Subscribe
Share
Share
Embed
Explore your curiosity - Interesting people with fascinating stories.
Life consists of three things:
-The stories we tell others
-The stories others tell us
-The stories we tell ourselves.
Paul, welcome back to the show. I don't think I've said that more than once before. You're the second guest to get a return invite. So welcome back to the show.
Speaker 2:Great to be here. Thank you.
Speaker 1:You were telling me you read a book about tool building. Tell me about it.
Speaker 2:We wrote we wrote a blog post about this recently. Yeah. So I was actually so this is a bit of a long story, but I'm building a paint booth because I'm restoring a car at the moment. So you ever play, like, a video game where, like, there's a main quest and then you go on a side quest and then you go on another side quest. Right?
Speaker 2:So so I'm I'm restoring this nineteen sixties MG, and I'm doing all the work myself. And as part of that, I'm gonna be painting it. So I decided to build a paint booth.
Speaker 1:Of course.
Speaker 2:You did. Of course. Right? That's what you do. And so
Speaker 1:As you do.
Speaker 2:And the paint booth, the internal walls are made of steel. So I had to cut some steel. I had these steel sheets laying around. And I was working with my father-in-law, and we wanted to cut a straight line. So naturally, because I've done a lot of woodworking, I grab a piece of plywood that I had laying around and convert it into a quick little saw guide for sliding the saw, and he hadn't seen that before.
Speaker 2:But because I've done a lot of woodworking, it's just a very natural thing to if you have a circular saw and you need to cut a straight line, you know, you can look on YouTube. There's ways to create circular saw guides for it, and we use that all day to to cut up the steel. And so it inspired writing this this blog post where because what what what I realized talking to some folks internally is that, like, woodworking has this special property to it, which is a lot of the tools that you use for woodworking are made of wood. Right? So you'll one of the first projects you might do is create a workbench or something out of out of wood.
Speaker 2:And you see the same with blacksmiths. So when you're learning to be a blacksmith, one of the first things you do is create a hammer, create a chisel, create tongs. Right? The sort of special property that these crafts can create their own tools. And it occurred that programming is kinda unique in in this way.
Speaker 2:Right? Like, engineers, other types of engineers, like bridge building engineers, don't don't kind of really do this to the degree that programmers do. We're kind of special amongst a lot of roles in that most of the tools we use on a daily basis are built by people like us. And and there's some unique things to that. And so we're doing a lot of work in the platform engineering space at the moment.
Speaker 2:And platform engineering, is is kind of the ultimate version of this. Right? Like, it's a team based version of this, but it's, you know
Speaker 1:So wait. For those of our guests who are kind of non engineering executives, what does what does that mean?
Speaker 2:Yeah. So, you know, if you're a CTO, you have this question of, I have a 100 engineers, you know, are they being productive? Are they are they being effective? And one of the ways that you might find that they're not being all that productive or or effective is because they're all sort of solving the same problems over and over again. Right?
Speaker 2:So this team over here has a way of deploying stuff to production, and this team over here has a different way to do it, and they're both spending time maintaining and building these sort of internal tools and and so on. Now we don't we don't talk about developer productivity in the same way we might talk about other jobs and and the productivity associated because measuring you can't measure engineering productivity anyway, really. It's more creative and lines of code isn't a great isn't No. So
Speaker 1:I mean, that's a great topic for us to dive into, but I'll let you keep going. We might come back to that one.
Speaker 2:Yeah. So I think the the sort of innovation more recently has been the focus on developer experience as an alternative way of talking about this. So I guess the hypothesis would be if we can set up our engineers so that their day to day toil of doing the engineering work is is less and they're just able to spend time, you know, writing code and and building features that benefit our business, we can kind of imagine that that will likely produce a more effective engineering team without a focus on productivity on itself. In fact, let's just focus on removing blockers and annoyances. So that's that's where developer experience comes in.
Speaker 2:And then part of the pursuit of developer experience becomes, well, like my example, I'm using my circular saw and it's gonna wobble all over the place. But if I take five minutes to build a quick tool that will help me create straight lines, well, we do the same in in software teams. And so the developer experience initiative becomes a platform engineering initiative, and it then the focus becomes building this platform. But a little bit like the side quest problem, at some point, the side quest becomes the quest. Right?
Speaker 2:And so the platform team
Speaker 1:In fairness, AWS was a side quest that became a billion dollar business. Right?
Speaker 2:That's the ultimate example.
Speaker 1:Yeah. Yeah. Yeah. Ultimate side quest.
Speaker 2:It is. Yeah. I mean, you go back to the the Jeff Bezos, the leaked email, right, of, like, everything is gonna be an API and teams are gonna be decoupled from each other, which, yeah, results in all of the the internal tooling kind of being easily something they can they can make available for other people to use. Yeah. And so, you know, that happens across enterprise companies all the time.
Speaker 2:Right? You you end up with these platform teams. But I think, you know, the tool building thing, I've definitely done woodworking projects where in order to improve the productivity of the actual project I'm working on, I'm gonna go and spend five minutes creating a tool. That becomes 20 creating the tool. That becomes the entire day creating the tool to the point where I've lost the motivation for the main project I was working on because I'm so frustrated about how the tool building process has gone.
Speaker 2:Right? And we see this, I think, with software platforms too, where the, that the platform can sometimes become this all consuming thing that ultimately doesn't get to the main benefit we were hoping for, which is a more engaged, productive, less blocked developer team.
Speaker 1:Yeah. Let me tell you a story that will probably get me in trouble should I still So be working for we wrote algorithms for Apple, Qualcomm, Honeywell, which basically did design for manufacturing, which is basically analysis of the electrical design and the physical design, meaning the layout. So I basically, you know, I was the test, I was the gatekeeper for the next version of Snapdragon. Right? That was the algorithms that I wrote.
Speaker 1:And it would take us three to four, sometimes, you know, two months, three to four weeks, sometimes two months to write these algorithms. And you know, I must've written 50 to 100 of these algorithms. And it hit me, this is just a mathematical abstraction that I'm writing again and again and again every time. So it hit me that there's a mathematical abstraction that could be written, a tool, you know, a layer, an API. And it hit me that that mathematical abstraction would take the time down.
Speaker 1:And I wasn't sure to how much. So I thought through it and I was like, in 80% of the times, what took up to two months would take thirty minutes. Wow. I went to engineering, they told me, Ari, that can't be done. I went to product management, Ari, that can't be done.
Speaker 1:I went to my boss, Ari, that can't be done. I said, okay, can I work on this? Like skunk works. I'll spend time on the weekends. I won't spend more than a couple hours a week on this.
Speaker 1:Sure, go do it. It could be done. So then I come to my boss and I'm like, give me a check. And he's like, do this. Five lines of code.
Speaker 1:Right? I just wrote the math. Right? I wrote the equations up and it did what it was supposed to do because everything at the end of the day is just geometric math when it comes to this stuff, right? I write it up.
Speaker 1:He's like, Ari, you can't tell anybody you did this.
Speaker 2:Why? Well,
Speaker 1:we were a consulting business. So now what took me months now takes me an hour. So what am I going to charge? Half an hour?
Speaker 2:What's the billable hours on that?
Speaker 1:What's the billable hours on that? So I was not allowed to tell anybody that I did this. In fact, he's probably going to curse at me if anybody hears this.
Speaker 2:Yeah. So that's that's the lesson. Right? Yeah. If you're being paid by the hour, maybe tool building isn't the best thing to
Speaker 1:It's not the best thing to do.
Speaker 2:Isn't that amazing that, like, you intuitively knew that a tool could be built to solve this problem though?
Speaker 1:Look, look, even a monkey, if they did the same task a 100 times, they would see patterns. So I don't think I'm much better than a monkey. Monkeys, by the way, you know, they lick a piece of stick and they stick it in to get some ants. So even monkeys tool build and I have literally written a 100 of these checks before these mathematical patterns emerge to me. I can't say that I looked at it the first one and I was like yeah no problem I'll do this.
Speaker 1:So, so I wouldn't, I wouldn't claim any kind of genius on my part, but, but yeah, I think that if you, if you become an expert, put in your ten thousand hours, right? Then things that, the shortcuts that are apparent to, that are not apparent to everybody suddenly become clear, crystal clear to So there's something about that, right? It's the ten thousand hours is what got me there.
Speaker 2:There is, but you know, if you're, let's say that you've spent ten thousand hours playing the piano, that doesn't mean you can build a piano. Right? The the skills required in learning to play a piano are not the skills required in building a piano. There's something unique, I think, about software and the crafts, but only some crafts too. Mean, sewing, for example, doesn't help you build too many tools, I think, to assist with sewing in the way that, say, woodworking Yeah.
Speaker 2:Or metalworking
Speaker 1:I mean, what's what's you know, this is such a timely conversation because the world is changing and now not only are we building the stupid stuff that I did, right, mathematical abstractions, I mean probably any mathematician would have looked at what I did and be like, oh, all right, that's simple. What you did isn't that impressive. The problem was I was the only aspiring mathematician in the group. That's why I saw it. But now what's happening is on a whole different level, right?
Speaker 1:Math and a little bit of neuroscience gave us neural networks. Neural networks in the combination with something called GAN, which is basically game theory of two agents fighting each other to get to efficiency, gave us generative. Then some, I don't know who it was, but somebody came and said, well,
Speaker 2:what if we gave it
Speaker 1:a hell of a lot of data? That gave us GPTs. And then I don't know what the next step was. Maybe you know this, I don't know. Somehow GPTs learned how to code.
Speaker 1:I guess they just trained them on code instead of language, but I'm not sure about what the answer to that is. Suddenly, GPTs can code. They are now the ultimate coding tool.
Speaker 2:Yeah. Yeah. Yeah. That's an interesting way to think about it, isn't it? So does vibe coding and and really just I mean, we've had this for a little while.
Speaker 2:Right? If you think about Excel, you know, a lot of people in a lot of business runs today on Excel because people have spotted and and these are people who haven't had, like, a programming background. Right? They're they're not they're not they wouldn't be fully able to build things, but Excel kinda gives them a bit of a starting point at least to build for certain types of workflows. And when we had this with Access, you know, a little bit as well, right, how many doctors build their own Access databases and and things.
Speaker 2:But, yeah, we're we're probably at a new place where I
Speaker 1:see You're dating yourself, Paul, with Access.
Speaker 2:That's right. Well, I like I I there's something interesting in the history of software that I think is worth studying even as you went around. But, but, yeah, certainly, I think the AI tooling is makes things accessible to to people that otherwise wouldn't be. And you find this, I think, the the debate around artists who spend a long time practicing the skills of being an artist in the pursuit of creating art or graphic design versus someone like me who just has never I'm not the person that sits around sketching something all day, but I I can kind of prompt an AI to create something that doesn't look too bad. And that's quite empowering for someone who just hasn't had the ten thousand hours to put into learning that skill.
Speaker 1:Yeah. I mean, you know, this drives a very important, I think, question. Are engineers out of a job? Are artists out of a job? I have an answer.
Speaker 1:What's yours?
Speaker 2:I've got a very contrarian view on this whole AI space actually that that comes more thinking from
Speaker 1:If it's contrarian, we might actually agree.
Speaker 2:Yeah. I I had this conversation with my daughter once around, like, what so she she she's very artistic and she spent a lot of time. Some of the paintings are behind me. And and so I think she's seen a lot of the things happening in the AI space and the art community reaction to that. And we had this conversation around like, I asked the question of what do you think an artist does?
Speaker 2:And there's sort of two levels you can answer it. Right? So the there's a basic level of answering, which is they sort of wiggle a brush.
Speaker 1:Beautiful question. I mean, that question is brilliant. I just love that you did that with your daughter.
Speaker 2:Yeah. Because the basic answer is they wiggle a brush on a canvas, right, and colors come out, which is the same if you said, what does a doctor do? The answer would be, Well, they have their hands in blood and guts all day and it's kind of like plumbing but grosser. And that is a true answer to the question. Another true answer to the question of the doctor is, Well, they save lives.
Speaker 2:Right? And so the same job has two different answers to the question, one very high level, one very low level. The same is true, I think, in the art question. I think, you know, artists challenge our conception of humanity and what goodness looks like and and all these things.
Speaker 1:Thank you very
Speaker 2:much. So, you know, can AI wiggle paint on a canvas as well as humans probably can? Yeah. Probably imaginable they will. Can AI do that high level function in the same way?
Speaker 2:Probably not. And so but it might encourage it might it might allow a whole set of people who haven't spent ten thousand hours learning certain artistic skills to express themselves. I think, you know, someone a human creating very derivative art is probably having less artistic impact than someone who really is challenging how people conceive of humanity even though they might have prompted an agent to do it. So I don't have a clear answer on it, but I don't think it's an answer that can easily I think it's
Speaker 1:the perfect answer. I'll tell you why it's the perfect answer, because I see a lot of artists, copywriters. I was speaking to a very accomplished, acclaimed author the other day and she was like, Ari, I'm depressed. Like AI has taken my job and it's killed my industry. And like yourself, I said, no, it's quite far from it.
Speaker 1:And I actually think that we are in front of ten years of economic devastation, but then we're going to have a renaissance. Unfortunately, of those things are going to happen. Why? Anybody who is not at the top of their game and is not going to be an early adopter of AI, there is a certain percentage of a chance that they will be impacted because their job will no longer exist in its current form or not at all and will be replaced with something else. So I think there's going be cascading effects across industries.
Speaker 1:So I think it's going to be scary for a bit. I don't know if it's ten years or two years or five years or three years. That's hard to say because all the numbers kind of we don't know we've lost our ability to create time predictions in this age. So I don't know really if it's going to be two years or ten years. I hope obviously that this economic devastation will be time limited.
Speaker 1:But what I think is going to happen after that is going to be a renaissance. And in fact, I think we'll see the renaissance very early in the game, but in small scale. And then that renaissance will roll out to the general public and that renaissance is a derivative of what you're saying. It's that anybody can cook right? It's the Gustavo.
Speaker 1:Anybody can code. Anybody can illustrate And then it's a question of what are you? Do you wiggle technique? Right? Which is wiggling a brush is technique.
Speaker 1:So if you don't need the technique, what's left? Inspiration, love, putting up a mirror to society and saying, is this who we are? Self reflection? Enlightenment? That's what's left.
Speaker 1:So either we degrade into war and chaos and all kinds of really bad stuff and that probably will happen for a while. We're basically there right now.
Speaker 2:That's quite depressing. Yes. I think to your friend though, yeah, I think the question is I've seen this with writers specifically. I think sometimes writers attach their sense of self worth or their belief in their own skills to the specific, the low level job of writing. Right?
Speaker 2:Like, I'm really good at forming sentences that are that are compelling. Actually, thing that the most impactful writers do, I think, is they have an idea worth sharing or a story worth sharing that adds some value to humanity. The actual sentence structure is a part of that, but it's actually not the most important impactful thing they ever did anyway. They maybe just need to attach themselves to that higher sense of what is my purpose
Speaker 1:right They're not copywriters. That's they don't write copy. Right? They're not the thing that they do. Yeah.
Speaker 1:They never were that. Right?
Speaker 2:No. And and for some of those roles like marketing copywriters, like, certainly the the challenge with the way AI works, and I don't know if this will ever change too much, is it's only gonna produce the kind of things that the average well informed person will produce. It's never gonna produce something that unique. In the in the world of marketing, you need something unique. Right?
Speaker 2:The the more derivative your marketing message is, the less people are going to pay attention to it.
Speaker 1:But if you think about it, it's a statistical model that is looking at averages and like what a neural network is literally a statistical model with a lot of bells and whistles. Like we can, we'll lose all our listeners if I start talking about math, but that's all it is. It's a statistical predictive model. End of story. There is no magic in AI.
Speaker 1:In fact, it's not AI. It's neural networks. It's two generative neural networks, one classifier, one generative that basically decided that each other were kind of okay. That's all generative is. So if all we have is a statistical model, I don't know, but according to my understanding of mathematics, statistical models are not creative.
Speaker 1:They have the illusion of creativity, but they're not creative.
Speaker 2:I think I'm with you on the changes to the job market. I think we saw this we talked about spreadsheets before. Right? But, like, where spreadsheets in Excel come from, people don't kinda know. These used to be actual huge pieces of paper or whiteboards or black blackboards.
Speaker 2:And it was people's job. Like, you know, when you change a cell in Excel and the other cells recalculate? Like, entire firms, they've had people the full time job was like, you change that thing. I'm gonna go and rub this out, and I'm gonna calculate the new thing and put it here. Right?
Speaker 2:And that's how that's how businesses were run. Well, of course, the spreadsheet comes out and overnight, that job just doesn't exist anymore. But that job like, the people doing that job went on to find other much more impactful things that they could do anyway that increased productivity of the the country. You know, so that that will happen. Do you know, I think there's gonna be a devastating loss though on the side of investors.
Speaker 2:I think investors are gonna lose so much money in this AI craze.
Speaker 1:Here here's a well, because they invest in in in basically tin tin wrap tinfoil wrappers or because they're gonna lose their power? Which which what's your reasoning?
Speaker 2:I I think it's the I feel like do you remember a few years ago, we had this sort of boom of, like, company valuations. Right? Every company was was sort of worth 30 x revenue. Yeah. And now the other thing that happened though is anything that looked remotely like a software company got valued like a software company.
Speaker 1:Yeah.
Speaker 2:So if you say, like, you know, like, Microsoft is creating Windows, and every copy of Windows they sell doesn't cost them anything. They could sell a billion copies. They could sell three copies. Yeah. There's no cost of goods sold.
Speaker 2:That's the thing that always underwrote software company valuations because Yes. If my gross margin's 90%, I can do that thing all day and my cost structure doesn't go up. AI is totally the opposite of that. So so the the idea of, like, you
Speaker 1:my This is brilliant. Explain why.
Speaker 2:Well, the the cost so let let's say, right, so today you have an assistant, and they cost you a bunch of money to have this assistant do basic tasks for you. And you say, oh, AI can do that. There's an AI agent that will do that. So the investors get really excited. Oh, the whole PA industry is gonna disappear, and, you know, this company doing AI stuff is gonna make all this money.
Speaker 2:Well, the problem is gonna be you're gonna charge $30 a month or something for this assistant, maybe maybe more. Every question that gets asked has to be answered through running everything through that neural network you described. There's a huge hardware cost with that. And what you see today in a lot of AI companies is huge variance in those costs, but it's usually the gross margins. The gross margins are not 90%, they're not 80%.
Speaker 2:They're typically, at best, 40% and at worst We're negative one hundred percent one hundred
Speaker 1:that's ridiculous. Yeah. We're adding, we're significantly increasing variable cost. Like that's one, so that's huge, that's brilliant. I haven't heard anybody say that.
Speaker 1:Insightful. Here's the other problem. We're also eliminating barrier to entry in software businesses. Hypothetically, the next two to three years, if vibe coding gets good enough, the, you know, you know, barrier to entry in software, excluding distribution, excluding sales, excluding relationship, just software. So the only barrier of entry, which is in the IP itself is calculated by the months taken to develop that IP.
Speaker 1:Now with the introduction of vibe coding that actually works, what if that five years of software can be created in five minutes? What happens is your barrier to entry from an IP standpoint goes away as long as the intellectual property associated to the domain expertise is not mystical. So there's some caveats here. There's some salt that we need to kind of put over it.
Speaker 2:We should come back to that. Kind of agree, I actually disagree a little bit with that. But I think the fundamental challenge for these AI companies is it's a it's a very different economic model where they're assuming it's gonna have software level multiples and I just don't see it. Do you remember during the COVID period there was that company there's a large US company that did like Peloton, it was. You know?
Speaker 2:Oh, yeah. Oh, yeah.
Speaker 1:They went up and then went down.
Speaker 2:Everyone got a Peloton.
Speaker 1:Now Everybody went we did too.
Speaker 2:They looked like a software company. Right? Because they were on the internet. So they got software company multiples. At one point, they had 4,000,000,000 in revenue and they were worth like $50,000,000,000.
Speaker 2:Right? That's You know what their gross margin was at the time? 40%. The same as these AI companies at best. Most AI companies are losing money on.
Speaker 2:So imagine like that's your like any any anyone should have been able to look
Speaker 1:at that a media company. Let's just agree that they were a a media company. Was not a software company.
Speaker 2:With physical hardware that had to get shipped to you. Like Well,
Speaker 1:there was the Peloton hack. Right?
Speaker 2:What was that?
Speaker 1:So the Peloton hack is where you just get the app and you use any old bike.
Speaker 2:Alright. Yeah. Yeah. Yeah. But there's still I I don't know.
Speaker 2:The the cost of goods sold on that thing wasn't like a software product was. And so the more customers they add, the they're making very little profit on that. And so what are they worth today? I think today they're worth like 3,000,000,000 or something. They're worth and the the revenue's gone down, but and their gross margins have actually improved, but the people realize this is not a software company.
Speaker 2:Right? I mean so I think most of these AI companies are not software companies. They have really high cost of goods sold on the thing. That's why I think investors are gonna lose a lot of money. You know, on software though so I think there's there's something if you think about product the act of product creation.
Speaker 2:Right? When you the only reason you create a product in the first place is you think that there's a better way to solve the problem than all the current ways that currently exist to solve the problem. Right? You've got this sort of unique insight.
Speaker 1:10 x kind of in the VC approach. Yes.
Speaker 2:Yeah. If if we could just bring this new technology that's come about, that will solve the problem better or just a new way of thinking about and framing the problem, you know, will will solve the problem better. And and then over time, that gets you you that initial insight, you know, you embed it you embody the knowledge into the product. People use the product. They give you feedback on it.
Speaker 2:You're continually kind of innovating it. It takes this sort of evolutionary pathway where it gets more valuable.
Speaker 1:Domain IP, not your software IP. Yeah.
Speaker 2:Yeah. Exact exactly. There's like like if if Atlassian disappeared tomorrow, there's still a lot of really clever things that they figured out with Jira in the act of asking how do we build a better way of managing work that needs to be done Yeah. Where humanity could kind of recreate the solutions to that. But in doing it, you know, like, they have a very high moat.
Speaker 2:Like, if you wanna start a bug tracking product today, you're gonna spend five years and tons of engineers just building all the table stakes stuff that Jira already has before you even get to build something innovative. Or you have to have something that's so innovative, people aren't gonna ask for where's your SOC compliance, why does it integrate with ServiceNow, you know, all the enterprise table stakes that come with it. And I'm not I don't see that coming out of, like, the vibe coding coding movement, really. Yeah. And almost like the evidence would would sort of suggest it's not going to happen because this stuff's now been around for a few years and these companies haven't been built yet.
Speaker 1:So we probably will disagree. Here's my here's if you have one atom on one side of a river in Switzerland, I think it was, and you can literally teleport that atom to the other side of the river. That's all you can do now, right? That's physically possible today. Obviously you're destroying the atom and recreating the information on the other side.
Speaker 1:What's the difference in creating a molecule from one side of the river to the other? What's the difference in creating a molecule from here to the moon? What's the difference if you can do a molecule from here to the moon? What's the difference in creating a frog from here to the moon?
Speaker 2:What's the difference
Speaker 1:in creating a human? Now the famous saying was, I don't care about this anymore because it's no longer a physics problem. It's an engineering problem. So my argument would be that, sure, Vibe coding doesn't work today, but how far are we from solving that engineering problem?
Speaker 2:Well no. But I think it okay. Let you know, we talked about the there's a low level way of describing a thing in a high level way. So the low level way of describing software development is I type code and the code compiles. Right?
Speaker 2:Right. The high level understanding of what programmers do, what the good programmers do is they find creative ways to solve problems in a better way than they'd previously been solved. Right. And they're continually finding those those those ways of doing it. That's to me, that's that's what software development is.
Speaker 2:So I I think it's quite different to the the sort of atom going up to molecules problem because it's not scaling up. Like, Jira is not more code thrown at the prompt of like, if if you said, you know, today, create me a bug tracker and GPT spits out 400 lines of code. Creating Jira is not a case of that same GPT spitting out 500,000 lines of code. It's actually 500 more prompts that you need to give. It's 500 more ways of thinking about the thing.
Speaker 1:Today.
Speaker 2:Now, like, if you said, okay. So for me as a software developer, what was the hardest part about being a software developer? It was never writing the code. I learned I learned the code is easy. It's actually, it was the people.
Speaker 2:It's the user. So if you start from the premise of like, if if if there's an imaginary universe where you say, the user perfectly understands their problem and the best possible way it might be solved, and they can express that clearly and succinctly.
Speaker 1:That's never
Speaker 2:And then the hardest part of our software development is turning that into code. Like, was not the problem.
Speaker 1:That's never true.
Speaker 2:No. That's never agile movement was actually the user doesn't know what they want, and that's totally okay. And they don't know what possible. The job of the programmer is to bridge that, know, say.
Speaker 1:Here's the thing, right? Let's talk about economics and let's talk about supply and demand. If all else equals, sorry, and we're going into economic terms, I'm going to apologize to the audience, but we'll try and keep it as simple as possible. All else equals, meaning if we don't change any other parameter and a job is 10 times easier to be done, then you're significantly increasing the amount of people who can do it.
Speaker 2:Yes.
Speaker 1:Meaning you're increasing supply 10x. If you're increasing supply 10x with all else equals, meaning demand, the number of jobs that need to be done, what that means is that your price or salary or whatever for that or unemployment is going to increase significantly unemployment. The price is going to go significantly down because now my 13 year old son can do the job because he has the mental capacity to do these iterative logics in his head. However, he doesn't, he hasn't spent ten years practicing a programming language, but he doesn't need to do that anymore because all he has to do is logical iterations and then he can visually see what it creates. And so he's really not learning C, C plus plus which is what I learned as a child.
Speaker 1:Now he's learning prompt engineering and he's learning software architecture, which he has to feed into the prompt. So he has to say, oh, this is the design pattern I want to use. I mean, that in itself is fundamentally problematic. If we are changing supply and demand by a factor of 10, a 100? Yeah.
Speaker 1:That's catastrophic.
Speaker 2:Yeah. Maybe. You know, we had this with with programming. Right? So we went from low level languages to high level languages.
Speaker 2:Yes. We also saw an increase in the number of programmers, and there was also more demand for programmers too, as we could see what programs were capable of doing. So and and there were periods of boom and bust with that. Right? Like the sort of the old joke of like through the .com boom was, you know, some 13 year old coming in, running the company or something.
Speaker 1:Yes. Well, that happened with Facebook. Right? I'm 13, but pretty much.
Speaker 2:Yeah. Yeah. That see, there there will be that boom and bust. I think generally I don't know. Maybe it's gonna be a return to you read about, like, the old scientists from, like, the seventeen hundreds, and they weren't just kind of focused on one thing.
Speaker 2:Like, someone wasn't just a physicist. They were also a mathematician, and, you know, all the knowledge was kind of available.
Speaker 1:Yes.
Speaker 2:Because the the amount of knowledge was less than today. And then more recently about these sort of hyper specialized roles, I think a if you ever watch old time woodworkers, you know, using hand tools or something, you watch that stuff and you think, like, I just don't have the patience to to learn all of that. But you know what? I can go down to the hardware store and I can buy a circular saw and drill, and I can create a table that's, you know, not gonna be as beautiful as as pretty or anything, but it will be functionally good.
Speaker 1:Yes.
Speaker 2:So there's probably arguably more woodworkers today. There are more humans on the planet capable of crafting something of utility out of wood than there previously were.
Speaker 1:Let's break down your argument because it's brilliant. I want this to be so clear because this is, I think your argument A is true, I completely agree with it, and I left you a way to kind of punch my argument down because I said all else equals. All else is not equals. It never is. All it is is all else equals, you're right, is a tool for economists to do modeling.
Speaker 1:That's all it is. All else is never equals. So what I think you're right and I think what we're going to see is, first of all, there's going be a dip. There's going be a crash. I'll tell you why.
Speaker 1:Because everybody's going to take a bath, right? In economic terms, they're going to want to do an, basically an accounting bath. They're going to want to show that they've cut all these costs, that they're hyper productive, hyper efficient, hyper profitable. They're going to want their stock prices to jump up. Like okay, so when we stop being greedy we'll figure out that there's massive demand.
Speaker 1:So now productivity has increased 10, a hundredfold. Sure, supply went down, demand was held equal, but then demand is going to go out of control. Now here's a few reasons why. If I'm an IT company, I have no developers. I have an, sorry, I'm an any company, I'm a B2C company, any company, gas, oil, chemicals, pharma.
Speaker 1:I don't have a engineering team, a software engineering team. If my IT person, my accountant, my head of design can now vibe code, why wouldn't they? So what I'm arguing is the resurgence of in house development teams, except they're not going to be developers. They're not going to be engineers. They're going to be anybody.
Speaker 1:So I think there's going to be a massive explosion of in house development. Now, if you're developing in house, now it's not one SaaS software, Calendly or Salesforce that's servicing thousands, hundreds of thousands of customers. Now you have IT people, 10, not a thousand because you don't need a thousand, maybe you have five or 10. Now you have five or 10 IT people who are coding, in vibe coding, in every single company in the world. So from, you know, sure, demand, you know, is held the same, supply completely crashes, right, but then suddenly demand explodes.
Speaker 1:That's my hypothesis of what the future holds. Now let's come back in three, four, five years and see if I was right or not, but it feels right to me.
Speaker 2:I think so. I mean, I think, and we've probably historically seen that, right, like if you ever remember seeing Yes. When IntelliSense came around. Right? Like like, autocomplete as you're coding.
Speaker 2:So for people who don't know, we used to read books about how to write a program, and then we would type in letter for letter, and you would type every single letter that appeared on the screen. And then, you know, at some point, the Internet came around so we could copy and paste the code, but you still had to write all of it. And then the the tools that we used to write programs in started getting this ability to as you would type the word, you know, foo, it would autocomplete what the suggestions would be, just like on your phone when it's auto auto suggesting things. Right? Yes.
Speaker 2:And I remember this this article at one point that was, like, arguing that this rots the minds of programmers. Like, programmers are gonna be dumber because they weren't they won't learn what the entire function to do a thing with the mouse in Windows is because they're just gonna remember the first three letters of it and the autocomplete will do it, and and they'll get dumber as as a result. What history kinda showed is actually just to have more people to become programmers and less people spending time memorizing this stuff. And more people doing, I think, with the high level job of what programming is, which is just solving problems. So I'd I'd almost argue that if you end up with a a networking engineer here and a IT admin here and a former accountant that just started doing some vibe coding over here, And they are spending all day in the purpose of solving problems better through the creation of software, that software just happens to be vibe coded or they are programmers anyway.
Speaker 2:They are. They're just using different tooling to do it, they're still programmers. They're still doing the job of programming. Yes. Yeah.
Speaker 2:And I think, again, if your if your entire sense of self esteem of, like, I'm a good developer because I memorized the name of this function in Windows
Speaker 1:Yeah.
Speaker 2:That's a problem. That ten years ago, you probably realized that wasn't that important. Today, the fact that you can write code that perfectly compiles, that shouldn't be the thing to hang yourself on anyway. What you should be really your sense of self esteem should be, I'm really good at solving problems that people present to me.
Speaker 1:Paul, you know what horrifies me? That's not how a lot of VP or CTOs hire. A lot of the CTOs and VPs, they ask, oh, do you know the jargon? Oh, do you know the API? Because they haven't figured out the fundamental truth in what you're saying that you should be asking, can you solve a problem?
Speaker 1:And it doesn't matter if you don't know the jargon. It matters if I gave you so let's say you don't know what jargon is for mutexes. I don't know what mutex is, but if I give you two loops that can happen in parallel and have a common shared data, can you tell me three problems that happened due to this? And you can spit out the problems? I shouldn't I shouldn't care if you know what mutexes are.
Speaker 2:That's right.
Speaker 1:I shouldn't.
Speaker 2:Yeah. That's right.
Speaker 1:But I hear time and time again, how do you measure your program? Oh, if they know the jargon.
Speaker 2:Yeah. Yeah. It's a shortcut. Right? Like, they know this, then they they probably maybe know how to solve the problems, but we don't have a good way of evaluating
Speaker 1:will they solve a problem?
Speaker 2:Yeah. Yeah.
Speaker 1:Right. So I can read a book and I can learn the jargon. Can I necessarily become great at solving problems faster than anybody else? Yeah. No.
Speaker 1:Yeah. The answer is no. So I don't know, I hear CTOs saying that, it horrifies me because I know a lot of great programmers who know the jargon, but they are basically savants. Right? They are sure, they're sometimes antisocial, but my Celebrite programmers, some of them, could decrypt 10 phones in the same time that another programmer could do one.
Speaker 2:Yeah, yeah.
Speaker 1:Right. I hired the team that decrypted, that hacked the Apple machine, right. So we could hack that. We gave that FBI, CA, NSA. Everybody bought it from us, the Met Police.
Speaker 1:I hired that team. So I look at those programmers and I'm like, they could do something that nobody else could do. So why do I care if they know or don't know jargon? That's right. Everybody knows jargon.
Speaker 1:Right?
Speaker 2:I think it's what people maybe don't especially people who consider themselves good programmers because they've got really good technical skills or knowledge of the how to code a son is so if you think about, like, there's a VP I I've I've said this internally sometimes when I'm hiring people who for a specific expertise, I say, like actually, so so law is a good one. Right? So some have you ever had this experience? If if you've ever had to go to a lawyer, you say to the lawyer, hey. I've got this problem.
Speaker 2:Right? I've got this contract I'm trying to do. And you give them the contract, and the contract is 20 pages. And what you get back is 30 pages of advice from them. Right?
Speaker 2:And the conversation I've often had with lawyers is, okay. I've got business problems. I don't have legal problems. I have business problems. And but I operate in a framework where the law exists.
Speaker 1:It's a thing.
Speaker 2:If I was a legal expert, I would be a total weapon in solving my business problems because I would know what I could could and couldn't do in the law.
Speaker 1:How many lawyers have you met that meet your criteria?
Speaker 2:None. Very hard. I haven't quite found someone yet. Right? So But
Speaker 1:you would have told me one, I would say, can I have his or her name?
Speaker 2:Yeah. Yeah. Yeah.
Speaker 1:Because I think Every business person in the world wants what you described.
Speaker 2:Yes. Yeah. And and that and that's and that applies for every function. So you would say, like, I'm a you know, I'm the head of revenue at this company or something. I have business problems.
Speaker 2:And if I was also an amazing coder, I could use coding to solve a lot of those problems. And that's the thing I'm paying you for, mister and missus programmer. I'm paying you to solve my problems by exploiting your knowledge of how to do these arcane things on a on a but don't lose sight of the fact that that's what I'm paying you for. I'm not paying you to code. I'm paying you to solve my problems because you have this expertise.
Speaker 2:You know? And so if you're wondering why you got a bad performance review despite writing really good code that compiles and all your unit tests pass, that's that's the problem.
Speaker 1:Paul, you are absolutely brilliant. This is not how most people work. Like most people, if the coders, engineers, developers, whatever built what they were told to build and it works, they're doing a great job. Even if it doesn't work, they're not gonna be fired. Maybe not in your organization, but most organizations, the engineers, even if they actually build stuff that fails all the time at customers, they're still not going to be fired.
Speaker 2:Yeah. Yeah.
Speaker 1:So there's a question here in the future about, you know, what does it mean to write code? What does quality mean? What does vibe coding mean? And I think what your fundamental question that you asked at the beginning to your daughter, I'm a firm believer that answers don't matter. Answers don't matter because tomorrow there's going to be a different answer but questions, questions matter fundamentally because what you asked your daughter is you asked her who are you?
Speaker 1:What are you? Why does what you do matter? And you asked it in the most simplest of ways. What you know, what does it
Speaker 2:Yep. Yeah. And what's the highest level answer to it and the lowest level answer to it and focus on the high level thing. So here's if you're a good programmer today and let's say that you're dabbling with AI tools and you're finding that they make you more productive Yeah. What do you do with that productivity boost that you've got?
Speaker 2:Right? Do you spend it writing more code, or do you spend it spending more time with the customer to understand their problem better? There's this book Domain Driven Design by Eric Evans, and it's a it's a sort of popular book within the software community as people get more advanced in their programming skills.
Speaker 1:Yes.
Speaker 2:And what it does is it it, the first the first part of the book talks about the importance of deeply understanding the problem domain that you're working in. Right? So you you wanna go and meet the people doing the job that you're coding for. And when they use a strange word that you haven't heard before, double click and try and understand what exactly it means and why they aren't using a different word
Speaker 1:because there's a The double click in the context is a very engineering term you're using. Double click in this context means use the wide tech technique in order to understand deeply into what is actually happening here. Paul, I can't believe I'm gonna do this. A, I'm cutting you off because we're at the end of our hour and I have another guest in a minute. So you're getting a third invite.
Speaker 1:That is the first time ever this show has ever happened. This is
Speaker 2:I loved it.
Speaker 1:Paul, you are brilliant. We haven't even talked about risk and compliance
Speaker 2:Nice.
Speaker 1:Which we really should have. But that's what you get when you have a curious host and a brilliant guest. Paul, thank you so much for joining the show today.
Speaker 2:Thanks for having me. It's lot of fun.