Paul Ford: Hi, I’m Paul Ford.

Rich Ziade: And I’m Rich Ziade.

Paul: And this is Reqless, the podcast about how AI is changing the world of software, and jeez, is it. Let’s just play that theme song and get right into a big subject.

Rich: Let’s do it.

[intro music]

Paul: Richard.

Rich: Yo.

Paul: How you doing today?

Rich: I’m doing well.

Paul: Yeah, me too. I’m doing okay. I biked in. It’s a beautiful day.

Rich: I worked out. Beautiful day.

Paul: I should be really clear. I e-biked in. I’m getting—

Rich: Yeah…

Paul: I got the Citi e-bike.

Rich: Let’s be honest.

Paul: Whew! Boy—

Rich: We are honest citizens.

Paul: You know, Clinton Street in—

Rich: I do.

Paul: —in Brooklyn? Just kind of goes up a ways?

Rich: Yeah.

Paul: And it’s kind of a chore. It’s not a bad chore. But it’s a little bit of a chore. It’s, like, a real mild incline—but not on a Citi e-bike. Just, woooooo…

Rich: You need that nudge.

Paul: Wooooo—that was my life. Anyway, that’s not what we’re here to talk about. That’s just all I can think about. But what we’re here to talk about: You know, OpenAI. Big company. Sam Altman runs it. It’s going to create artificial super babies. And everybody’s weird and right-wing now. It’s all very confused in the, in the tech industry. But they went and they did something that, that tech companies often do.

Rich: Oh, what’s that?

Paul: They are trying to buy Windsurf for zillions of dollars.

Rich: Okay, what is Windsurf?

Paul: It’s one of those tools that helps you code. Like, and they, I think they also tried to buy Cursor and Cursor said no.

Rich: Yeah.

Paul: Which is another one. So it’s one of those things where like, you open up the, the development environment and it says,”Hey, what are we coding today?” And you’re like, “I gotta do this, that, and the other.” And it’s like, “Hey, I did it over here for you. Want me to run it?” And you know, it’s your buddy. It’s your buddy programmer friend.

Rich: Yeah.

Paul: It’s what we talk about on the podcast all the time.

Rich: Yeah. So these, it’s worth noting these tools, the code-assist tools, tools that, like, improve the productivity and help engineers get work done. They really actually are a profound productivity breakthrough. Like, it’s an invention that has landed in the engineering space that I don’t know if anyone won’t be using them in the future, right? They’re immensely, immensely useful.

Paul: I think that’s right. They’ll be part of programming. And I don’t know if AI art generation will be part of illustration in the future because it’s a little tacky. Like, we stopped doing it on our website.

Rich: Yeah, yeah.

Paul: Because, because it was like, not because we were a great moral cause, but because it kind of sucked. And it’s got that, that weird sheen. There’s that AI glaze that everything has.

Rich: I think if you are an engineer and you understand the ins and outs of code and how code should be glued together, these tools are incredible. I think if you’re a non-engineer and you’re just messing around, this is the whole vibe coding trend, I think it’s a different… Different set of challenges kick in. But the Cursors and Windsurfs of the world, which essentially glue on to the environments that their work engineers are working within anyway. It’s a game changer again. And that’s not me tooting their horn. It just is. [laughing]

Paul: No, there’s a, there’s a marketing video out I saw today. It’s Satya Nadella, the CEO of Microsoft.

Rich: Mmm hmm.

Paul: And so he recreated the very first product—don’t worry about this, what this means—but the very first product that Microsoft came out with was BASIC for the Altair computer.

Rich: Mmm hmm.

Paul: And he recreated it with, like, a virtual Altair in 10 minutes. And it took Bill Gates and, was it Ballmer? No, it was Paul—

Rich: Paul Allen and Bill Gates. It took them months.

Paul: It took him about—yeah, it took them about six weeks. And it was like this incredibly accelerated, super-genius thing. They did it on paper tape. And Nadella just kind of sat down, was like, “Hey, can you do this for me?” And it’s exactly this sort of like there’s tons of prior art about how it would all work. And so…

Rich: Yeah.

Paul: It’s exactly the thing. So, so basically this is the world, that’s Copilot. He’s, you know, they own GitHub. GitHub has Copilot. This is the world of software engineering kind of forevermore. I don’t think we’re ever going to walk back from it.

Rich: I think that’s right.

Paul: So here we are, here’s this moment, and I’m going to ask you a foundational question. That’s what I want to talk about for the next few minutes, and then we can all go back about our days.

Rich: Mmm hmm.

Paul: If AI is so fricking good at coding and can code really, really fast and it has all these systems and it can really kind of get everything done for you, why in God’s name would you spend billions of dollars for a coding-assist product that already kind of calls out to your APIs? Why would you buy such a thing for so much money if you could take a team of five and go build a whole new one using AI? You’re an AI company. What are they thinking?

Rich: Well, I think there’s a couple things going on. First off, I think the productivity boost and the adoption rate. And it’s worth noting, both Cursor and Windsurf charge money. They’re not free tools. There are free tools out there, but they charge money.

Paul: They really have to because they’re talking to these APIs that also charge money, like Claude…

Rich: That’s right. What, what is happening here is that OpenAI is coming to realize that the retail end of the offering has taken hold really, really fast, that companies like Cursor and Windsurf have a huge customer base already. And it’s been like two years, three years or whatever it is. And—

Paul: Buying—so they’re buying users, that’s one thing.

Rich: But they’re also buying—you got to remember, OpenAI, they love to trumpet how they are the PhD geniuses who are making the brain.

Paul: Mmm hmm.

Rich: And the truth is, wrapping all that stuff up in product that people want to take a credit card out for, that integrates seamlessly in, in very familiar environments that they work within? That’s, that, that’s pedestrian shit as far as OpenAI is concerned.

Paul: I mean, when, when you look at how they name their product, I think that—

Rich: Exactly, exactly.

Paul: GPT-4, blah, blah, blah. And then you got 3o and o3. And it’s just an absolute boondoggle to figure out what you’re using, when, and why.

Rich: That’s right. And the likes of Cursor and Windsurf, it’s born out of the engineering community. They understand the tools engineers love and want to use. And they augmented those tools, and that resonated with an audience that understands that language. It’s a particular community. And they’re making a lot of money. Like today, they’re making a lot of money.

And OpenAI, for OpenAI to decide, we are going to launch the code-assist wing of OpenAI? Things have moved so fast, and they probably view it as a distraction. And if anything, it’s like, I don’t need to create another company that only worries about a particular cohort of people that use tools in a very specific way. I’m changing the world. Right? Like, so—and, and the revenue is strong. Like, these companies are not like they have a handful of customers and some IP. They have big, big customer bases already, right? And they want that.

And I think this highlights another thing, which is the commoditization of these tools. If I’m not mistaken, Cursor doesn’t care which LLM you use. You can sort of pull a drop down and plug its suite of tools into any of them. It’s verticalization, right? You’re essentially saying, now Windsurf, if there was a drop down—they’re acquiring Windsurf, by the way, Cursor turned them down.

Paul: Yeah, that’s right.

Rich: If there was a drop down in Windsurf, it’s gonna go away. It’ll probably happen very subtly and slowly over time. But you don’t buy a company so you have a drop down to use any of the other LLMs. You buy a company so that you can find efficiencies and whatnot. I think. They may decide to let it go.

Paul: Well, they will if it’s sort of, like, Google where they’re, like, anticipating antitrust.

Rich: Yes, that’s true. There’s another thing they can do, which is, “Hey, we’re going to launch key code-assist features that will only surface on Windsurf first.” You could still use the other ones. Essentially, there’s now a symbiotic relationship between the LLM part of the company and the code-assist engineering part of the company.

Paul: So that, that is my bet. Right?

Rich: Yeah.

Paul: Like that they, you know, ChatGPT will come out with “Super Python Mode” and Windsurf will be like, “Oh, do you—”

Rich: You’ll need windsurf.

Paul: Yeah.

Rich; Yeah, that’s right.

Paul: If you need, if you use Windsurf, you can get five times more effective with Python.

Rich: Yeah.

Paul: Or game programmer mode or whatever, right? Okay, so, but let me go back to my original question, which, you know, that is a very sensible response. It’s a business.

Rich: Thank you.

Paul: It needs to drive revenue. Here are this, here’s this thing that has a lot of users, probably a very talented team has built it. It’s already built on the technology that OpenAI produces. So let’s go, let’s all get in a room together and let’s make even more money and even more success.

But let’s go back to my original question, which is I’m not supposed to need humans. I’m supposed to be able to accelerate everything and I’m supposed to be able to just spit out code and spit out productivity. But things aren’t moving so fast that I have to spend a billion dollars to save six months.

I should be able, if I’m ChatGPT, to go get those tens of millions of users, and I should be able to get my bot army to enable this, right? And I really want to call this out, because that’s the story. And yet the action is I’m going to spend a couple billion dollars over here and I’m going to go and, just like a software company got acquired in 1996, I’m going to go ahead and do that because I want that revenue, I want that growth, and I want those users. So is there anything radically different about where we are now compared to where we were 20, 30 years ago?

Rich: No. Humans pay money for things. The greatest AI invention doesn’t result in AI paying another a AI product money. That’s weird.

Paul: I mean, this is kind of the fantasy, though. The fantasy is that bots with blockchain marketplaces will simply transact all day.

Rich: Okay, that’s a different podcast, bro. I can see agents unleashed into markets where only agents—like, when I say agents, I mean AI agents are allowed, and they have long lunches together and then they go back to the trading desk and they trade with each other, and then they meet for drinks at a steakhouse. Like, you could do that. And I guess someone could make money, say, well, what do you do for a living? “I have 300 AI agents that trade stock for me.” I don’t know.

Paul: I mean, Richard, it’s not another podcast. That was the Sarah Chips podcast.

Rich: Oh, yeah, that’s true. [laughter] Yeah, I don’t know! I mean, look, I’m not—we can go futurist here. Hard. We could do that. But for now, humans are still very much, especially specialized, skilled humans, like computer engineers, are still very, very much in the mix, and there’s a massive amount of opportunity to make them more productive.

Now, will we eventually show them the door and tell them, look, just take a, take a day off the machine will do this today. Probably, in some, some respects. I have, I think I’d love to talk at another, on another podcast about the distinction between generating output and actually being thoughtful and internalizing intent and doing things in a bigger way. And I think that’s a very uniquely human thing. But, you know—

Paul: Don’t wait for the next podcast. I made my point, which is…

Rich: Yeah.

Paul: And a lot of times I come in as the soft liberal arts type, and I’m like, [nebbishy voice] “You still need humans. Humans are very important.” Right?

Rich: Yeah.

Paul: But wait. But wait. So—but I’m going to tell you as someone who is an industry observer and is objectively looking at the capabilities of this technology, what this technology does is generate artifacts that seem a lot like human artifacts. An artifact like a product-requirement document that is pretty close to what a human might generate is something that can produce action. The next action is to start building the product.

Rich: Yes.

Paul: A PowerPoint that says, we should do this instead of that. Now people can look at that PowerPoint and they can say, “What action will I take?”

Rich: Yeah.

Paul: And so we’re building this new world where we generate these artifacts, and the artifacts lead to people kind of doing things based on them. And it seems really magical because it used to take four months for someone to kind of organize their thinking around that PowerPoint.

Rich: That’s right.

Paul: And now, Deep Research will make you a list of all the things that learn by looking at the web. And it’s a pretty good list.

Rich: Yes.

Paul: Okay? But the humans come in and take action, and then there’s this big AGI conversation that just eats up all the oxygen in the room. Very similar to the blockchain conversation where people go, “But soon you won’t need the humans. It’ll just, it’ll just do it itself.” That does not appear to be happening. Objectively, everyone is starting to see the outer limits of LLMs, and they’re saying, “Oh, I guess we’re still going to need those people,” to the point that the giant AI company that’s supposed to be spending money with some sensibility is willing to spend billions of dollars to get someone who can do what it does just a little bit better. It’s an interface to their equipment.

Rich: What we’re really talking about here is promotions.

Paul: Mmm hmm.

Rich: Like, we keep calling them assistants and, you know, buddies, and we give them cute names. But there may come a time where if I don’t address the, you know, AI with, you know, a mister or miss, it won’t respond because it commands respect. I don’t know. It’s going to get to a point where it’s going to tell me what to do with my day, rather than me telling it what to do. I, you know—

Paul: I so look forward to that moment. Let me stop you [laughter] before we talk about you actually listening to anyone.

Rich: [laughing] I’m not talking about me. I’m not talking about me. This isn’t about me.

Paul: Rich, let me just reframe this, actually, back to the audience and kind of what I think people think about.

Rich: Yeah.

Paul: Okay? We know, we’ve actually had a lot of conversations about what engineers are going to do with this technology.

Rich: Yes.

Paul: Okay? But clearly, product is still in the mix. Okay?

Rich: Product.

Paul: Product managers are going to continue to exist.

Rich: Product managers, I would argue, are going to become much more critical.

Paul: So let’s go ahead and I’ll be a product manager and you can interview me and you can ask me some questions about AI and tell me how you need me to think about AI in order to be a good product manager in the future.

Rich: I wouldn’t do that. I wouldn’t do that first. Maybe I’ll get to AI eventually.

Paul: Mmm hmm.

Rich: Look—

Paul: So you don’t think that—you still want the same PM skills?

Rich: Well, look, let’s talk about, there’s PMs that can do a job and then there’s the best PMs. And let me tell you about the best PMs. The best PMs do one thing better than anyone else. Before—when we talk about AI, we often talk about execution, about generating stuff that gets you to the final thing, right? But what does a great PM do? I’ll tell you what a great PM does. A great PM takes in information from people. They may be executives, they may be managers, they may be customers. They take in information and they triangulate on the motivations and intent of what everyone seems to be surfacing. They don’t—

Paul: That was the most corporate statement I’ve ever heard.

Rich: Okay.

Paul: Make that into human language.

Rich: Okay. I’ll give you a—

Paul: Triangulate on the intent of what seems to be surfacing. You said those words.

Rich: Yes, I did. And I meant that.

Paul: I ate with you.

Rich: Yes.

Paul: We had—

Rich: Here’s the thing.

Paul: Appetizers yesterday. Okay.

Rich: If someone comes to me and says, “I need a shack in my backyard,” right? And it’s 10 by 12 and it has to be made out of metal? You don’t need a product manager. You just need a project manager. You just need someone to go buy the metal, to draw out the dimensions, and cut up the metal, and nail it together, and there’s your shack. You essentially went and executed a task that was put in front of you. Now, there may have been some ambiguities and you may have come back with some questions and, like, “Do you need it to be rainproof?” Et cetera, et cetera. Okay? A great product manager pauses and says, “What’s the shack for?”

Paul: Mmm.

Rich: “Well, I’ve got stuff strewn about all over my backyard.” “Okay. Why—” You know, that interrogation and that sort of deeper dive into the motivations of what people want and what people need is what separates an average consultant from an exceptional consultant. An average product manager from a great product manager.

And what does that mean? What that means is very often people think they know what they want. Right? And have particular needs and goals. But a great product manager is actually internal—it’s like a great architect who’s building your house. They don’t just say, “Do you want a bathroom?” They say, “What matters to you?” Right? That aspect of it, that sort of internalization—now eventually they have to turn that into a real blueprint and actually build the thing and show you that I’ve internalized what you want and what you need.

Why am I saying all this? Because before you go and work in Notion or whatever you’re going to use to build, whatever you’re going to build, if you don’t get this part right, and if you’re not good at this part, you’re not a great product manager. And I don’t know how or if AI is going to have that conversation to understand what people want.

Paul: Well, let me play it back to you, which is, okay, let’s, let’s make AI the architect. And…

Rich: Okay.

Paul: You’ve trained it, you’ve trained it to be good architect.

Rich: Mmm hmm.

Paul: And someone’s like, “I really need a bathroom on the second floor.” AI as architect doesn’t care. It can kind of be trained to produce patterns.

Rich: Yeah.

Paul: It doesn’t care. And so if it might end up producing 30 gold toilets for you to build.

Rich: Yeah.

Paul: It might tell you that you don’t really need a bathroom. It might make a bathroom for an extraterrestrial body, just a long tube, nothing else.

Rich: Yep.

Paul: Like, it’s not—you know, there’s a funny aspect to the PM that you’re describing. I know this PM very well, like, as an abstract entity, and they’re, here’s why they’re rare. The domain knowledge is not enough. You might know, like, hey, here’s every single tool. Here’s how we build it. Here’s what it will cost. But—

Rich: Yeah. They’re so rare. They’re so rare.

Paul: But then when the person says, “We need to get Salesforce in here.” And the PM goes, “Why?” It can’t be confrontational. It has to actually be a connection.

Rich: That’s right.

Paul: And they’re cutting scope and they’re doing things and they’re moving and manipulating in a social, relationship-driven context. And you can’t Q&A your way out of that. You actually, that is the relationship. You build the relationship and then the product emerges out of the desire to connect to the person.

AI can simulate that. That’s, this is what’s going to really throw people is they’re going to find that they can simulate that all day long. But that last little bit is just going to be out of reach. And why? Because the buyer was looking for that actual connection and they weren’t even able to articulate it. They’re like, “I know the thing I want, but I don’t fully, I can’t express it to you.”

And the bot can help you, it can absolutely help you. But that last little bit is still about sort of primate dynamics in an environment, and it’s very abstract and people are going to hear this and they’re going to say, “He’s full of it. Absolutely no way. Claude’s going to be able to do that in six months.” And God bless.

Rich: Yeah. Yeah, yeah, yeah. I mean, look, I think there’s another component, by the way, which is constraints. A good product manager also asks, “When do you need this by? And what is your budget? And what kind of team do we deploy?” Like, one of the weaknesses of AI, right, is it has infinite budget and infinite time. Like, it just goes buck wild. And that isn’t helpful, actually, because nailing it within constraints is actually a huge part of really delivering something exceptional. It’s like, “Wow, not only did it come in within budget, but you actually didn’t even bother with the things that would have been superfluous about even having that. Like, I’m glad we didn’t do those things.”

That—what you take out and what you, the limits you put on, money and time and whatnot, are huge components around this, which, by the way, inevitably bear down. Because nobody has endless budget, nobody has endless time. There’s always pressure, right? Like they need the thing or whatever it may be.

Paul: Let me close with a prediction and then give pushback on the prediction.

Rich: Okay.

Paul: Five years from now, when we come and look at the software industry, it will look a lot like the software industry of 2019, but certain things will go much, much faster. Same roles, same systems.

Rich: Okay.

Paul: People are still going to be shipping JavaScript code that runs in web browsers and apps for phones, but there will be these new components and toolkits and certain things will be between 5x to 500x faster.

Rich: Yes.

Paul: I think that’s, I think that’s it. I think we’re—

Rich: Certain phases, you’re saying?

Paul: Yeah, because otherwise, if the magic trick was still there and the super robot builders were going to be able to build any software in a minute, then it wouldn’t make sense to buy a company for billions of dollars.

Rich: I think you’re right. I think you’re right. Now, does that mean, does that mean that certain jobs, certain people with certain roles should think about how they can be useful in this new context? Yes, it does. Because, look, man, a lot of human labor used to go into all sorts of things that got automated away, right? Like people yell at robots in factories. Like, that’s life. That is, that is the march of technology.

I do think what holds it together is that human, it’s the glue, really. Across these things that used to take six weeks now takes six minutes. Great. But someone has to interrogate it in that broader context where you understand the motivations and needs of what’s around you.

Look, there is a part of me that’s like, “Listen, here’s what I care about, Claude. I care about light. I care about an island in my kitchen. This is what I—I care about my children.” Like, I can tell Claude what I care about. That’s what’s a little scary about all of it. But boy—

Paul: That’s a… Boy, what a sequence that was. [laughter] That was, that was great.

Rich: No, but, like, Claude will do a pretty credible job. But you know what? I can sniff out when Claude is not doing anything extra and it’s just spitting back what I threw at it, right? Like, it’s always—that’s the tell every time, right?

Paul: Rich?

Rich: It’s not, there is no surprise. Yeah?

Paul: A database will never love you.

Rich: [laughing] There’s your title. Podcast title. Bingo.

Paul: Here’s how I want to close, which is we’re talking all these, like, nice, huggy, feely, family, kitchen-island things. When we say relationships, I’m talking about some of the most conflicted, difficult, nightmare business people.

Rich: Oh, yeah.

Paul: Like, just—

Rich: Totally.

Paul: Environments that would make you want to just rip your own face off. [laughter] A pure, sheer software development pain.

Rich: Yeah, yeah, yeah.

Paul: The fantasy that you replace that with the robots and the pain goes away. I’d be very, very careful. So, yeah, learn to use the tools, obviously. But it’s interesting to me. Your product manager interview doesn’t change.

Rich: No, no, no, no. Look, I need you to be curious and willing to use tools that make you more productive so you’re not wasting a lot of time. But the instincts you look for are the same instincts for sure.

Paul: Right back where we started from. All right, well, look, if anybody wants to get in touch, they should. There’s a lot going on, and, boy, are we—we’ve got a demo in hand that is hallucinatory is what, how I would describe it right now.

Rich: [laughing] Well, that’s often considered a bug in the AI world. You’re hallucinating, or is it hallucinating, Paul?

Paul: No, it’s not hallucinating. It’s just, like, I can’t, it’s doing so much that it’s hard to tell, like, we’re just going to land somewhere very, very interesting here. I’m very excited.

Rich: Same, yeah.

Paul: And so this is, you know, you guys are all in the room with us. Hello@aboard.com if you need anything. We’re excited. We’re growing. And be in touch. We’re looking especially for partners. We’re looking for people with big engineering teams and big sort of client bases who would like a little free, a very free or inexpensive taste of how AI could work in their organization. So get in touch.

Rich: Yep. Have a wonderful week. Take care of yourselves.

Paul: Goodbye.

Rich: Bye.

[outro music]