Paul Ford: Hi, I’m Paul Ford.

Rich Ziade: And I’m Rich Ziade.

Paul: And this is the Aboard Podcast. It’s a podcast about how AI is changing the world of software. Rich, how are you doing?

Rich: I’m doing good.

Paul: I want to talk about something that nobody talks about.

Rich: [scoffs] Is it AI-related?

Paul: No, it’s this thing I have on my leg. No, it’s… [laughing] It is AI related. It is the fact that AI can’t finish anything.

Rich: No.

[intro music]

Paul: Okay. So the thing I want to talk about here is what we’ve been calling “the last mile.”

Rich: Yes.

Paul: Okay? The last mile. Just, you know, The Last Mile was a movie in the ’30s. The last mile has a whole rich history. But here’s what I mean. I mean that, and there’s an old, this is a classic rule of software. You get the first 80% done, and now you have to do the next 80%. We’ve talked about this before.

Rich: Yes. Yes.

Paul: So we are currently getting our product into the market.

Rich: Yeah.

Paul: And, like, every software project, including those that are AI-enhanced, the last mile is very long. Just acknowledge that, like, this is not a secret. Every, our head of engineering, everybody knows this. Like, that, those last bits show up and you go, “Oof. Okay. That’s going to be a lot of work to get that across the line.”

Rich: Yes.

Paul: Okay? AI is a funny one, because the promise is you’re going to be able to describe an app and it’s going to ship an app.

Rich: Yes.

Paul: You got Lovable. Bolt. You’ve got Google Gemini. You’ve got all these frameworks saying, “We’re going to get you done.”

Rich: Yes.

Paul: Let me pause for a second there. And here’s what I’m finding. I’m finding that you don’t actually get things done. I’m finding in my own coding, I’m finding I sort of have to, like, pull back, stop using the same hammer, meaning a prompt, and figure out what I really need to do. I might still use a little more AI, but I probably have to go either write some code or organize stuff myself. I want you, as someone who has shipped over 2 billion products—

Rich: Mmm hmm.

Paul: —to tell me, when is it done? When is something done? And then let’s start to explore why AI and “done” aren’t necessarily lined up.

Rich: So before we talk about AI, I think it’s worth talking about how hard the last mile is for humans. It’s very hard for humans. Like, forget AI. And the reason for that is that as you get further and further into an effort—it could be, by the way, building a tree house. You tend to lose sight of the goal line and your judgment of what matters gets worse and worse over time. I don’t know why that is. Probably psychologists will talk that through. But what happens is you’ve decided from the very beginning that your treehouse is going to be beautiful and it’s going to be stained and you want carpeting in it.

Paul: Mmm. That’s gonna be nice up there.

Rich: Now it turns out that there’s an issue with rain and carpet, right?

Paul: Yeah, there is.

Rich: And those issues arise as you’re going through it. But you stay so committed to this, like, idealized vision of what the thing should be.

Paul: I mean, you could feel what it would be like to just be, you know, drinking a bottle of bourbon up in your treehouse.

Rich: And things just happen to you. Just happen that surprise you. Look, directors are notorious for, like, blowing out budgets and blowing out timelines, right? And that’s because, especially, like, really famous directors, they don’t, they’re not held to any—they’re like, “Go create your creative work. Go do your thing.”

Paul: I mean the legendary one there was Coppola. Who, like Francis Ford Coppola, like, Apocalypse Now.

Rich: Yeah.

Paul: Huge overruns.

Rich: Yes.

Paul: State of absolute anxiety and depression the whole time, like, just absolute misery.

Rich: The first thing you have to do is if I’m going to throw a treehouse reveal party and invite all my neighbors.

Paul: Mmm hmm.

Rich: And I’ve sent the invites out and they’re bringing cupcakes, right? There’s this intense, incredibly powerful tension that kicks in.

Paul: It’s a treehouse, too. Like you’ve got a lot of risk here. You could have your 65-year-old neighbor falling out there.

Rich: Yeah, exactly.

Paul: 85, I’m not putting her up on the ladder.

Rich: So what happens is this. What happens is you made a commitment to people. Now there’s a deadline. Now there is—when you say “last mile,” by the way, what’s implied in last mile is that there is a finish line.

Paul: You want to know the most unpopular thing I’ve ever said in my life?

Rich: Yes.

Paul: There was a point [laughing], there was a point about, many years ago where everybody was looking for things that weren’t computers but still could let you type.

Rich: Yeah.

Paul: So there were all these devices that were real lo-fi, almost, like, typewriter plus, so that you couldn’t get distracted and you would get your writing done.

Rich: Oh, I remember this.

Paul: And I had a couple of them and they were great. I enjoyed them. And somebody once was like, “Yeah, so you’ve written about that. What’s the one that you would really—what’s the one technology you would recommend in order to really get something written? You write a lot.”

Rich: Yeah.

Paul: And I was like, “Deadlines.”

Rich: Yeah.

Paul: Nobody liked that answer. They want to go buy a thing. And there’s actually, we talk about it, you know, I joke about the—

Rich: Deadlines.

Paul: I’m into synths. But there’s this whole concept of Gear Acquisition Syndrome, or GAS, because I’d rather buy something than learn something or do something.

Rich: 100%.

Paul: And everybody would. This is not—we make people feel guilty about this. This is just humans.

Rich: It’s just humans. And look, the truth is, the feature articles can come out because the printing press is revving up. Like, it’s like it’s an actual physical situation.

Paul: I used to work at a magazine and if you were, like, an hour late getting something onto the FTP server, it was like $10,000 an hour.

Rich: Yeah.

Paul: Literally, the presses were ready.

Rich: Yes, yes. Now, if you have a larger effort and there’s different participants and there are team members and whatnot? There are different levels of sympathy with hitting a deadline.

Paul: Mmm hmm.

Rich: There just are. Because some people are so enamored with their own craft and creating something special and exceptional that they find the deadline actually kind of a nuisance. Like, it’s almost a thing, like, “I’ll get you the thing, just leave me alone, I’m going to do it again.”

Paul: Yeah.

Rich: And they won’t tell you, by the way. Sometimes they throw out what they had three days before and they do it again. Right? And that could be literally an illustration for an article, or it could be a pile of code. But what good managers, good leaders, good product managers have to do is essentially make very ugly calls late in the game. They’re essentially, they start to say things like, “Don’t worry about that.” Like, “What? Don’t worry about it? I spent five weeks on that part.”

Paul: One of the most brutal aspects of software development, and this is sort of, is when you truly start to cut scope at the end. And it is, like, limbs coming off of a marathon.

Rich: It’s limbs coming off, right?

Paul: You’re going to have to run that marathon with one leg.

Rich: Yeah.

Paul: You’re like, “No, I need two legs.”

Rich: That’s right. That’s right.

Paul: And they’re like, “No, no, you get one.”

Rich: So now let’s bring AI into the picture. And AI right now is very, is like, is sort of this ridiculous concierge. It’ll just do, like, “What do you want? Do you want a recipe? Do you want a picture? Do you want a poem? Do you want whatever?” Right? And the truth is, when those tasks are, when they’re tasks, when they’re discreet and kind of finite, and you can kind of see the edges of them very quickly? AI will give it a go and be, like, “Here you go: Recipe.”

Paul: Well, I’ll give you an example. Doing geographic queries in a database is difficult.

Rich: Sure.

Paul: Like, you need to actually understand lat long, and you have to do great-circle calculations and there’s libraries and so on. But, like, it’s a lot of prior art because it’s literally all of cartography is wrapped up in there.

Rich: Of course, yeah.

Paul: You’re trying to measure the distance and be like, “Hey, how far is this going to be for me to deliver it, or do a search in the region?”

Rich: Yes.

Paul: Right? It’s really good at writing those queries for you.

Rich: Sure.

Paul: It’s really good. And that is, that’s, like, an ugly half-day task that goes to Task Heaven. You just don’t have to do that anymore.

Rich: It’s a task.

Paul: That’s right. It is a task. It is not conceptually a thing, it is just a task.

Rich: Do you know what AI doesn’t do? It doesn’t say, “Just curious, what do you need that for?”

Paul: No.

Rich: It’ll just do it.

Paul: Well, they can, they can teach it to do that, but it doesn’t, it’s a database query itself. It doesn’t care.

Rich: Exactly. So back to the last mile. What really—what makes the last mile so painful is that there is someone in the mix somewhere that is more worried about crossing the line than all the little tasks. In fact, the little tasks become very subordinate to crossing that line. And the reason they’re worried—and look, crossing the line? Shitty. Still sucks, by the way. You could be a project manager. Like, “I just got to get this out,” and then the product is shit. Right? It’s not good. And that’s a problem.

Paul: Let me ask, would you say I’m an empathetic person?

Rich: Yes. A little too much.

Paul: To a fault.

Rich: To a fault, yeah.

Paul: But as deadlines approach?

Rich: Yeah.

Paul: Empathy dies, because there’s no way to be empathetic anymore. It’s like, “Oof, that is a tough one.”

Rich: You are the steward of the thing. So there are two things you’re trying to achieve. One is whatever the goal is, just because—if you put a rickety box across the finish line, and it’s not good?

Paul: Mmm hmm.

Rich: You didn’t succeed just because you got it done, right? And also to get it done on time, there’s usually a commitment, a deadline, whatever it may be, because there are other dependencies or whatever.

Paul: Don’t ever take it—but when your boss or manager gets real chilly at the end?

Rich: Yeah.

Paul: That’s not personal.

Rich: Yeah, it’s not.

Paul: That’s just, any empathy or, like, emotional connection to you, as the deadline gets closer? It just increases project risk.

Rich: I had, like, a two-year window where I worked for someone, the CEO of a media company.

Paul: It was bad for everybody.

Rich: It was bad for everybody. I used to do this thing. He’d ask for stuff. Because he didn’t care. He’s just used to, like, he’s used to the caterers doing what he wants.

Paul: “Get me a car.”

Rich: He’s used to redoing his kitchen. He’s a billionaire, this guy.

Paul: Yeah.

Rich: Right? So he would come to me, like, “That’s nice, but why doesn’t it also do that?” And we were, like, five weeks away from, like, launching an app—a pretty big one, too. And it’s like, “That’s a great idea. We’re going to put that in the queue, and that’s going to be the next version. We’re going to definitely do it.” He’d never, he would never mention it again. He would never bring it up again. Why? Because my job was to get this thing across the line at that very moment.

Paul: You want to see what, like, the ultimate force for never shipping a product is a rich person doing a home reno project.

Rich: Oh, it could go for years.

Paul: They’re just like, “Well, you know, I don’t know if I like this blue.”

Rich: I’ll tell you a funny story. Another very wealthy individual wanted a particular architect to build their house, and they wanted to use a particular wood for, for the house.

Paul: Yeah.

Rich: And he, they sent them off to the, like, a special forest.

Paul: [weary sigh] Yeah.

Rich: And the architect came back and he said, “Another two years. They’re growing still.” Like, this happens, right?

Paul: Mmm hmm.

Rich: Zero deadline. They’re like, “I don’t care. I want the best. The best of the best.”

Paul: “I’ll wait two years. I got, I got a, I have a Ski-Doo.”

Rich: Here’s what AI doesn’t do. And I know that they keep talking about how it’s, “Oh, it’s reasoning longer.”

Paul: Mmm hmm.

Rich: Its reasoning window. There’s terminologies that seem to kind of come up, and I think a lot of it is sort of a campaign to show that AI is thoughtful in thinking deeply and not just knee-jerk, right? There’s a lot of, like, signaling like that. But here’s what it does not do. What it does not do is come out of the context of the task it’s working on and think about the overarching goal, like, that is sort of hovering over it.

Paul: It can’t. And I’m sure they’ll find ways to kind of simulate that in the half-hour thinking process and so on. But even then, there’s always a slightly bigger picture that you really like—ideally you’re, you’re always thinking a little bit bigger. There is no ceiling to, like, where things, where you shouldn’t, where you should stop thinking.

Rich: Yes. I agree.

Paul: But you know, you know what’s interesting. I want to come back to something. I asked you, how do you quantify done? And you don’t have an answer, because I don’t think you believe anything is ever done.

Rich: I don’t. But deadlines are incredibly critical.

Paul: Certain thresholds get crossed and then a business can move forward.

Rich: Yes.

Paul: But that is not done.

Rich: But you may have to create it. It may be artificial. There’s a term, “artificial deadline,” which is just because we would meander forever. And so I’m the boss. I’m going to give you a date.

Paul: So that’s the anti-AI technology is the artificial deadline.

Rich: Yes. It is.

Paul: Artificial intelligence on one side.

Rich: Here’s, I think, what we’re coming up against. What we’re coming up against is something that I don’t know how to articulate around, like, human instinct around real time decisions that you make to get something finished. Right? I have that instinct.

Paul: Mmm hmm.

Rich: I can tell when things are starting to get messy. You’ve seen me do it.

Paul: Mmm hmm.

Rich: It’ll be, like, two weeks prior to things getting messy. Like, this is about to get messy.

Paul: I think of it as, like, a shrieking siren.

Rich: Fine.

Paul: I mean it is.

Rich: It is.

Paul: It’s, like, the Rich siren goes off.

Rich: I mean, and that instinct, which I don’t write down, like, it’s not a, it’s not a, it’s not an algorithm. I don’t think, I don’t think we’re there yet.

Paul: I can describe—

Rich: Maybe we’re now wading into AGI, and that’s what’s going to do that.

Paul: No, I can fix all this for you.

Rich: Fix it.

Paul: First of all, the way your brain works is you have five or six components of a platform in your head, and you rotate between them about every 30 seconds, wondering how they’re doing. And if they’re okay—

Rich: I do break it up into parts.

Paul: If they’re okay, you never speak about them. And if they’re not okay, you speak about nothing else.

Rich: This is true.

Paul: It’s true.

Rich: By the way, we’re hiring at Aboard. [laughter] It’s a very healthy culture.

Paul: No, it is, it is a healthy culture.

Rich: No, I know.

Paul: But—

Rich: You’re right. What you’re describing is correct.

Paul: This is your job and we put a little buffer around you because I don’t know if that’s, like, if a junior employee was to experience the full force of that?

Rich: It’s very disorienting.

Paul: It’s a lot.

Rich: Yeah.

Paul: But it’s for senior employees to get that feedback at that level, and that’s what people want, because then they grow. It’s all good.

Rich: Yes. Yeah.

Paul: What I keep coming back to and what we’re talking about more and more inside of the company because we’re shipping products, we’re using AI, our tool uses AI. By the time you’re hearing this, it is out and we’re in the real world with this stuff. The processes that we’ve used for software that we’ve evolved over 50 years, things like Agile or even Waterfall—

Rich: Yeah.

Paul: Or just, like, extreme programming, pair programming, all the stuff that, that we do is just not quite up to the task of this new robo buddy that spews code everywhere.

Rich: It’s a lot of pieces.

Paul: And so, you know, what we talked about recently was how developer experience is becoming really important at places like Google around AI, or Anthropic is going to help you with GitHub, which is wild when you think about it. Like, hey, I thought we were going to revolutionize every aspect of knowledge management and coding. No, actually we’re going to write your pull request with this AI. We’re going to bring it back to the old way of doing things.

Rich: Yeah.

Paul: But we’re going to accelerate that because the new way is too disruptive and nothing actually gets done.

Rich: Yeah.

Paul: So I think what we’re going to see here is that this becomes a component of the whole old way of doing things, and that a lot of our processes have to evolve. I don’t love Agile, but it worked to get a lot of software written.

Rich: It did. It got people in the room. That was the whole point.

Paul: You have a stand-up, you talk, and then you got use cases and story points and then everybody talks about the fricking process instead of working, which I hate.

Rich: Yeah. Yeah. Yeah.

Paul: But regardless if it’s applied and people say, “Okay, I’m now going to go to my desk and do some work,” and they need do it? It works pretty well.

Rich: Yeah.

Paul: It’s like going to the gym. You can talk about going to the gym and never go to the gym.

Rich: Yeah, articulating—

Paul: I’m doing that right now.

Rich: [laughing] Yeah, yeah.

Paul: So that, to me, when we talk about the last mile, it’s never going away. Don’t assume that AI is going to close up this gap. You’re going to be able to get a lot of stuff shipped, but now it has to go and actually find users and be tested in the market. The last mile is very long, it’s going to remain very long. And you need a process that allows you to assess things, keep humans in the loop, make sure you’re going to be able to ship. And that process, I hate to say it, it’s probably going to look a lot like the old ones. It’s going to be Agile with AI, it’s going to be with bots, it’s going to be with LLMs.

Rich: And look, we’ve gone into a lot of different cultures and shipped a lot of software. People tend to underestimate, when you impose tools on culture, you often have to navigate a lot of different changes. A lot of change. There’s just a lot. It’s just a lot around—what you thought you were doing in the beginning is actually quite different by the end. And that is just humans and culture and how people work.

Paul: One of the things I’ve seen many times is Salesforce jammed into do-gooder not-for-profits as a case-management tool.

Rich: Sure.

Paul: So it’s like we’re going to help these, you know, we’re going to help the kids by making a record in Salesforce for each one of the kids.

Rich: Yeah.

Paul: And years later, the people are grieving that they have to use this.

Rich: They have to use it. Oh, they don’t. And they pay the bill.

Paul: They miss their old system so badly.

Rich: Yeah.

Paul: Because it was—the old system was built with crappy old tools, but custom for their needs.

Rich: Let me, can we go devil’s advocate for a second?

Paul: Let’s go and let’s, let’s do that and let’s end this.

Rich: In a previous podcast, we talked about just constant innovation and invention and Jony Ive has made a cough drop you put in your mouth that understands you and it has a little microchip in it.

Paul: It’s made of aluminium.

Rich: Yeah. You love saying that.

Paul: I just— [laughing] The way he says aluminium, man.

Rich: Can we get there? Why can’t we get there? It’s a mile. It’s not 100 miles. Why not, like, a master agent that seems to think about intent and goals, and then there’s little baby agents that do all the tasks and, and they report up to the master agent.

Paul: Richard. Of course we can get there. But that won’t be the finish line when we get there.

Rich: What do you mean?

Paul: Everything will keep moving further and further out. You’ll get the cough drop that monitors your body and tells you things and answers your email for you.

Rich: Yeah.

Paul: And then that won’t be enough.

Rich: It won’t be enough.

Paul: It won’t be enough. Humans are never satisfied.

Rich: Yeah.

Paul: And we do the dumbest stuff when we’re unsatisfied, like, elect certain people.

Rich: Yeah.

Paul: We get real anxious and we can’t stop.

Rich: Yeah.

Paul: So the assumption that somehow we’re going to put a dome over all of this and everybody’s going to line up and be, like, “It’s AI-enhanced now and I’m happy.”

Rich: Yeah.

Paul: No.

Rich: Yeah.

Paul: You will be amazed at the perversity that is about to unfold over the next 20 years.

Rich: Yeah. I think you’re right. And I think if history is any indicator, when we do invent things that seem to open up free time and convenience, we fill it up with more bullshit.

Paul: Otherwise we’d have a 30-minute work week and robots would be, like, massaging our feet.

Rich: I gotta tell you, try to do anything, like get new glasses, and it does feel like some people have a 30-minute work week. I’m not gonna lie.

Paul: Okay, well, that’s a whole different old-man conversation there.

Rich: Join us on The Old Man Podcast.

Paul: That’s right. [laughing] Rich, we haven’t talked about our product.

Rich: We haven’t.

Paul: Okay, go to aboard.com. There is a text box right there, front and center. And you put the kind of software that you want into that text box and we’ll do a good first pass. It’ll be, I mean, now look, if you want—

Rich: It’s gonna have a conversation with you.

Paul: Yeah, if you’re like, “Photoshop for llamas?” No. But if it is like, “I need to manage my llama farm?”

Rich: Knock yourself out.

Paul: We will dynamically, in the browser right there, build you real software for managing your llama farm. You can’t just use it just like that. You got to talk to us, all sorts of stuff like that. You know, we’re still humans and humans stay involved so that the software we ship is actually really good and works as opposed to—

Rich: It’s a sampler of the capabilities of the platform.

Paul: Like a delicious Whitman Sampler.

Rich: I don’t know what that is.

Paul: Oh, it’s those little, it comes—

Rich: Oh, it’s that garbage chocolate from the Midwest.

Paul: You have just made 50 grandmothers cry.

Rich: Whatever. It’s, like, See’s, S-E-E-S.

Paul: That’s very different. It’s a Whitman Sampler.

Rich: Okay, fine. I know what you’re talking about. I know the box.

Paul: When I was a kid, man…

Rich: It’s got that stitched diagram.

Paul: You know, we’ve talked about Dipsy Doodles in the past, but when I was a kid, you’d see that Whitman Sampler—

Rich: You can’t sell our product for two minutes.

Paul: No, I have—

Rich: You’ve moved on to Whitman’s chocolate.

Paul: I have a slight attention issue.

Rich: Check it out at aboard.com. Our pitch is getting sharpened as we speak. [laughing]

Paul: Yeah. All right, let’s go get—

Rich: And reach out. Hello@aboard.com. We love talking about stuff.

Paul: Let’s go get a Whitman Sampler.

Rich: Let’s go.

Paul: Bye.

[outro music]