uh, for the, for the memes.
0:34:38 And then for my writing, um, go to, uh, creator.economy.so.
0:34:41 That’s where I share my writing and my YouTube channel is
0:34:41 like a new channel.
0:34:43 It’s like just getting started started.
0:34:44 Awesome.
0:34:46 Well, thank you so much for, for joining us today.
0:34:48 This is, this has been a great conversation and we really
0:34:49 appreciate you taking the time today.
0:34:50 All right guys.
0:34:51 Yeah, just pleasure being here.
0:34:52 Thanks Peter.
0:35:02 Yeah.
0:35:03 Yeah.
0:35:04 Yeah.
0:35:05 Yeah.
0:35:07 (gentle music)
0:35:17 [BLANK_AUDIO]
Episode 20: How can AI tools revolutionize your business and content creation strategies?
Matt Wolfe (https://twitter.com/mreflow) and Nathan Lands (https://twitter.com/NathanLands) dive into this with Peter Yang (https://twitter.com/petergyang), a principal product manager at Roblox and writer of creatoreconomy (https://creatoreconomy.so/).
Peter reveals actionable insights on leveraging AI tools for everything from generating YouTube thumbnails and headlines to summarizing transcripts and editing newsletters. With a deep dive into specific use cases and the balance between engaging content without falling into the trap of clickbait, the guys offer practical tips for integrating AI into your daily workflow. They also explore the broader implications of AI on job roles, productivity, and even personal relationships.
Check out The Next Wave YouTube Channel if you want to see Matt and Nathan on screen: https://lnk.to/thenextwavepd
Show Notes:
00:00 Using AI for practical everyday use cases.
04:10 Enthusiastic about Claude’s project feature for reusing content.
08:30 AI aids in summarizing information efficiently.
10:27 Free HubSpot guide helps understand and use AI.
15:27 Enthusiasm for AI art tools in YouTube.
18:38 EA uses AI to put 11,000 players’ faces.
21:31 Meta’s imagine me feature disrupts AI startups.
25:07 “OpenAI is the Kleenex of AI.”
26:38 Rethinking management: AI, small teams, big impact.
30:00 Future education: One teacher reaches millions globally.
32:36 Samsung Galaxy Buds 3 feature AI translation.
–
Mentions:
Get the free ChatGPT Bundle here https://clickhubspot.com/chatgpt
Roblox: https://www.roblox.com/
Claude AI: https://www.anthropic.com/
Superwhisper: https://www.superwhisper.app/
Runway: https://runwayml.com/
Suno AI: https://suno.ai/
EA Sports: https://www.ea.com/
—
Check Out Matt’s Stuff:
Future Tools – https://futuretools.beehiiv.com/
Blog – https://www.mattwolfe.com/
YouTube- https://www.youtube.com/@mreflow
—
Check Out Nathan’s Stuff:
Newsletter: https://news.lore.com/
Blog – https://lore.com/
The Next Wave is a HubSpot Original Podcast // Brought to you by The HubSpot Podcast Network // Production by Darren Clarke // Editing by Ezra Bakker Trupiano