WEBVTT 1 00:00:18.640 --> 00:00:19.840 Hey, Ben. 2 00:00:19.840 --> 00:00:21.240 Hey Matt. 3 00:00:21.240 --> 00:00:22.400 How are you today? 4 00:00:22.400 --> 00:00:23.580 I'm doing alright. 5 00:00:23.580 --> 00:00:25.230 Excellent. Excellent. 6 00:00:25.230 --> 00:00:25.740 Thank you. 7 00:00:25.740 --> 00:00:30.900 Well, we... we surprisingly have a topic today. In fact, we went out for drinks last night. 8 00:00:30.900 --> 00:00:30.900 Mm-hmm. 9 00:00:30.900 --> 00:00:47.520 And we discussed many things, some of which we wish we had a microphone for so that we could have just recorded it and put it out to our good audience here. Much of it... was absolutely unbroadcastable for a variety of reasons, usually around intellectual property and or just... 10 00:00:47.520 --> 00:00:48.190 Yes, right. 11 00:00:48.190 --> 00:00:49.200 inappropriateness. 12 00:00:49.200 --> 00:00:50.100 Too many drinks, that's... 13 00:00:50.100 --> 00:00:51.850 Too many drinks, yeah. 14 00:00:51.850 --> 00:00:52.260 Yeah. 15 00:00:52.260 --> 00:00:53.080 Ha ha. 16 00:00:53.080 --> 00:01:01.100 But one of the things that you brought up, which I thought was really interesting... Well, I'm gonna let you say it. What is your topic for today, Ben? 17 00:01:01.100 --> 00:01:05.760 Is programming still fun? 18 00:01:05.760 --> 00:01:08.560 Is programming still fun? And is this... 19 00:01:08.560 --> 00:01:09.220 Yeah. 20 00:01:09.220 --> 00:01:16.260 Do you mean... Have you gotten bored of programming? You know, now you're... ancient and old. You'll never be as old as me, but... 21 00:01:16.260 --> 00:01:20.640 Yeah, ancient and old, and you've gotten tired of it. Yeah, you just want to do, I don't know, weaving or... 22 00:01:20.640 --> 00:01:28.080 I don't... You know, it's sort of the, you know, the cliche breakup line is it's not you, it's me. 23 00:01:28.080 --> 00:01:29.540 It's the opposite of that. 24 00:01:29.540 --> 00:01:29.540 Ha ha! 25 00:01:29.540 --> 00:01:29.540 Ha ha. 26 00:01:29.540 --> 00:01:32.580 It's not me. It's you, comma, Claude. 27 00:01:32.580 --> 00:01:32.840 Oh. 28 00:01:32.840 --> 00:01:34.800 Oh. 29 00:01:34.800 --> 00:01:41.750 So I am posing this question specifically in the context of these modern, powerful tools that we have. 30 00:01:41.750 --> 00:01:42.320 Yep. 31 00:01:42.320 --> 00:01:52.840 Which I'm certainly not suggesting that we give them up. But is a side effect of this that programming is no longer fun? 32 00:01:52.840 --> 00:02:08.910 Or... You know, and I... I don't think that there is an obvious answer to this question. This is not like a super leading question. It also may be in flagrant violation of Betteridge's law of headlines, which is the answer is not obviously no. 33 00:02:08.910 --> 00:02:10.060 Interesting. 34 00:02:10.060 --> 00:02:18.000 The answer is not also obviously yes. So it's like, I think it's worth a podcast to talk about. 35 00:02:18.000 --> 00:02:21.860 Yeah, worth a podcast episode. So yeah, I mean, you and I obviously, we spoke about this 36 00:02:21.860 --> 00:02:21.860 Yeah. 37 00:02:21.860 --> 00:02:46.520 last night a little bit. And I think... As we've alluded to in other episodes, we've both... been on a journey of discovery through this process, as have many people in our position, where, you know, first of all, we're in the luxurious position of we have established careers, and so we can look at it from that point of view. And there's a whole bunch of topics that go in the direction of like, but what does this mean for the next generation? 38 00:02:46.520 --> 00:02:50.760 I don't propose we talk about that today. But from a very selfish, 39 00:02:50.760 --> 00:03:04.160 individualistic point of view, has the advent of coding assistants of all sorts, has that taken away the thing that we actually enjoy about 40 00:03:04.160 --> 00:03:06.340 programming? 41 00:03:06.340 --> 00:03:07.290 So let me start with this. 42 00:03:07.290 --> 00:03:07.480 Mm-hmm. 43 00:03:07.480 --> 00:03:13.180 Okay, so back in the day, I would edit code in black and white with no syntax highlighting. 44 00:03:13.180 --> 00:03:13.180 Right. 45 00:03:13.180 --> 00:03:26.350 In a vi terminal or in an editor, a very straightforward editor. And it was hard work. Right? It was hard work to hold what I needed to hold in my head. 46 00:03:26.350 --> 00:03:27.000 Mm-hmm. Yeah. 47 00:03:27.000 --> 00:03:36.900 I sort of squinted to be able to understand the code that I was working on. And the first... improvement in my life as a developer 48 00:03:36.900 --> 00:03:42.020 was syntax highlighting and or colorization of what I was working on, I think. 49 00:03:42.020 --> 00:03:42.020 Yeah. 50 00:03:42.020 --> 00:03:53.020 StrongED was the name of the editor that I was using. And it was like an absolute step change in my ability to look at a page of code and kind of get the gist of it straight away. And that was... 51 00:03:53.020 --> 00:04:08.160 unambiguously an improvement. Then... Time ticked on. And the first sort of next, rather, the next stop point, I think, where like tooling came along and changed the way we program was 52 00:04:08.160 --> 00:04:17.180 IntelliSense. And that materially changed the way... I didn't have to remember all of the documentation or forever go hunting for it. 53 00:04:17.180 --> 00:04:17.280 Mm-hmm. 54 00:04:17.280 --> 00:04:17.720 I would type 55 00:04:17.720 --> 00:04:23.640 object name dot and then hit control space, and I'd get a list of all the things that that object could do. 56 00:04:23.640 --> 00:04:24.380 Yeah. 57 00:04:24.380 --> 00:04:35.360 And I remember people at that time saying, oh, this is... You don't really know it anymore. You're just kind of feeling your way through the problem. Right. And that means that you're not thinking about it holistically. 58 00:04:35.360 --> 00:04:35.480 Uh-huh. 59 00:04:35.480 --> 00:04:41.110 You don't really understand it. You're just kind of auto-completing your way into a solution, right? 60 00:04:41.110 --> 00:04:41.260 Right. 61 00:04:41.260 --> 00:04:56.070 And that's another step change in the way that we do things. And then maybe we could say refactoring tools have a similar feel, but I could argue that maybe they're more freeing, I don't know. And then obviously we get to the point now where I, rather than thinking about something myself, I say, 62 00:04:56.070 --> 00:04:56.540 Mm-hmm. 63 00:04:56.540 --> 00:05:05.430 Hey, agent, can you refactor this thing for me? And then I go away and make a cup of tea, or more likely I have four or five of them going and I'm doing something kind of different. 64 00:05:05.430 --> 00:05:05.680 Yeah. 65 00:05:05.680 --> 00:05:09.760 And maybe that's the thing. What do you think about that? 66 00:05:09.760 --> 00:05:22.190 One of the observations that we had last night is that I certainly find, and I think other people have found this as well, that it is very difficult to be in a flow state while programming with agents. 67 00:05:22.190 --> 00:05:23.360 Yes. 68 00:05:23.360 --> 00:05:32.120 I think I've had some moments where it sort of comes back a little bit and it feels a little bit like a video game in the sense of like, you know, you're interacting in like a frequent enough 69 00:05:32.120 --> 00:05:32.330 Yes. 70 00:05:32.330 --> 00:05:44.620 way to where you get a little bit of it. But it is definitely not the same feeling of, you know, putting on a pair of headphones with the, you know... 71 00:05:44.620 --> 00:05:46.640 Tron soundtrack on. 72 00:05:46.640 --> 00:05:48.080 Oh, excellent choice, sir. 73 00:05:48.080 --> 00:05:58.200 And... you know, sitting down with the editor or IDE of your choice and solving a problem and being in flow state the entire time. 74 00:05:58.200 --> 00:06:08.690 And, you know, especially like some of the practices that I had built up in my career around, you know, test driven development and continuous testing and running tests 75 00:06:08.690 --> 00:06:09.060 Right. 76 00:06:09.060 --> 00:06:20.400 on a continuous basis and getting that feedback loop, where you have like the constant feedback loop of like take a very small step, get feedback on that step, take another very small step. That can very quickly put you into a flow state. 77 00:06:20.400 --> 00:06:20.800 Mm-hmm. 78 00:06:20.800 --> 00:06:32.320 Which many people, including myself, find the flow state to be very enjoyable. That is half the reason I play video games is because... they put you into that state and it's fun. It's just fun to be in that state. 79 00:06:32.320 --> 00:06:41.130 A friend of mine referred... This is in the context of playing video games, usually on the Game Boy, which tells me exactly how long ago this was. 80 00:06:41.130 --> 00:06:41.420 Right. 81 00:06:41.420 --> 00:06:58.870 But at the time I was commuting a very long distance on the train, and my friend pointed out that it's okay because you have a time travel device, which is that you have a Game Boy in your hands. You put on Pokemon and then before you know it, you're about to miss your stop an hour later, right? And I think that's true. That's what that flow state is. It's like... 82 00:06:58.870 --> 00:06:59.180 Yeah. 83 00:06:59.180 --> 00:07:00.440 Right. Right. Yeah. 84 00:07:00.440 --> 00:07:10.920 Yeah, I definitely feel you on the flow state. And in particular, when I am managing multiple sessions at once, the context switching overhead is just decimatory 85 00:07:10.920 --> 00:07:10.920 Yeah. Yeah. 86 00:07:10.920 --> 00:07:12.680 to flow. 87 00:07:12.680 --> 00:07:13.280 Yes. 88 00:07:13.280 --> 00:07:34.900 I think the times where I get into flow where I'm working with agents is when I have the screen split in half, and I have my IDE up and I'm still reading the code, and maybe I'm even commenting the code, I'm looking at the code, while meanwhile I have more of a pair programmer, but a pair programmer that I can alt-tab to rather than have to like speak to. And that's like the nearest thing. But then 89 00:07:34.900 --> 00:07:35.530 Yeah. Yeah. 90 00:07:35.530 --> 00:07:35.840 Right. 91 00:07:35.840 --> 00:08:01.540 that feels like it's not as efficient, if efficiency is what you're looking for, as working on five things at once and getting, you know, progress on multiple fronts. So it feels like maybe I'm not doing it right, you know? I'm still doing the old school programming, it's just I've got Claude to help me, you know, and like, oh, can you fix this, make that test pass, now actually I've now changed the code, that feel. And then it's kind of back and forth. I don't know. Is... 92 00:08:01.540 --> 00:08:01.600 Yeah. 93 00:08:01.600 --> 00:08:05.120 Is that close to your experience when you said you could get into a flow state, or are you thinking of something else? 94 00:08:05.120 --> 00:08:27.660 Yeah, that is close to my experiences when you start doing that sort of more, almost pair programming type approach. And in my day job, a couple of weeks ago I actually did a presentation at work because we've had a lot of folks that, you know, have lots of questions about like, you know, how should we be using agents at work and, you know, other things like that. 95 00:08:27.660 --> 00:08:33.480 And I did a presentation that I called Agentic Patterns. And it was like, 96 00:08:33.480 --> 00:08:56.300 these things, you know, like we've talked a bunch of times about programming patterns and, you know, software development patterns. Like you've got, you know, architectural patterns like the singleton and the flyweight and all of those other kinds of things, but you've also got patterns that are essentially practices, you know, test driven development, refactoring, continuous integration, continuous deployment, pair programming. These things are all kinds of patterns really. 97 00:08:56.300 --> 00:09:08.020 And so, you know, I tried to sit down and like, with lots and lots of caveats and asterisks, saying like, these are some things that kind of seem to work for me right now, given the current technology, but all of this is going to be different. 98 00:09:08.020 --> 00:09:08.040 Mm-hmm. 99 00:09:08.040 --> 00:09:10.030 And I have no idea what the long-term ramifications of this is. 100 00:09:10.030 --> 00:09:11.210 Right, I suppose unlike like Gang of Four... 101 00:09:11.210 --> 00:09:11.360 Yes. 102 00:09:11.360 --> 00:09:17.380 style patterns and practices, whatever, all these things change week on week, let alone maybe, if not day on day. 103 00:09:17.380 --> 00:09:23.500 Right. Yes. Ah yes. I am making no claims whatsoever that these are the patterns or that even that these are good patterns for everybody. 104 00:09:23.500 --> 00:09:31.380 They just happen to work for me in certain contexts. And so I wanted to at least start a conversation about like sharing them and maybe giving them some names, right? 105 00:09:31.380 --> 00:09:31.720 Right. 106 00:09:31.720 --> 00:09:42.360 And so this sort of pattern of like when you are essentially pair programming with the agent, right? Like there's one agent running, you're reviewing the code fairly closely. 107 00:09:42.360 --> 00:09:56.040 You're doing things like adding TODO comments into the code and saying like, hey, fix this, fix this, fix this. You're very much focused on like method and class signatures and function signatures as a way to drive the direction of the agent rather than strictly just using English. 108 00:09:56.040 --> 00:09:58.580 Right. 109 00:09:58.580 --> 00:10:08.220 You know, like that kind of interaction, I definitely have had situations like that where it feels like pair programming in a lot of ways, and you definitely can get into a flow. 110 00:10:08.220 --> 00:10:18.410 But those are... That's not the everyday, right? Like that's a situational type of thing where you're like, it's really important that I get this right, or I'm trying to understand this, or I'm trying to like 111 00:10:18.410 --> 00:10:18.940 Yeah. Yeah. 112 00:10:18.940 --> 00:10:32.740 fix a problem or something like that. That's not the normal default mode, at least that I have when I'm doing agentic programming. And so most of the time, that means these days I'm not in that flow state. 113 00:10:32.740 --> 00:10:35.600 Right. Which is very different than it was two years ago. 114 00:10:35.600 --> 00:10:45.060 Right. Right. And I mean, especially, I mean, I'm seeing more of a push towards sort of the software factory approach where, you know, you have Slack integration or... 115 00:10:45.060 --> 00:10:45.060 Yeah. 116 00:10:45.060 --> 00:11:11.000 or whatever, and you kind of, you're kind of like doing one shot, like, hey, just fix this comment, right, to a thing that then makes a pull request to your repo or whatever, and then some other bot reviews it, and then you once over and go, yeah, that's about right, and then you check it in. And all that, the software factory of, like, almost like a Kanban board of, here's all the tickets that are in progress, and I'm just basically the ticket 117 00:11:11.000 --> 00:11:11.000 Mm-hmm. 118 00:11:11.000 --> 00:11:13.290 administrator stroke 119 00:11:13.290 --> 00:11:14.440 Mm-hmm. 120 00:11:14.440 --> 00:11:24.660 last sign-off part of the process. And it feels... like that is not, yeah, I'm... I'm a mediocre product, a project manager 121 00:11:24.660 --> 00:11:26.120 Yeah, yeah, yeah. 122 00:11:26.120 --> 00:11:28.630 at best, and I'm an adequate code reviewer. 123 00:11:28.630 --> 00:11:28.940 Yeah. 124 00:11:28.940 --> 00:11:43.190 In fact, I like to think... I'm better than adequate code reviewer, but... The thing that I like about reviewing code is the ability, first of all, to learn something new, which maybe I can do still with agentic 125 00:11:43.190 --> 00:11:43.300 Mm-hmm. 126 00:11:43.300 --> 00:11:57.910 code. Maybe. But also the ability to mentor and shape and help others become better. So like it's a two-way thing. I learn something, they learn something. And especially, you know... 127 00:11:57.910 --> 00:11:58.460 Yeah. Yeah. 128 00:11:58.460 --> 00:12:10.950 And that just doesn't happen. That side of it doesn't happen. No matter how many times I implore an agent to stop writing ludicrous comments that explain the things that I told it not to do rather than the thing that it is actually doing. 129 00:12:10.950 --> 00:12:12.110 Yes, right. Uh huh. 130 00:12:12.110 --> 00:12:17.150 It's not going to learn that. And that is unsatisfying from a human point of view. 131 00:12:17.150 --> 00:12:17.500 Yeah. 132 00:12:17.500 --> 00:12:22.240 And I know that's nothing to do with flow necessarily, but it's like if we're looking for what makes 133 00:12:22.240 --> 00:12:22.580 Yeah. 134 00:12:22.580 --> 00:12:25.690 our job, rather than programming per se, 135 00:12:25.690 --> 00:12:26.470 Right. 136 00:12:26.470 --> 00:12:38.640 rewarding and fun. Then... I don't mind a good back and forth with a human, knowing that we can go for a drink afterwards and kind of have a chat about that silly thing that we discovered in the blah blah blah blah. That's cool. 137 00:12:38.640 --> 00:12:39.340 Right. 138 00:12:39.340 --> 00:12:55.060 But I won't have that with my army of agents. That's not the same experience. And so if I'm trying to rationalize my new role as a middle manager of a bunch of robots, then maybe I don't get even the benefit of that out of it. 139 00:12:55.060 --> 00:12:55.360 Right. 140 00:12:55.360 --> 00:12:55.660 Yeah. 141 00:12:55.660 --> 00:12:55.960 So. 142 00:12:55.960 --> 00:13:08.340 Yeah, well, and that is another dimension. Like of all the things that makes programming fun, what are some things? Well, flow is one thing. The ability to mentor and teach through sort of like the craft of software development is another one. 143 00:13:08.340 --> 00:13:18.660 You know, I think there's definitely something, I don't exactly know how to describe this, but there's definitely something about like just tickling that part of your brain when you understand a complex thing. 144 00:13:18.660 --> 00:13:26.840 And you still get some of that, but you don't get it at the same level that you got it before. Right. 145 00:13:26.840 --> 00:13:29.950 Like... You've always had to understand the architecture. 146 00:13:29.950 --> 00:13:30.300 There is... 147 00:13:30.300 --> 00:13:30.640 You know, 148 00:13:30.640 --> 00:13:43.120 But you also had to understand the code underneath that architecture. And quite often if you had to choose between the two, you kind of started from the bottom and worked your way up, because that's... kind of like the only way to really get it. 149 00:13:43.120 --> 00:13:52.300 And now it seems like, yeah, you don't really. In many cases, not all cases, but in many cases you can be very ignorant of how the code actually works 150 00:13:52.300 --> 00:13:54.560 and still produce working software. Right? 151 00:13:54.560 --> 00:13:55.050 Yeah. 152 00:13:55.050 --> 00:13:56.040 Right. 153 00:13:56.040 --> 00:13:58.520 Which is weird. 154 00:13:58.520 --> 00:14:10.260 And I think that's... It is weird and... One of the things that... I received feedback recently about the fact that there were some parts of the code that I clearly didn't know as well as I perhaps should have done. 155 00:14:10.260 --> 00:14:15.520 And it was because while I'd reviewed it, I had... It was not the same, you know? Like, yeah, people 156 00:14:15.520 --> 00:14:15.680 Yeah. 157 00:14:15.680 --> 00:14:34.350 joke about, you know, writing stuff down. You know, I've got... pads of paper around me right now. I write post-it notes and stick them to the screen, even though I have a Jira system to do ticketing in, which is far more convenient because it follows me around, unlike the piece of paper that's stuck to my computer back in the office, right? 158 00:14:34.350 --> 00:14:35.160 Right. Yeah. 159 00:14:35.160 --> 00:14:40.380 But there's something visceral about having done it yourself, even if it's as mechanical as writing it down. 160 00:14:40.380 --> 00:14:40.880 Yeah. 161 00:14:40.880 --> 00:14:53.610 But there's also, if you typed it yourself, I'm sure that's another layer of it. And even moreover, if you thought about it yourself, you thought about it in a different way, you experienced it, you went through it, you are more... 162 00:14:53.610 --> 00:14:53.860 Yep. 163 00:14:53.860 --> 00:14:54.600 Right. 164 00:14:54.600 --> 00:15:04.700 simpatico with it. And then... We've said this before. It's like managing, right? Like when I've managed a small team before now, I know what my individual reports are doing. 165 00:15:04.700 --> 00:15:16.460 I've reviewed their code. But if I go into a meeting with a bunch of other middle manager people and we talk about the project, they go, well, how exactly does X work? I'll be like, we're going to have to go and ask. 166 00:15:16.460 --> 00:15:17.870 Yeah, the people who actually know. 167 00:15:17.870 --> 00:15:18.200 Of, you know. 168 00:15:18.200 --> 00:15:18.420 Yeah, right. 169 00:15:18.420 --> 00:15:28.580 Yeah, I don't know it well enough to teach it. And yet I don't, I can't go and ask, at least the exact agent, the one I worked with. And obviously, I can go and ask an agent. 170 00:15:28.580 --> 00:15:29.340 Right. 171 00:15:29.340 --> 00:15:33.360 It's unsatisfying and I feel fraudulent, honestly. 172 00:15:33.360 --> 00:15:33.860 Yeah. Yeah, yeah yeah. 173 00:15:33.860 --> 00:16:00.780 If that's the right... phrase. Now I've just, and this is becoming new, new experience coming to me as I'm talking to you. I think... I'm worried that... if I... don't use the agents, I won't be as productive as I could be, and I will be measured, but by myself mainly, no one's asking me to be, but as being not as good as I could be. And certainly you and I have in our wider friend group some amazing engineers who are doing some incredible things with 174 00:16:00.780 --> 00:16:05.150 AI technology writ large, right? 175 00:16:05.150 --> 00:16:06.360 Mm-hmm. 176 00:16:06.360 --> 00:16:16.380 But... I feel fraudulent for like passing off someone else's work as mine, even though I have spent ages... reading it, crafting it, 177 00:16:16.380 --> 00:16:16.840 Yeah. 178 00:16:16.840 --> 00:16:20.510 and explaining back and forth and ideating and going, that's rubbish. 179 00:16:20.510 --> 00:16:36.020 No, we're going to have to try again. I still feel like it's not mine. And then when I do represent it, even though I could, I feel confident in what I've done. And I don't know how to square that circle because maybe I... Maybe it is enjoyable. Sometimes it is enjoyable to be doing a little bit of that iterating. 180 00:16:36.020 --> 00:16:40.280 Yeah, but has the fun come out of it? Because it's not mine. 181 00:16:40.280 --> 00:16:48.960 Yeah, I mean, so there is... there's a powerful... element there. You know, and I could, I think. 182 00:16:48.960 --> 00:16:49.080 A powerful elephant? 183 00:16:49.080 --> 00:16:51.540 A powerful elephant, right? 184 00:16:51.540 --> 00:16:51.620 Wow. 185 00:16:51.620 --> 00:16:54.950 There's an elephant in the room and we're all trying to feel the different parts of the elephant. 186 00:16:54.950 --> 00:16:55.620 I've... and I've definitely... 187 00:16:55.620 --> 00:16:57.380 And it feels very powerful. 188 00:16:57.380 --> 00:17:08.200 I've definitely got, like, there's the elephant with an extra I in it, E-L-E-P-H-A-I-N-T, right, for the AI in the room. I think there's something in there. I've got a slide somewhere that's... anyway, never mind. Carry on. 189 00:17:08.200 --> 00:17:17.500 Uh huh, right. Yes. There's a... I forget where this came from, but there was a poster or something that's been popping back in my mind of the Octophant, 190 00:17:17.500 --> 00:17:23.330 which is an elephant on top and an octopus on bottom. And it's carrying a little flag that says, I should not be. 191 00:17:23.330 --> 00:17:23.930 Whoa. 192 00:17:23.930 --> 00:17:24.520 Whoa. 193 00:17:24.520 --> 00:17:26.140 I forget where that's from. 194 00:17:26.140 --> 00:17:26.540 Well, I... 195 00:17:26.540 --> 00:17:28.160 Where is the Octophant from? 196 00:17:28.160 --> 00:17:38.860 There was a little post-it note on someone's desk when I went to Google, and it was just a little... caricature of an elephant, and underneath it it said, anything that's not about elephants 197 00:17:38.860 --> 00:17:41.740 is irrelevant. It just tickled me. 198 00:17:41.740 --> 00:17:41.900 Yeah. 199 00:17:41.900 --> 00:17:42.500 I like that. 200 00:17:42.500 --> 00:17:43.560 And I still think about it. 201 00:17:43.560 --> 00:17:45.320 Anyway, the elephant in the room. 202 00:17:45.320 --> 00:17:45.580 Yeah. 203 00:17:45.580 --> 00:17:46.320 Where were we with this? 204 00:17:46.320 --> 00:17:46.360 No, the thing that I have to wonder is... 205 00:17:46.360 --> 00:17:51.780 There was some point we were having. 206 00:17:51.780 --> 00:18:04.760 I've gone back and forth on this over the last few decades of my career of... There are lots of people that I have worked with that are very good at their jobs 207 00:18:04.760 --> 00:18:15.330 that do not code in their free time. It is strictly a profession for them, right? Like they use it to, you know, pay their rent, and it's like, they go home and they do other things. 208 00:18:15.330 --> 00:18:16.560 Right. Hmm? 209 00:18:16.560 --> 00:18:24.520 But I have worked with a lot of people who are very good. And also... you would have to pay them not to code. 210 00:18:24.520 --> 00:18:24.840 Yeah. 211 00:18:24.840 --> 00:18:40.340 Right? Like, they just do it all the time. I think, right, a good chunk of the open source software that modern society rests on comes from these types of people, where they're just like, I am interested in solving problems and I will do it for free. 212 00:18:40.340 --> 00:18:40.480 It's... 213 00:18:40.480 --> 00:18:44.980 And it's kind of cool that I get paid for it, but I will do it for free because I find it to be so much fun. 214 00:18:44.980 --> 00:18:48.230 Right. This, like, these... It just comes from them. 215 00:18:48.230 --> 00:18:48.300 Right. 216 00:18:48.300 --> 00:18:50.100 It's like artists that have to express themselves through art. 217 00:18:50.100 --> 00:18:50.100 Yes. 218 00:18:50.100 --> 00:18:57.380 It's like I have to express myself through code. That's kind of, I mean... Yeah, I feel that way, but the people you're talking about 219 00:18:57.380 --> 00:18:57.860 Exactly. 220 00:18:57.860 --> 00:19:01.940 seem to act that way too. And then, yeah, society sort of takes advantage of their better nature 221 00:19:01.940 --> 00:19:02.980 Right, 100%. 222 00:19:02.980 --> 00:19:04.860 by saying, thank you so much for your compression library. 223 00:19:04.860 --> 00:19:10.990 Like... What are all these LLMs trained on? It's trained on all of that, you know, despite what people may claim. 224 00:19:10.990 --> 00:19:11.200 Oh, God, yeah. 225 00:19:11.200 --> 00:19:12.750 I mean, it pretty obviously is. 226 00:19:12.750 --> 00:19:13.180 Yeah. 227 00:19:13.180 --> 00:19:24.860 And so I think a strong argument can be made that this phenomena that we're experiencing here is essentially the... the death of open source in the way that it has been, because it does two things. 228 00:19:24.860 --> 00:19:34.490 One is it removes the fun element of creating these things. And two, it makes the actual efficient way to create solutions like this very expensive in the form of tokens, which it wasn't before. 229 00:19:34.490 --> 00:19:34.860 Mm hmm. 230 00:19:34.860 --> 00:19:35.040 Ah. 231 00:19:35.040 --> 00:19:46.720 Another thing we were talking about in the pub last night is... Our profession has had this amazing, wonderful property. And I think we maybe have talked about this on a separate episode, of you can just kind of get like a laptop. 232 00:19:46.720 --> 00:19:54.280 And that's it. And now you've paid this one fixed cost and yeah, you get some open source tools and you can build a lot of things with just that. 233 00:19:54.280 --> 00:19:55.080 I mean... 234 00:19:55.080 --> 00:20:04.800 Heck, you can get a $20... or $40, like, Raspberry Pi, plug a monitor into it, and you've got a good enough development kit to be able to get on with... Yeah, it's... yeah, yeah. Yeah. 235 00:20:04.800 --> 00:20:04.980 Right. Right. 236 00:20:04.980 --> 00:20:08.000 Yeah, the barrier to entry is... is fairly low. 237 00:20:08.000 --> 00:20:10.180 Yeah. Yeah, compared to a lot of other... 238 00:20:10.180 --> 00:20:14.930 There are even, apparently, websites you can use that do this. You know, as long as you've got internet access, you don't even need a computer as such. 239 00:20:14.930 --> 00:20:15.780 All right. 240 00:20:15.780 --> 00:20:17.120 You know, you can do it, albeit badly, on your phone. 241 00:20:17.120 --> 00:20:17.180 Exactly. 242 00:20:17.180 --> 00:20:21.120 I'm sorry, everybody who tries to use my site on the phone. Anyway, but yeah, so the barrier to entry has been... 243 00:20:21.120 --> 00:20:21.120 Yes. 244 00:20:21.120 --> 00:20:23.140 very low for a long time. 245 00:20:23.140 --> 00:20:24.060 Quite low. And I... 246 00:20:24.060 --> 00:20:26.650 In fact, lower than perhaps it was when we were growing up, right? 247 00:20:26.650 --> 00:20:26.680 Yeah. 248 00:20:26.680 --> 00:20:26.760 When I was... 249 00:20:26.760 --> 00:20:28.620 I think so. 250 00:20:28.620 --> 00:20:33.290 When I was, you know, single digit years old and I got a computer, I was incredibly fortunate 251 00:20:33.290 --> 00:20:33.440 Yeah. 252 00:20:33.440 --> 00:20:45.720 that I was in a position where my parents could afford, and did, and that's, you know, so life... life-changingly, path-settingly, but you know, that was a privilege. And... Yeah. Anyway. 253 00:20:45.720 --> 00:21:07.620 Yeah. Yeah, yeah, yeah. And now I have to wonder if all of the forces that created the open source ecosystem that everything now rests on are being removed, because you no longer have the sort of fun element that people get when they're like, oh, I'm going to go solve this problem and I'm going to build this open source library and it's going to be really fun. 254 00:21:07.620 --> 00:21:08.300 I'm going to enjoy it. 255 00:21:08.300 --> 00:21:12.710 And oh, I guess maybe someone else will use it. I don't even care if anyone else uses it. I want to build this for me. 256 00:21:12.710 --> 00:21:12.950 Yeah. 257 00:21:12.950 --> 00:21:13.200 Yeah. 258 00:21:13.200 --> 00:21:22.090 Combined with, well, the actual way that people build software in the modern world is you use an LLM and that costs a bunch of money and, you know, there's all these dependencies. 259 00:21:22.090 --> 00:21:22.320 Yeah. 260 00:21:22.320 --> 00:21:24.600 I think we talked about that in the last episode. And so... 261 00:21:24.600 --> 00:21:24.600 Yeah. 262 00:21:24.600 --> 00:21:35.440 I think there's a lot of kind of elements of funness that could be getting removed through efficiency. 263 00:21:35.440 --> 00:21:36.820 Right? And... 264 00:21:36.820 --> 00:21:36.820 Yeah. 265 00:21:36.820 --> 00:21:42.900 Well, let me tell you what, let's, one sort of counter to that, if I may. 266 00:21:42.900 --> 00:21:43.080 Okay, yeah. 267 00:21:43.080 --> 00:21:54.720 And that is... lumber mills or CNC machines or whatever you... Insert your big technology that came and changed the whole industry around type thing. 268 00:21:54.720 --> 00:22:01.110 3D printers. There's another example of like, here is a transformative thing that means that nobody needs to do like whittling things down. 269 00:22:01.110 --> 00:22:01.220 Hmm? 270 00:22:01.220 --> 00:22:13.660 They don't, you know, you don't have to machine it from like parts and send it off to whatever. Those have come. But if you enjoy that kind of stuff, I mean, I sit and watch loads of YouTube videos on people who sit and machine these beautiful little parts because they enjoy 271 00:22:13.660 --> 00:22:25.440 the making of it. You know, there are people who make watches from scratch out of little tiny pieces, and how do you do this? It's like, you could just go into a shop and buy something that's been hand, you know, not handmade, machine made, for like 272 00:22:25.440 --> 00:22:25.440 Mm-hmm. 273 00:22:25.440 --> 00:22:27.520 nothing. 274 00:22:27.520 --> 00:22:28.440 Right. 275 00:22:28.440 --> 00:22:33.130 But you enjoy the process. And so you choose to do it that way. And I respect that. There are people who crochet stuff. 276 00:22:33.130 --> 00:22:33.360 Right. 277 00:22:33.360 --> 00:22:38.410 I mean, there are people who knit stuff, even though knitting machines are a thing. 278 00:22:38.410 --> 00:22:38.600 Yes. 279 00:22:38.600 --> 00:22:40.400 And, you know, and it's almost ridiculous. 280 00:22:40.400 --> 00:22:40.660 Right. 281 00:22:40.660 --> 00:22:46.140 You're paying for the privilege of spending hours and hours to make a garment that you could wear yourself. 282 00:22:46.140 --> 00:22:46.440 Right. 283 00:22:46.440 --> 00:22:46.740 Right. 284 00:22:46.740 --> 00:22:47.040 Yes. 285 00:22:47.040 --> 00:22:50.270 People do this because they love it, right? And I think... 286 00:22:50.270 --> 00:22:50.680 Right. 287 00:22:50.680 --> 00:23:01.020 This is something else we touched on, I think, last night when we talked about like how maybe specifically for programmers, this is different from other uses of this kind of technology. 288 00:23:01.020 --> 00:23:16.700 In as much as like we still have a job out of this, right? We're still able to at least, you know, again, in our privileged position. Whereas... you know, and I don't know, hobby knitters will never become... 289 00:23:16.700 --> 00:23:27.740 I don't know. Maybe I don't know where I'm going with this, but there feels like there's a difference between folks who have done all right out of the technology and folks who... now have been having their jobs totally taken away. 290 00:23:27.740 --> 00:23:30.800 And maybe I'm naive and ours will go and that will be the end of it. I'm not sure. 291 00:23:30.800 --> 00:23:48.320 Well, yeah, my concern is that handcrafted code, which I almost am reluctant to use that phrase, but I don't know a better way to describe it, is going to become a hobby. That's not something that anyone would ever do professionally, except in very rare circumstances. That's my worry. 292 00:23:48.320 --> 00:23:50.540 Because that's been my experience for like the last 293 00:23:50.540 --> 00:23:50.540 Right. 294 00:23:50.540 --> 00:24:01.620 year, nine months, where, you know, 90 to 95% of the code that is being created, both by me and the programmers around me, 295 00:24:01.620 --> 00:24:11.850 is being created by agents and not in this sort of handcrafted hobbyist way. Right. Whereas before there was no real distinction between the hobby and the profession. 296 00:24:11.850 --> 00:24:12.040 Yeah. 297 00:24:12.040 --> 00:24:14.860 Right. Like you loved building software. You did that by writing code. 298 00:24:14.860 --> 00:24:17.610 That's exactly what you would do in your day job. It's the same thing. 299 00:24:17.610 --> 00:24:18.680 Right. 300 00:24:18.680 --> 00:24:38.090 If you, yeah, if you were a master craftsman and you spent all day building chairs with beautiful dovetail joints and whatever, and you still loved it, you went home, you'd do the same thing at home. But now you work in the IKEA factory and you're feeding the machines that build the flat pack, and then you go home and then you do something completely different that sort of only overlaps in the Venn diagram by 10% of what you do. 301 00:24:38.090 --> 00:24:38.100 Right. 302 00:24:38.100 --> 00:24:38.180 Right. Right, yeah. Yeah, yeah. 303 00:24:38.180 --> 00:24:41.550 Right. Right. Maybe you could break even on your costs by selling your chairs at the local... 304 00:24:41.550 --> 00:24:41.560 Yeah. 305 00:24:41.560 --> 00:24:43.500 art fair. 306 00:24:43.500 --> 00:24:46.770 But it's not a profession, not really. 307 00:24:46.770 --> 00:24:47.180 Yeah. 308 00:24:47.180 --> 00:24:54.730 And maybe, I mean, maybe there will be a market. I mean, I've seen a bit of, you know, there's obviously backlashes against AI usage in many, many places. 309 00:24:54.730 --> 00:24:55.180 Mm-hmm. 310 00:24:55.180 --> 00:24:59.370 And there may be some, you know, certified 100% organic free range code, right? 311 00:24:59.370 --> 00:25:00.000 Right? Yes. 312 00:25:00.000 --> 00:25:01.540 That's been written by real... 313 00:25:01.540 --> 00:25:03.800 Artisanally crafted from the finest ones and zeros, you know? 314 00:25:03.800 --> 00:25:05.190 Absolutely. Yeah, exactly. 315 00:25:05.190 --> 00:25:06.360 Yeah. 316 00:25:06.360 --> 00:25:16.360 Yeah. Not these... Yeah, maybe there will be a market for that. But yeah, unlike many other things, I think, because the output is... Well, I guess that's a debate is whether the output is better or not. 317 00:25:16.360 --> 00:25:16.540 You know, this is... 318 00:25:16.540 --> 00:25:16.580 Right. 319 00:25:16.580 --> 00:25:23.380 Who knows? I mean, we're in a period of discovery, but, you know, we're also in a period of where it changes week on week on week. So... 320 00:25:23.380 --> 00:25:24.140 Yeah. 321 00:25:24.140 --> 00:25:33.830 Well, you know, the interesting thing is a little sidebar, but it's like you talked about knitting, and then like one of the ways that you can spot a hand knitted sweater is there are tiny imperfections, and there's this sort of weird, 322 00:25:33.830 --> 00:25:34.490 So, you know, I mean, like... 323 00:25:34.490 --> 00:25:34.600 Yeah. 324 00:25:34.600 --> 00:25:51.780 I don't know, almost uncanny valley type thing where it's like, as a, like, you don't want it to be absolutely perfect. You never would be really able to make it absolutely perfect like a machine would, but you want to like limit the imperfections, but you still want the little thing of like, you knit that yourself, didn't you? It's like, yeah, I did. It's like, oh, that looks really good, you know? 325 00:25:51.780 --> 00:25:52.240 I'm... 326 00:25:52.240 --> 00:26:04.220 Yeah, two sort of observations. I was like watching my son reply to an email about college admission-y type stuff. And like he made a typo and he said, oh, I'm going to leave that in because then they know it's not AI. 327 00:26:04.220 --> 00:26:04.970 And I'm like, oh, my God. I mean, you're not wrong. 328 00:26:04.970 --> 00:26:06.040 You're not wrong. Yeah. 329 00:26:06.040 --> 00:26:06.310 Honestly. 330 00:26:06.310 --> 00:26:06.580 Yeah. 331 00:26:06.580 --> 00:26:18.720 And then, so secondly, you know, like when they built, these master craftsmen of old would build these great temples to their gods, they would always make one deliberate mistake somewhere so they could show that they... 332 00:26:18.720 --> 00:26:25.780 they weren't perfect, unlike the god they would want to, like, be. And I'm like that... But the fact that you chose to not be perfect when you could have been kind of undermines the whole thing. 333 00:26:25.780 --> 00:26:25.780 Oh yeah. Yes. Yes. 334 00:26:25.780 --> 00:26:28.160 But that's a whole other debate. 335 00:26:28.160 --> 00:26:33.770 Well, you know, that was the backwards justification for why they made the mistake, right? Like, that's... They didn't do that on purpose. 336 00:26:33.770 --> 00:26:34.180 Yeah. Right. 337 00:26:34.180 --> 00:26:36.400 They're like, oh, well, no, of course, we put that in there. 338 00:26:36.400 --> 00:26:37.620 because 339 00:26:37.620 --> 00:26:40.700 Crap, we've put the bottom... block of the pyramids is upside down. 340 00:26:40.700 --> 00:26:41.000 Right. 341 00:26:41.000 --> 00:26:41.700 We meant to do it. 342 00:26:41.700 --> 00:26:44.480 Yes, exactly right. Exactly right. 343 00:26:44.480 --> 00:26:50.560 Otherwise, you know, the god would be annoyed at us having done it perfectly. Right. Okay, invoice is in the post. 344 00:26:50.560 --> 00:27:04.320 Yeah. Right. Right. Well, okay, so... This, you know, the first part of this episode is a little bit of a downer in the sense of like, oh, I don't have fun at my job anymore. But there are actually some really good things about this stuff. 345 00:27:04.320 --> 00:27:09.680 And the part of the struggle for me, and part of the reason why I think this is an interesting question, 346 00:27:09.680 --> 00:27:30.660 is because being able to use these tools does... unlock a lot of really cool things. And I don't want to give those things up either, because the ability to like speak words into my phone and have working software come out the other side is kind of amazing. And I don't want to give that up. 347 00:27:30.660 --> 00:27:34.240 Right. 348 00:27:34.240 --> 00:27:46.520 It's... it's, you know, the problem is it doesn't take very long, right? If I could spend my whole day doing that, maybe I wouldn't hate it quite so much. But it's, this, a lot more of like, as you say, you wind up... If you want to actually do this all day long, it's a lot of context switching. 349 00:27:46.520 --> 00:27:49.620 It's a lot of like spinning multiple plates at the same time. 350 00:27:49.620 --> 00:27:49.720 It's... 351 00:27:49.720 --> 00:27:54.460 It's not focusing on a single problem and seeing it from beginning to end, which is what I want. 352 00:27:54.460 --> 00:27:54.940 exhausting. 353 00:27:54.940 --> 00:27:57.040 It's exhausting in a different way that focused time 354 00:27:57.040 --> 00:27:57.040 Yeah. 355 00:27:57.040 --> 00:28:12.090 is, for two reasons. I think one is the massive amount of context switching that you're doing in order to sort of feel like, while you're waiting for one thing to reply and it's off grepping the code base looking for whatever, you get another one up and you make a new work tree and you're doing something different or whatever. 356 00:28:12.090 --> 00:28:13.580 Yeah. 357 00:28:13.580 --> 00:28:21.940 The... But that also, the way that you're programming is by... engaging the social part of your brain, which, you know, for many of us 358 00:28:21.940 --> 00:28:21.940 Yeah. 359 00:28:21.940 --> 00:28:26.100 in this profession is... not as developed 360 00:28:26.100 --> 00:28:26.100 Taxing. Yeah, it's taxing. 361 00:28:26.100 --> 00:28:33.920 as maybe other people, or at least takes up energy that we keep in reserve for actual human 362 00:28:33.920 --> 00:28:33.920 Yeah. 363 00:28:33.920 --> 00:28:33.920 Right. 364 00:28:33.920 --> 00:28:39.980 con... And, you know, there are outliers amongst us. But... 365 00:28:39.980 --> 00:28:40.280 Right. 366 00:28:40.280 --> 00:28:53.980 For the majority of people, I can't help but think that like, if you've been arguing for 25 minutes with an AI on and off, that has drained a battery that doesn't replenish quite as quickly as the, I sat down and solved the logic puzzle for 25 minutes. 367 00:28:53.980 --> 00:29:05.980 Right. Right, right. Yeah. The whole, like, you know, I have no basis in sort of like psychology or anything to say whether the whole, like, introvert extrovert model of like, you know, 368 00:29:05.980 --> 00:29:17.040 introverts, you know, spend their energy on social interactions and extroverts recharge their body on, or their battery on, social interactions. I have no idea if that's like a real thing. 369 00:29:17.040 --> 00:29:21.150 But definitely seems appropriate for me and a lot of the other people that I know. 370 00:29:21.150 --> 00:29:21.560 I... 371 00:29:21.560 --> 00:29:26.720 Where it's like, it's not that I don't enjoy social interaction. It's just that it's taxing. Right. 372 00:29:26.720 --> 00:29:27.860 Yeah. 373 00:29:27.860 --> 00:29:38.900 And I can only do it for so long before I just need to go sit and like, you know, again, listen to Daft Punk, write some code, or play Factorio, as I've been doing a lot more recently, right? 374 00:29:38.900 --> 00:29:39.340 So. 375 00:29:39.340 --> 00:29:44.400 Oh, you're that Tron soundtrack. Oh, you see, I mean, that's a great one too, but like the Nine Inch Nails one is 376 00:29:44.400 --> 00:29:44.400 Yes. 377 00:29:44.400 --> 00:29:46.150 pretty darn awesome as well. 378 00:29:46.150 --> 00:29:52.920 You know, I think I've only listened to that like a couple of times, like the new Tron movie. 379 00:29:52.920 --> 00:29:53.640 Oh, shame. I haven't actually watched the movie. 380 00:29:53.640 --> 00:29:53.880 Yes. 381 00:29:53.880 --> 00:29:54.650 People have told me it's... 382 00:29:54.650 --> 00:29:54.940 I am. 383 00:29:54.940 --> 00:30:03.120 not great, but I would watch it just to listen to the soundtrack. There's a number of... The number of movies that I've got like that in my head, I'm like, yeah, the soundtrack is what makes this movie great. 384 00:30:03.120 --> 00:30:03.720 Not great. Yeah. 385 00:30:03.720 --> 00:30:04.120 Yeah. Yeah. 386 00:30:04.120 --> 00:30:04.520 All right. 387 00:30:04.520 --> 00:30:29.020 But yeah, there are a lot of things that make... using these tools to build things fun. But I think a lot of it just comes from the building, right? It's like, in the same way that like, you know, putting together a little Arduino circuit is fun, or like cooking even is fun. It's like, you know, I get to make something that people enjoy and I enjoy, and it's cool and I like that. Right? 388 00:30:29.020 --> 00:30:29.260 So... 389 00:30:29.260 --> 00:30:48.420 Let me take that, because that was definitely something that I was... going to... bring up is that, you know, and we've mentioned this before, there are people who like the craft of building something and they enjoy the edifice that they've made, not the outcome. And there are people who just want the outcome and they're excited about that. I think everyone's a... there's a mixed, you have a mixture of that. And I think probably where it's for me is like, I've got a thing. 390 00:30:48.420 --> 00:30:48.490 Right. 391 00:30:48.490 --> 00:30:48.640 Yeah. Yeah. 392 00:30:48.640 --> 00:30:55.140 I can make a thing. I can almost call a thing into being now by just naming it, almost, which is like an amazing 393 00:30:55.140 --> 00:30:55.140 Okay. 394 00:30:55.140 --> 00:31:06.140 magician ability that we've all gained, or for those who can afford the tokens can gain. And that's great. But then I didn't make the thing. 395 00:31:06.140 --> 00:31:06.380 Right. 396 00:31:06.380 --> 00:31:10.500 And so anything that I still, this comes back to this fraudulent feeling, is that 397 00:31:10.500 --> 00:31:10.500 Yeah. 398 00:31:10.500 --> 00:31:16.120 I don't feel like that is worth as much, right? 399 00:31:16.120 --> 00:31:16.520 Yeah, yeah. 400 00:31:16.520 --> 00:31:27.000 And I mean, if it ticks boxes for me, like if it's, I think we talked about this before, if it's a scraper of YouTube that goes and finds unanswered comments on other people's videos that include me, 401 00:31:27.000 --> 00:31:36.420