AI Meets Marketing with Noah Brier
AI Meets Marketing with Noah Brier
How is AI changing the marketing industry? This week Paul and Rich welcome Noah Brier, a marketer and startup founder who’s excited about the ways AI could be used to solve the industry’s problems. Topics discussed include his early interest and adoption of generative AI tools, the types of problems his marketing clients are trying to tackle with AI, and why the tech industry seems to be missing the true potential of AI in its messaging.
Read the transcript
- Noah’s newsletter is here, and you can also find him on X and Linkedin.
- The Fast Company article on Noah’s initial brand mashup site.
- Nat Friedman and Daniel Gross on Stratechery.
- Noah is starting to work on a new project called Alephic.
Paul Ford: Hi, I’m Paul Ford.
Rich Ziade: And I’m Rich Ziade.
Paul: You sound a little subdued, Rich. What’s going on?
Rich: And I’m Rich Ziade!
Paul: There we go! Everybody get ready for this podcast. It’s called Reqless: R-E-Q-L-E-double-S. It’s the podcast about what happens when AI smashes into the software industry like a Cybertruck smashing into a brick wall.
Rich: Into the side of a Waffle House.
Paul: Yes.
Rich: Play that intro, Paul.
Paul: [makes a trumpet fanfare] Here we go.
[intro music]
Paul: All right, Rich—we did it. We finally did it. You ready? You know what we did?
Rich: We started a company.
Paul: Yeah, we’ve done that before.
Rich: Yeah.
Paul: Yeah, it goes great! No, we have—are you ready?
Rich: Uh huh.
Paul: A guest.
Rich: [whistles]
Paul: His name is Noah Brier, and he is here, hopefully we’re going to have him on a couple of times. He is someone who is out there, he is a marketing expert focused on sort of digital stuff, like, managing all the assets and working with the marketers and how they get, like, big companies got to tell their story, and there’s a zillion things that happen because of that.
Rich: Okay.
Paul: So he’s that guy, and he’s really an AI, and he’s starting to see how AI is changing all the big companies. And he’s having probably more direct conversations about marketing and sort of platforms and that kind of stuff than anybody else.
Rich: Yup.
Paul: And he gets into some interesting rooms. So we’re going to talk to him about what he is seeing from the big, broad world as to how people are understanding this technology and sort of how he came to figure this stuff out for himself.
Rich: Amazing. Let’s do it.
Paul: I think it’s going to be good.
Rich: Noah, welcome.
Noah Brier: Thank you.
Paul: Noah, how are you, my friend?
Noah: I am very good.
Paul: So I don’t even know how we met. I think it was just in the context of being business operators in New York City, [laughter] focused on technology and brand and stuff.
Noah: People who like to be on the Internet.
Paul: And so we’re really glad you’re here. Tell the people, give them a little bit of a story, because I think everybody is coming to AI with a lot of preconceptions and sort of on a journey, and some people are throwing their hands up and going, “Oh my God, this is the worst thing ever.” And some think the robots are going to come alive. You came from a background in sort of brand, and out of agency world, but with a real technology bent. And so, like, let me stop talking and just, like, give us, like, a minute on the story. How did you end up here in this room and what are you doing?
Noah: So I started my career in advertising agencies, and I was a creative director and a copywriter and I ran a strategy department, and then I taught myself to write code. And I started a software company in 2011 in New York City called Percolate. And we grew that and we sold it in 2019, and then I started another company, and I was taking a break from all startup things in 2022 in the summer, which was really good timing because it was also the time when OpenAI was starting to release APIs for GPT-3, and it was just starting to be available in a different kind of way and—
Paul: Percolate was not in any way an AI company.
Noah: Percolate was not in any way an AI company. It was a SaaS company. We built, we called it the system of record for marketing. It was a content-marketing platform. And we sold it to big brands and they managed their content workflow in there. But it wasn’t AI. There was no AI in there. It was regular old database and code and workflow.
Paul: So 2022, you’re like the rest of us, maybe you’ve heard about this stuff and it’s been kind of coming up and down for years, but oh my gosh, something is happening.
Rich: Well, it really wasn’t like the rest of us. The rest of us just kept saying, “Tell me a joke,” or, “What should I do about this weird notice I got from the IRS?” “Draw me a picture.”
Noah: Well, you couldn’t really even do that yet, because ChatGPT hadn’t come out yet.
Rich: Yeah.
Noah: So you were still sort of at the whim of what was available—
Rich: Right.
Noah: Via the API. So I was just like, I need a break. I need no employees. I need no investors. [laughter] I need to just sit and write some code and just sort of find my place in the world again. And I had been using, it’s like the extent of my sort of AI stuff was using Copilot in VS Code. Right? So I’d seen it, and I was pretty impressed. I thought it really made life a lot easier. I particularly had noticed—
Rich: As a coder?
Noah: As a coder—
Paul: So your home base when like, okay, I’m done with that last major, huge project, it got kind of big—was to go back to programming and like, just actually kind of get your hands dirty again.
Rich: It’s a comforting place.
Paul: Yeah, I get it. I fully get it.
Noah: It is a very comforting place. And yeah, you know, I think like a lot of peoples’ stories, as the company got bigger, I got further and further away from building and making things, and that’s where I feel at home. And so I just started playing around and started messing with this stuff and I didn’t really know what it did and I just started kind of writing some code and building things that popped into my head. The first thing I noticed was that I could get this thing to take raw text from a web page and turn it into any JSON I wanted. And I think actually I was one of the first people to publish a prompt that showed how to transform plain text into JSON.
Paul: For the audience, like, what’s that good for? What would that get you?
Noah: That is particularly good for web scraping. So if you’ve spent any time on the Internet trying to take a web page and figure out what it says, or actually, I was working on a thing where I was trying to extract pricing data from pricing pages.
Paul: Sure.
Noah: And so that is a very annoying—it’s not a hard programming problem, it’s just an extremely annoying programming problem without AI.
Rich: The variability of the web.
Noah: Just because the variability, you need to find the classes, you need to find the different pieces.
Rich: Yeah, yeah. Impossible.
Noah: It just sucks.
Paul: Yeah, there is no way to program and just be like, hey, find the price-like object. Sometimes it’ll work and you’ll get all optimistic, but then the next page on Amazon, it just doesn’t work at all. And you’re down this rabbit hole of like, how did they build this page? So you throw ChatGPT—or something—at one of these tools and it would just give you back.
Noah: It was GPT-3.
Paul: Okay.
Noah: The API, and I sort of figured out a prompt where I would give it a JSON TypeScript definition, and it would give me back reasonably close-to-valid JSON with the, it would say, you know, I could say, give me all the plans, and within a plan you have a price, and the price has a timeframe, and it just broke it all out, and I was like—
Paul: Now your eyes are lighting up. Why is this so exciting, as you’re telling me this? What was this moment?
Noah: It was magic. I mean, this is just like, if you have spent any time parsing unstructured data, it sucks. And it’s part of almost every project you ever have to do, and it always sucks. And you know, GPT-3 wasn’t great and it gave me bad JSON, I don’t know, say, 40% of the time. And I was still like, “Oh my God, this is the most magic thing I’ve ever used in my entire life.”
Paul: You know, what I compare it to is the experience of, when Citibike in New York City, the bike-share program rolled out e-bikes, people, they started to have weird crashes. They would accelerate, and people would be injured. And so they had to, they had to pull them all back. But people were like, “No, no, no, I want them back. Like, I don’t care. Like, it’s worth it to occasionally be really injured just so I can just shoot over the bridge that way.” [laughter]
And it’s—that’s how I compare it. It was just. It’s so, you’re just like, “Oh my God. This used to be grisly, horrible work that everybody had to do. And suddenly—” I’ll give you an example, like, creating an array that has all the months in it, or, like, in multiple languages. And I probably did that 500 times in my life, and I will never have to do it again. Right? And if you’ve done that, it’s like somebody builds a house from a seed. Like, you’re just like, “Whoa!” It’s so exciting. You had that conversion experience, which I think a lot of us have had with this. Then what happened?
Noah: So then I had had this idea in my head. One of the sort of notions I had with AI was like, it would be really fun to build something where it generated things that could be real but weren’t. And so first I did it with dog breeds. So I took all these dog breeds, and I, like, crossbred them, and I generated names for dogs and the description of the dog and the image of what the dog would look like. And I was like—
Paul: See, now we are deep in the, “I no longer work at the company” moment. [laughter]
Rich: Yeah, I’m gonna play.
Noah: And I’m just messing around. And then I built that, and I was like, that’s kind of cool. The next project I did, which really sort of, like, changed the trajectory of what I was doing, was I did something very similar, but for brand mashups. So, I think lots of people have seen brand mashups done with AI.
Paul: Sure.
Noah: But I built a whole site that let you take any two brands and mash them up. It was just sort of another idea I had in my head. And again, it started with this notion of, like, well, what if you could build kind of infinite pages on the web of things that could be real but aren’t.
Paul: Sure. So, like, Nike plus Honda, right?
Noah: Exactly.
Paul: Okay.
Noah: My very favorite version actually was a Grateful Dead Hermes scarf.
Paul: Honestly, it would work. It would sell.
Rich: Beautiful.
Paul: Yeah. Don Henley sang about that.
Noah: So I built that. Uh, it got picked up by Fast Company, who wrote about it, and in the conversation with Fast Company, I told them I was going to have an event for marketers about AI, even though I had no plan to do that. I just, I just sort of said it.
Paul: No, you are roughly our cohort. The reality is you can swing an event, if you’re like, all right, I can make 100 people show up in a room if I need to. We can—
Noah: Yeah. And it seemed like a good idea, and I did it genuinely because I was disappointed by the conversation—even then, this was 2022, and I was feeling disappointed by the way the industry was talking about it. But, you know, on the—
Rich: Talking about what?
Noah: AI and marketing. And it just, like, it felt like everybody was already onto these really boring things. Like, they were like, how do you do better SEO content with AI? Or like, when will AI replace creative directors?
Rich: It wasn’t creative.
Noah: It just wasn’t creative or interesting. And, you know, even at that time, my experience was already like, this is the most magical thing I’ve ever touched. Like, I’m producing at this incredible level. I feel amazing. I’m like, I feel like a superhero. I’m, like, writing code, building whatever I want, anytime I want.
Paul: I want to derail for one sec, because I struggle with this, too. So many of my beloved friends in the technology industry have seen this technology and gone, “It is an enormous thievery factory that can only be used for evil.” And I have the same experience you do, which is like, well, that can be said about a lot of tools, and I get why you might be saying it about this, but it also applies to lots of other things in our industry, and the upside can be immense for individuals. Like, individuals can be empowered through this stuff in ways they couldn’t before. How are you wrestling with that? I find it hard, frankly. Like, I go up and down.
Noah: Yeah, I find it hard, too. I mean, I think maybe I don’t find it that hard. I don’t really buy the thievery argument, at least on the tech side.
Paul: Mmm hmm.
Noah: I don’t believe that these things are doing something massively different than what we all do when we read things, like, the idea that they’re copying wholesale what comes out, you know, even the sort of hoops that the New York Times had to jump through in order to get it to reproduce exactly. Like, yes, it has that capability, but fundamentally it’s not really doing that. And so I’m a little on the tech side specifically, I think I have a little bit of a different feeling.
But, yeah, I mean, it’s just more than anything else been hard to ignore my own feeling of empowerment. Like, I’m not the best developer in the whole world. I taught myself to write code in 2008. I have always lived a life where I’m like, I can build a prototype and then if it gets good enough, I’ll bring on some people who are better at writing code than me.
Now I have this thing where like, I feel like I can do anything, partially because I can work alongside the AI, right? I can sort of, I’ll be like, “Oh, well, today I want to build an iOS app, or I want to do something else.” And I know how to get started and I know that I know enough about writing code that I know how to ask the questions. But probably more importantly, I just have this thing where it’s like, whatever problem I’m going to run into, ultimately I can figure out how to use AI to sort of like stitch two things together.
Rich: You have to have some technical proficiency. The fact that he found it appealing to play with the API rather than just mash into the prompt is a big deal. Like, you have to understand basic concepts. You know what I would love to see? I would love to see kids play with this. Like, I think kids can be taught basic programming knowledge and then say, well, that’s all you need for now. Actually go play with this incredible thing, and then let them play because they’re going to think more creatively and out of the box than someone that’s all about frameworks and libraries and DevOps and all the other stuff that slows down innovation, I think.
Noah: Yeah, I think, I mean, for me at least, I sort of settled on my stack now and it’s just life is so much easier. I just don’t worry about anything anymore.
Rich: What is your stack? Nerd out for us.
Paul: What are you doing? Well, tell us what you do and then tell us about your stack.
Noah: Okay, what I do now is I have been running an organization called Brxnd—B-R-X-N-D—.ai. We put on events, so I put an event on for about 200 marketers in New York City in May. We do that for brands too. And then I do consulting work. So I sort of like, work with brands in all sorts of different capacities, helping them both figure out kind of how to just utilize AI, what to do with it, but also sometimes kind of building and solving problems with and for them, writing code and building kind of multi-step workflows. I think the big focus I’ve had with Brxnd has just been like being hands on. Like, I’m trying to kind of differentiate what we’re doing with what everybody else is doing by actually doing things and not just talking about it. Like, the rule at the conference—
Paul: [overlapping] Sure.
Rich: [overlapping] Just writing papers and articles…
Noah: And the rule at the conference is nobody’s allowed to come pontificate. You’ve got to show stuff you really built.
Rich: Your conference?
Noah: My conference. Like, I don’t want to talk about the future. I don’t want to talk about AGI. Like, maybe AGI will come, maybe AGI won’t come. [laughter] My point of view is, if nothing changed from today, if we never got a better model than GPT-4 or Claude 3.5—
Rich: There’s ten years of invention.
Noah: Ten years of invention, easily.
Rich: 100%. 100%.
Noah: And so, you know, all that stuff’s exciting and interesting and whatever, but I generally don’t tune in because, I don’t know, I’m just too busy building whatever I want. Yesterday I built a thing again in, like, 30 minutes. I was like, I’d really like to be able to pull from a CRM and send it into Slack. But the annoying thing is you have to format everything. You have to figure out what the data structure is. I was like, what if I just skip that part and just send all the data to the GPT-4.0 and just ask it to make me a Slack message? And you know what? It worked perfectly.
Rich: Yeah. Yeah.
Noah: And so I took a project. The hard part of that project was not calling the API. It was figuring out how to write a templating language for all this data. I just eliminated it. I was just like, whatever. I’ll pay the 10th of a cent for every message to have it formatted by GPT-4.
Paul: Party at Noah’s house, is what I hear. Yeah!
Noah: That’s what it’s like on a Sunday.
Rich: Before you talk about what you’re doing for companies, very curious about how companies perceive AI, and how they think about it. Like, literally the first conversation—because it’s been such a consumer story up till now. But companies, I think, are starting to get wonder if they can help them and whatnot. Yeah. So tell us, tell us your impressions of, like, their reaction to AI and how they think about it.
Noah: Well, I think the first thing to know about what’s going on inside companies, at least in my experience, is that there’s an incredible amount of top-down pressure to use AI.
Rich: Why?
Noah: The CEO is getting pressure from outside investors and their—and again, my view is pretty marketing-centric. So I’ll talk about marketers. But I’m sure this is happening in other parts of the organization. But the CEO is getting market pressure, like, what’s your AI strategy? And so they go to the CMO and they’re like, what’s your AI strategy? And the CMO goes to the VP and they go, what’s your AI strategy? And the VP goes to their agency and they’re like, what’s our AI strategy?
Rich: Why do you think they’re saying that? To save money? To not fall behind competitively—
Paul: [overlapping] Aw, man—I think that, you know, you see an article in HBR and you get a little pressure from your board, and you get out that pair, of like, those children’s scissors and a paper clip, and you go drop it on the CMO’s desk.
Noah: I think Paul’s basically right. I think that’s it.
Rich: Okay, so if you—
Noah: I mean, I don’t think it’s—
Rich: If you pursued that and you said to that executive, “What do you mean by AI strategy?” They’ll just start sobbing? [laughing]
Noah: I think that they all mean different things and they’re getting different pressure. I think fundamentally, though, people have heard about it, they don’t really know exactly what it means. [laughter] And they know that it—people think it’s important, and it probably is important, and they’ve seen some evidence that it’s important, particularly if you’re a marketer.
Rich: Uh huh.
Noah: And so they’re like, hey, what should we be doing that with this? How is it going to change the way we work?
Rich: That’s not a ringing endorsement of executives in America.
Paul: [overlapping] No no, but hold on, here’s what’s different. This happened with blockchain. It happened with metaverse.
Rich: What are we doing about it?
Noah: No, it happened with mobile, it happened with social.
Paul: Yeah.
Noah: I mean, to be fair, like, important things tend to come this way, right? Like, there’s consumer interest.
Rich: Yeah. Yeah?
Noah: Like, I mean, I remember working with brands at Percolate on social stuff. Like, there was. Everybody was talking about social, they were talking about Facebook, they were talking about Twitter. Somebody was like, “Hey, what are we doing about Facebook? What are we doing about Twitter?”
Paul: Aw, those are the words: What are we doing about…? And now the entire, everybody’s, like, month is ruined.
Rich: I want a memo. Write me a two-pager.
Noah: Right.
Rich: Okay, okay.
Noah: And so, you know, I think to some extent, this is how the market works. Is it perfect? No, but it’s actually, like, a relatively efficient way to operate, right? You’re like, hey, there seems to be enough attention on this thing. We should probably at least explore what it means.
Rich: I’ll tell you what I have been hearing is, “I’ve had this problem. I think it would take a year. Do you think we can do it a lot faster?” There’s a lot of that just getting tossed out at you. “And I probably can’t afford it, but now there’s AI, and it seems really smart.” They talk about AI as if it’s like a guy named Joe down the street. “Do you think we can get it faster?”
Paul: Yeah.
Rich: Which also means cheaper.
Noah: I would say, to be fair, like, I’ve had less of that, and I’ve had more of, like, “We think there’s something important here. Can you help us just, like, figure out what that is?” You know, I mean, I had a funny one. I was in with an agency, talking to their exec team, I don’t know, a few months ago, and, they said to me, “Well, like, what are all the other agencies doing with AI?”
Rich: That’s the paranoia right there, right?
Noah: And I was like, “Nothing.”
Paul: Yeah.
Noah: And they’re like, “What do you mean, nothing? They’re all telling the brands that they’re doing all this AI stuff.” And I was like, “Well, yeah, I mean—”
Rich: That’s different. [laughing]
Noah: What they’re doing and what they’re saying are different things.
Rich: Exactly.
Noah: That’s pretty fundamental. So, yeah, I think that I have had a lot of questions of just at a really fundamental level, like, what is this? What does it mean? Have we been left behind? I think is a really big fear.
Rich: It’s a big fear. That’s a big fear.
Noah: They’re all like, “Hey, have we already missed the boat on AI?” And, I mean, which is such a totally crazy question to ask.
Rich: Yeah, it’s ridiculous.
Noah: We’re all still, you know, anybody who is deep in it, we’re all still out there just figuring out what it does on a day-to-day basis. And so the idea that you could be left behind already is crazy.
Paul: What are the things that either you’re doing or that you see brands doing with these technologies that are real, that are concrete? Because so much of it is really abstract. And a lot of the conversations, especially in the world of marketing, like, you have that no pontificating rule for a reason, in this world. It’s marketing. Like, people pontificate all the time. It’s part of the job. And so, like, what’s real? What could you do? What can I do? I am a big company trying to do, I don’t know, organized assets or get a message out, like, what, Noah comes in. What are you gonna tell me?
Noah: Yeah, so, what would I tell you?
Paul: First of all, get your checkbook out.
Noah: Yeah.
Paul: Second of all—
Noah: I tend to start with just, what does this do and what doesn’t it do. Like, I try to sort of ground everybody in kind of some foundational pieces and, you know, I think when I think about AI, when I look at the way I build things and the way I’ve used things, you know, the most fundamental thing it does amazingly well is transform data. This is our JSON thing at the beginning, right?
Paul: Okay.
Noah: So any problem you have where like, you have some people who need to transform data from one format to another, you know—
Rich: Integrations.
Noah: Integrations, anything—
Rich: Migrations, yeah.
Noah: You know, I want to take the CRM stuff and I want to put it into Slack. Like, those kinds of problems which are a huge part of how we all work. You can just kind of solve those things—
Rich: Quickly.
Noah: Very quickly.
Paul: Each one used to be weeks of developer time, frankly, like, that’s just how that was.
Noah: Yes, of course.
Paul: Okay.
Noah: What is difficult about it, though, is, honestly, I sort of end up having these conversations and I’m like, I say one of two things. One, let’s just talk about the sort of landscape of challenges that you have, right? Where are the kinds of places, and let’s try to line those up with what the AI is really good at. Right? So it’s like the kinds of things it’s really good at are, it’s really good at operating in places where you have lots of examples. So it’s like, I’m working now with someone on a proposal writer.
Paul: Okay, sure, sure.
Noah: You have lots of example proposals. A lot of what you do when you write a proposal is you take a bunch of data that is unstructured and you try to turn it into a structure, which is like a scope—
Paul: Right.
Noah: —and allocations and all these things, and you have lots of examples. It’s actually like a really good problem. And also it contains this sort of critical piece, which is you don’t actually need the AI to generate it and send it to the client. You just need the AI to get you to 50% or 60% or 40% or whatever. Like, if it can do that, it’s already saved you such a tremendous amount of time.
Rich: I mean, he just, he just pitched our value proposition around getting you to a good place, then handing it off to you to finish it off. Right? Versus like, am I getting a fully baked bundt cake? Nothing against punt cakes, just—
Paul: Everybody likes bundt cakes.
Rich: Yeah.
Paul: Let me give you a nerdy one. I’ll throw a project at you, which is, I’m a big government agency, I work directly with…millions of people check my website, I’m making this up. But, you know, it’s got, and I have 50,000 PDFs and other documents that I need to route them to, and I need to sort of measure how that’s going and so on. This, and this is, this comes out of my advertising and marketing budget, is to improve the relationship. Right? So that feels to me like a very AI-ish thing, Noah. If I gave you all those documents, what could you do to make a better experience for people?
Noah: Well, I mean, one of the first and sort of most amazing things, this was another one of those, like the JSON one is, and this is a little bit of a different thread, but the first time I used embeddings and vector search.
Paul: Yeah, so tell people what those are.
Noah: Okay, so embeddings essentially are a different kind of AI model. It’s utilized in ChatGPT, but they essentially allow you to kind of take the AI’s conceptual understanding of a world or set of words and it maps it into vector space. And so what that allows you to do is say the word apple and the word orange and the word pear all have a relationship between each other and we can measure the distance and the closer distances, they have more to do with each other. They’re more closely aligned.
What’s amazing about this, and again, speaking of problems that you spent a ton of time on, if you’ve ever written any code, basically, is like, dealing with search is hard. And what you can do by just getting, taking those PDFs, grabbing all the content out of them and embedding them, grabbing the embeddings, grabbing the vector, and then allowing people to search the vector space is just, like, a really high-quality search engine out of the box.
Paul: So if we go back to my example, let’s say it was the EPA, if somebody searches “floods,” it’s also going to find water damage and it’s going to find—and you’re going to be able to kind of package, and you’re, you also, I mean, one of the things that’s interesting about this stuff, right, is like you can extract a lot of context and rewrite things as you go and then search that as well. So everything can get incrementally smarter in a way that used to kind of just belong to big players.
Rich: Look, if I wanted good search against the body of data, it’s really hard to get it to be really good and smart. You can do like basic stuff, but holy hell, like, forget the word AI, forget vectors, and forget anything else. You just got search. Jesus, I mean, that’s—
Noah: In 15 minutes.
Rich: In 15 minutes, like, can we pause and take that in? Before we get to general intelligence. We just wrestled away some profoundly powerful technology away from, like, three companies that dominated it in the world and put it in your hands. I don’t know why we aren’t seeing that kind of explosion. Why do you think that is? Why isn’t that the story right now?
Noah: I think that the reality is sort of inside, like at the actual point of execution with this, is the story—and in fact, like, the one we’re talking about now is commonly called RAG, retrieval-augmented generation.
Rich: Yeah.
Noah: And this is probably the biggest enterprise AI story already because—
Rich: RAG has taken hold, yeah.
Noah: It’s such an obvious thing. You’re like, hey, we have a ton of documents. Nobody likes to read them. We try to get our employees to read their, you know, employee agreement in PDFs, and we can actually just very simply built a thing that lets them say, how many days a year do I get off? And it just answers their question instead of pointing them to page 17 of the 60 page employee manual.
Paul: Why is the conversation happening in such a way that is actually divorced from the real value? I think it’s because the main AI companies are focused on just getting billions in and telling a story to the market that is transformational at a level that the internet was transformational. If suddenly Sam Altman, um, was like, look, we’re not going to build a new consciousness. We are going to make sure that onboarding is better for hundreds or thousands of companies. You think that the entire market is going to lose its mind with excitement?
Rich: Yeah.
Paul: So that’s where we are. And then on the other side, I think you just have people who are just seeing Vulcan’s capitalist hammer smash into the world of cool art online and get real, and they’re upset. Then in the middle, you have the nerds like us who are like, gosh, it’s great to solve enterprise problems so that peoples’ lives are not boring and excruciating all the time.
Rich: I think it’s, I think, you know what it is? The utility is so vast that people jump to the controversial and sort of really emotional stories that come out of it. Like, because it’s, look, this stuff is interesting to us.
Paul: Yeah.
Rich: And I don’t know if this particular podcast episode is going to go viral. I doubt it.
Paul: No. This—
Rich: We’re kind of nerding out a little bit.
Paul: This is like fixing your own car.
Rich: Yeah.
Paul: That’s the kind of people we are.
Noah: I do think that’s a hugely important point, which is like, you know, one of the challenges I have when people ask me, it’s like, a lot of the best use cases are really boring. Like, Paul is very excited about this thing being able to transform anything into JSON. But I recognize that, like, that’s a relatively small universe of people who are like, we are super stoked that we now have a machine—
Paul: That alone, I can point to, like, seven companies, like, let’s say, like, MuleSoft, where that’s their entire value proposition, and they’re worth tens, hundreds of billions of dollars in aggregate. And instead, this is like a feature of the new robot that’s kind of built in.
Noah: I’m excited about it.
Paul: You know what it reminds me of, actually? Here’s an—remember how mobile kind of took away cameras? Like, cameras just, it’s a totally different marketplace. That’s what this is like. It’s like all those features that are on your phone that kind of ate whole industries, and then those industries had to reinvent themselves. It’s not like, you can still go out and buy a really nice Nikon. It’s out there. But that’s, I think it feels like a lot of things that used to be culturally a lot of work are now features in this giant system.
Noah: I also think the other piece we just have to remember is it’s like we’re all just kind of feeling our way through.
Paul: Boy are we.
Noah: I’ve really enjoyed the Stratechery episodes he’s done with Nat Friedman and Daniel Gross.
Paul: Oh, from GitHub. Yeah, yeah, yeah.
Noah: They said something in, I think, December of 22 episode where they were like, hey, you know, one of the things you have to understand is you have these people who are, you know, we were calling them prompt engineers, I think, more than we are, than we are now. But to a large extent, they are learning about these models in ways that the creators of the models never imagined. They’re doing things and learning things. And that’s, I think maybe the biggest piece is this is just such a strange bit of technology. We’ve all grown up in this world where computers are deterministic machines, right?
Paul: And this is pure trial and error.
Noah: And these are probabilistic computers, right? Like, we, none of us have a sense for what they are. And, and, you know, I mean, I think we are like, every day I spend tons of time with these things, and every day I figure out something that I hadn’t thought about doing before. And, and I think that’s kind of just what’s going on in the world. And I mean—
Rich: I don’t think the world’s internalized that yet.
Noah: No, I think it’s impossible—
Paul: No, because if you go on social media, there are ads where people are saying, it’s almost too late. Now you have to learn AI.
Rich: Yeah, yeah.
Paul: So let’s do two things. One is I want you to talk about the kind of work you’re going to be doing. Like, where do you see what are you going to be doing over the next year?
Noah: Yeah, sure. So I think over the last six months, the big transformation, I’ve seen beyond, I do lots of events, I do lots of sort of education with brands. The biggest change I’ve seen is brands bringing me really hard problems and not asking me, like, do it faster and cheaper, but being like, actually we have not been able to figure out how to solve this. Like, is there some way to solve this using AI?
Rich: I mean, what a fun entry point.
Noah: It’s awesome, it’s great.
Rich: It’s a fun way to, like, you’re essentially, they’re not giving you hard requirements, they’re not saying you must use a particular technology, they’re just saying, I’ve got a problem, and I just never thought I could ever really deal with it.
Paul: No, this isn’t $100 million problem. With a $50,000 budget, what could we do?
Noah: Right? And that’s exactly, and you know, I got one recently about content strategy that involved, you know, looking at 20,000 articles from the competitive set. And like, I, it’s just wild to me because in a short amount of time you can—I mean, it’s sort of like we can do things cheaper and better. And that’s not usually the combination that you get.
Paul: Yeah, there’s usually just like kind of faster.
Noah: Right. We got used to faster and then we would sort of extrapolate from there. And now it’s cheaper, better, and faster, and it’s, and actually sometimes slower but cheaper and better.
And the other thing that’s really interesting about it is when you solve—so I’ve sort of found my kind of sweet spot, I’ve realized is like, I have found that you can produce kind of world-class output from these models if you take people who know a lot about something, like people who work inside a brand, and you combine them with code and AI, you sort of like build this triangle, right? Code, AI, and the sort of individual expertise inside the organization, and you can build absolutely amazing things to solve almost any problem.
And if you have access to the right people who know how to solve those problems in the organization and you can write the code—the reason the code matters is because I think there’s a sense that probably, because of the AGI conversation, that AI is a thing where you just put in a question and it gives you the answer fully formed. And the reality is that it can do that, but that answer is going to kind of suck. And if you want an answer that doesn’t kind of suck, you almost always need to build a multi-step workflow. Right? You need to sort of chain prompts together, you need to include outside data, you need to do all these different pieces, right? And if you want to solve problems that way, you need code.
Rich: This goes back to sort of a theme we’re starting to see, which is it leaves me kind of optimistic. I’m glad that we still have to do some of the work. You know, what’s so interesting about your journey is that it’s born out of conversations with people. And, you know, there’s so many AI startups right now, they’re like, I’m going to automate this and I’m going to automate that and I’m going to, and there’s hundreds of little, like, little players that are like, I’m going to make, I’m going to get rid of those steps that you deal with and that’ll be that. You just hit a button.
You could have holed up, Noah. But I think what’s interesting about your story is that you went out and started meeting people. Like, okay, I think there’s something really interesting here. I’m going to go have a bunch of coffees and see what’s up.
Paul: So wait, wait, what comes next? You are now the marketing consultant for brands around AI. Like, that’s your, that’s the plan?
Noah: I wouldn’t say there’s a plan because I like, part of what I’m enjoying is living life without one. [laughter] But I have been, I’m gonna keep putting on the events just because, like, I really like it to the point. Like, I, this is exciting for me. I’m just like, having a good time and I love being around people who are excited about it and want to talk about actually building things. So I’m gonna keep doing that. I’m very excited about these strategic problems. And so I’m working with more and more brands, and I think that’s a big piece of it is like solving problems using this sort of triangle where you have code, you have AI, and you have the expertise of people.
Paul: Let’s say people want to get started. Let’s finish this, I want to finish with this. They want to get started. They want to figure this world out a little bit, talk a little bit about your, what’s your stack? And maybe like what’s a good starter stack here?
Noah: I’ll just say like, generally, I think when people ask me this question from a non-technical point of view, I think there’s a lot of people searching for the perfect tool. And the reality is just go sign up for Claude Pro or OpenAI ChatGPT Pro and just play around with it.
Paul: Claude’s a better programmer. I’ve been using it more.
Noah: Claude’s a better programmer and a better writer, I would argue.
Paul: Yeah, I think that’s right.
Noah: There are still certain things I prefer ChatGPT for.
Paul: So are these kind of your—when you’re making the JSON out of the documents, are you using these platforms as APIs?
Noah: I’m mostly using those platforms. I almost exclusively use those two APIs. I think that there’s lots of talk about other models. I bought a super fancy MacBook Pro so I could run things locally—
Paul: Like Llama.
Noah: Yeah, and I sometimes play with those things. Ollama is great for running on your laptop. But from an API standpoint, I almost exclusively use GPT 4.0 now and Claude 3.5 Sonnet, depending on what I want to do. The biggest difference honestly, between those two things is that OpenAI APIs have a JSON mode, which allows you to get guaranteed JSON output. And since a huge amount of what I do with these models, even when I’m not doing coding stuff, is, I still ask for JSON so that I can parse it.
Paul: No, we went through this, it makes a big difference. Just to note, to guarantee that something.
Rich: A lot of Aboard’s platform—
Paul: Yeah.
Rich: Relying on that.
Noah: They just released a new update that allows you to pass in schemas and get guaranteed schema output.
Rich: Very cool.
Noah: So, I mean, it’s pretty amazing. That’s so the sort of core, if you’re not writing code, just like don’t overthink it, just start trying to play and use these things and just find the ways to use it. Like at my first conference I talked about it, there’s a German word called Fingerspitzengefühl.
Paul: Not that kind of podcast, dude.
Noah: Which is about building fingertip feelings.
Paul: Oh, okay.
Noah: And that’s what I think where we all are. On the coding side, I am a big TypeScript fan.
Paul: Ah.
Noah: And so I actually have found—so TypeScript Serverless AI has been really magical for me. TypeScript, because I’m a kind of crappy programmer and TypeScript makes me be a little less crappy.
Paul: It forces you to think about, yeah, what is my data, for people who don’t know, it makes you, yeah, you have to organize your thoughts.
Noah: Serverless because it’s easy and I can’t stand servers.
Paul: Yep.
Noah: And then the AI. So I use, Next.js almost exclusively because I can write frontend and backend code in the same framework. And I just don’t need to overthink it. I don’t go look at other frameworks. I don’t worry about it. I’m just like, this thing works for me. I’m going to keep using it.
Paul: Well, this is one of the benefits of wisdom and age, right? You’re just like, I’m sure there’s other great stuff going on, but I don’t care.
Rich: There’s always great stuff going on.
Paul: So. okay.
Rich: Noah, we gotta have you back on.
Paul: Yeah, we really do.
Rich: There’s always stuff to talk about.
Paul: If people want to get in touch, what do they do?
Noah: They can go to Brxnd.ai, B-R-X-N-D.ai. I also have a newsletter at newsletter.brxnd.ai.
Paul: It’s a good newsletter. People should check it out. All right, and Rich, what if people want to talk to us?
Rich: Why would they talk to us, Paul? What company did we found together?
Paul: It’s called Aboard. And Aboard is an—oh my God, I’m gonna say it—AI-powered software development framework that anyone can sort of pick up and use. But what we do is we work with organizations. They tell us their problems. We spin up software in record time.
Rich: Fast.
Paul: Fast. So instead of running around gathering a lot of requirements, you can just get a prototype on screen, figure out what you want to do.
Rich: Yep.
Paul: And what we’re bringing to this is decades of software development experience.
Rich: Hundreds of years.
Paul: So if you’re a midsize company and you want to be able to have code that is as good as the ones that the giants have.
Rich: We love big companies too, Paul.
Paul: No, I know, I know. But if you’re like, I can’t pay $2 trillion and my firstborn daughter for Salesforce.
Rich: Yeah. Hit us up.
Paul: Check out Aboard. Hello@aboard.com.
[outro music]