AI burnout: the hardest parts of your job all day

[00:00] Eric: [upbeat music] Okay. [00:10] John: What a way to start. [claps] [00:13] Eric: Welcome back to the Token Intelligence Show. Today we're gonna talk about AI burnout, which is a fascinating subject. The reason this is on-... my mind is that, uh, you and I both have a friend who has been going back and forth from the East Coast to Silicon Valley for well over a decade. Um, entrepreneur, investor, ran an accelerator program, you know, sort of-... you know, has been around the block. And he poked his head into my office the other day. I haven't told you this story-... so this is all new for you. I have a couple new things for you. I have this story-... and then I have a great quote for you. [00:21] John: Yeah [00:31] John: Mm-hmm. [00:36] John: Yep [00:45] John: No Yeah. I'm excited. Awesome Great. [00:51] Eric: And, you know, I go out to San Francisco for Vercel a good bit and, uh, but I'm usually just sort of working on Vercel stuff. And he went out and, for, uh, for a conference with his daughter, and he took the opportunity to c- The gaming one. Yes Oh yeah, we talked about this on-... on the, on a-... previous episode. Yeah Yeah. Uh, escape your AI girlfriends. [00:59] John: Right. [01:05] John: Oh, the gaming. Which, which-... previous episode. Mm-hmm Yep, the companion... companion thing. Yeah. [laughs] Mm-hmm. [01:16] Eric: Yeah. He took the opportunity to connect with a bunch of people that he and I had connected with over the years, uh, in San Francisco and, and sort of the greater Valley area. And I said, "Well, how was it?" You know, "How are these people?" Um, and the very first thing he said was, "AI burnout is real." Which was really interesting to me. That that was his immediate gut response to, you know, how are all these people doing, was AI burnout. Does that surprise you? [01:29] John: Nice. [01:40] John: Okay. [01:43] John: Yeah. [01:51] John: No. Yeah, it really doesn't. And what... Did you, did you drill in on, like, what do you mean by that? 'Cause we've had... 'Cause burnout, you know, has had lots of flavors over the last-... 10 or 15 years ago [02:03] Eric: Sure Burnout is definitely not, burnout is definitely not new. Um- [02:06] John: Not new, but it, but it takes these flavors. It takes, like, the, I don't know, like, social media era flavor-... of, like, the distraction of-... social... Like that, or it takes the... You've just worked with startups a long time. Like, there's a part-... you know what I mean? Like, there's all these flavors. Did the... Is the flavor new? [02:13] Eric: Mm-hmm Right [02:21] Eric: Sure [02:25] Eric: I would say there are two... I did drill in, and there were two categories or two flavors, if you will. One was aggregate and one was personal, which I found really interesting. So the personal side was that, v- you know, in order to thrive in this age of, you know, sort of ultimate AI efficiency, if you will-... the amount of work demanded from a talented individual has increased. Or at least the expectations have increased-... and if not explicitly, certainly implicitly. And one of the really interesting things that he said about this was, you know, not only is there this sort of baseline... Basically the, the, the low water mark is, is higher now-... because you have AI as an accelerant, right? And so, okay, if you're talented, we expect, you know, there to be a significant amount of output due to a- adding AI to the equation, right? Which is interesting, but he also said a lot of the people he talked to, uh, struggled with the fact that in order to figure out how to use the tools as they're changing, that adds even more-... workload-... because the models are changing, the methodology is changing. We've talked a lot about the last couple episodes, you know, markdown files-... uh, and sort of, you know, running grep over markdown files-... as the oldest convention in computing actually turns out to be the best way to accomplish certain things with AI. Right? And so- And Bash, right. Exactly. Um, and so... Well, that's interesting, right? Like, if before I was uploading all of this stuff into the contacts window, and now I need to do it as markdown files, you know, your workflow changes, right? And there's a t- there's a ton of thrash there. And so on the personal level, it was sort of, okay, it seems like there's higher expectations, and the workflow keeps changing in order to, you know, maintain sort of the latest-... tool set-... or methodologies, right? So that was on the personal side. The aggregate flavor was the amount of disruption it's causing in tech-... which isn't a surprise to anyone, you know, s- anyone watching closely because, you know, valuations of companies that were considered unicorns, you know, maybe a generational company, have plummeted due to-... you know, uh, due to AI, right? And, and the impact of that. Whether that's, you know, the technology is less relevant anymore, whether that's people are building instead of buying. I mean, there are all sorts of reasons for that, right? But a lot of companies that people worked at, loved, you know, believed in, um, the ground is shifting. And so that was... You know, those are sort of the two flavors. [02:34] John: Okay. [02:38] John: Oh, okay. [02:52] John: Yeah [03:00] John: Yes. [03:03] John: Right [03:22] John: Yeah [03:26] John: Right. [03:35] John: Yeah. [03:50] John: Hmm Yeah [04:01] John: Right [04:05] John: Right [04:11] John: Right. And Bash. [laughs] Yeah. [04:27] John: Yeah. [04:37] John: Yeah Right [04:48] John: Mm [05:03] John: Right [05:22] John: Yeah. [05:26] John: Interesting. So I've got a hot take on this, and I, I really can't remember who told me this, which is... You might be surprised when I tell you how, like, aggressive- [05:35] Eric: Just looking granola [05:36] John: ... I, I probably should- How aggressive it is. But one of the takes that I've heard in the last two weeks, I cannot remember who told me, it was a non-technical person, was essentially technical people have had too much control for too long, and then, like-... happy they're being disrupted, is essentially the take. [05:37] Eric: [laughs] [05:49] Eric: [laughs] [05:54] Eric: Okay. [laughs] [05:55] John: Like, their, like, their wages have been inflated. They're, like, you know, just, like, a whole thing. I was like, gotcha. Right Right. Yeah. [06:00] Eric: Yeah. Come back down to reality-... people. Tech people. How... Do you agree with that? [06:08] John: No, I don't think so. Um, it depends on the company. There are some companies that, that their tech teams have the company hostage for whatever reason. But that's a people problem. That, that's a... You know, there's just companies that end up with tech teams that hoard information and knowledge and build an empire, and then everybody hates them, but, like, you ki- have to keep them around because, like, you can't... It's too hard to get rid of them, basically. So that's a problem that's always been a problem. Um, and so I understand that reaction, but the, uh, those are pretty broad strokes. [laughs] [06:18] Eric: Sure. Yep. [06:35] Eric: Yeah. Yeah, yeah, yeah. [06:45] Eric: Those are very broad strokes. Yeah. I, I'm tempted to get into the basic economics of, you know, supply and demand on skilled labor, but-... that's probably another episode. Uh, [06:47] John: Oh. [06:53] John: Yeah [06:58] Eric: personally, have you experienced AI burnout? You're running a company. You are probably one of the most advanced people I know in terms of testing out use of AI in almost every context. So running it, you know, actually, you know, having agents interpret meeting notes-... create issues, write code, you know, open pull requests, and running this entire cycle of essentially having agents as digital employees, uh, to reference-... a previous episode. [laughs] Um, and so you're very, very deep in this, and in theory, if I describe a workforce of, you know, digital robots that are doing these things, acting as employees, it would theoretically create a lot more margin for you, but I get to see you every day in the office-... and we have to, you know-... find time to record the show. And it doesn't seem like you have, you know, immense amounts of margin that didn't exist before. [07:01] John: Yeah. Right. [07:11] John: Right. [07:17] John: Mm-hmm [07:28] John: Yeah Sure. [07:49] John: [laughs] Right Yeah, yeah. Right. [07:58] John: Yeah. And I, and I think... I mean, some of that's, some of that's the current state of reinvestment-... of, like, if we find some time to be more efficient here, like, what's the use of that time? And, like, for now, I think a lot of companies are in this space. We reinvest it into learning and getting better at the thing. Right? Um, and I think we almost, we almost talked about the, the generalist skill set today. But we kind of pivoted. But I think it's still relevant-... here. And the, the thing that I think is relevant is in thinking through the, let's call it productivity, like generically. Um, I... Like, it's this yes and no of, like, yes, more productive, like coding is faster, like there's some, like, back office or, like, project management or stuff-... that, like, can be faster. But no, in that, like, there's this, like, there's some skill set alignment problems of, of like, ugh, like I'm pushing, like, technical people to, like, kind of be less technical and, like, do some of their own PM work with AI-... or I'm pushing, like, non-technical people or less technical people to do some of their own dev work with... [chuckles] Like, that's hard. Um, and then the analogy I could think of for this was imagine you took, um, a collegiate athlete that is a soccer player, let's say. And then said, like, "Hey, like, we're gonna take you and put you on the football team, and you can be the kicker." So that's, like, a common... That's a fairly common-... thing. But it's not a given. Like, not every soccer player is a good football-... kicker. Um, and I don't- [08:06] Eric: Mm [08:11] Eric: Mm-hmm. [08:17] Eric: Yep. Yep. [08:24] Eric: Mm. We did pivot, yeah. Mm [08:35] Eric: Mm-hmm. [08:46] Eric: Mm-hmm, mm-hmm [09:03] Eric: Mm [09:10] Eric: Mm-hmm. [09:25] Eric: Okay. [09:34] Eric: Sure Yep. Right. Sure Yep. Well, the entire culture of the sport is very different as well-... beyond just the skill set of-... of kicking a ball. [09:45] John: Right Right Yeah. And that's more obvious to, like, a normal person that those are different sports. But if you zoom out, and you're at this more macro level of, like, I don't know, it's, like, a tech person. I don't know what they do. Um, then you have this, like, s- weird skill set thing where, like, "No, I'm, like, a soccer player, and I kick the ball." And they're like, "Well, just play football and use AI and, like, kick the ball in football. Like, it's a ball. Just kick it." You know? Um... [laughs] [09:59] Eric: Mm. [10:12] Eric: Mm. [laughs] Such a good analogy. [10:17] John: And, and it is... And, and, and then the conversations are tough. Well, it is a ball. They're both balls. They're different shapes. Like, the rules are different. And then, and then you have, like, context awareness issues at playing the new sport-... that are specific. 'Cause is it... Like, say the person's, like, an athlete, and they're, like, a decent athlete. Okay. Great. They can flex a little. Like, they're familiar with, with, you know, kicking a ball into a net versus feel like there's some transla- But then you get into the more details, and then you have these, like, misses on, like, smaller things that people are-... like, just feel like, "Come on." [chuckles] So. So that's tough. [10:25] Eric: Mm-hmm. [10:31] Eric: Mm Yeah. [10:38] Eric: Mm-hmm. [10:46] Eric: Right, right. [10:52] Eric: Mm Yeah, yeah. Yeah. Interesting. Do you feel like your team is at higher risk of burnout than you? Because as a business owner, you're trying to figure out how your business can utilize these tools and thrive in this new age. Um, how you can run as efficiently as possible. [11:09] John: Right. [11:16] Eric: But... And so for you, in many ways, it's, okay, well, how am I reinventing my business to operate-... you know, and thrive in this new world? But for your team and the people who work for you, do you feel like they're at higher risk of burnout than you? Because it's a very different problem than trying to figure out how to operate a business in a new environment. [11:21] John: Right [11:33] John: Theoretically, yes, but I think practically, no. The theoretically yes is that, like... [11:35] Eric: Hmm. [11:41] John: I mean, there's two mindsets here. There's one of, like, a manager leader that... And we were talking about this before the show. This is great. So you've got, like, an executive level at a business that's being told by the board, like, "Hey, use AI." Like, make things more efficient, faster, increase our margins. Okay, great. We'll use AI. And then the mandate down is, like, use more tokens, whatever. Um, the... And then, and then some, you know, some of the executives are kind of playing around just to get a sense for the tools.Then there's like this middle management, like depending on your business, like VPs or directors or maybe they're just managers, who are kind of being told like, "All right. You gotta make all your team use AI. You have to figure out how to use AI." And then some of these roles are these like weird like player/coach roles at the same time where like, "We want you doing work too because like you can use AI." I think that position is like the worst squeeze-... right now. And I fall into both a bit, like I... But I, I've used like AI personally enough where I feel like I can allocate work given the new AI era, and I actually know like how long-... things are gonna take. Whereas if I was allocating work like previously, and then with AI, like it would just be wrong, right? It would just be not enough. But then if I was just like not really in... Like knowing how things worked in AI, and just being like, "I don't know, like use AI. It'll take you like five seconds," then I think there would be a lot of burnout, [laughs] people. Um- So, so I do think since I've spent a ton of time like digging into the piece of the technology and understanding like reps with like how... Like what are the edges-... and like how long things take, that, um... Yeah. But you'd probably actually, actually have to ask my team. [laughs] It's a good question. Yeah, we should. Um. Yeah, live It's gonna be live on air. [laughs] Um, but, but I think the most interesting part with it is, [11:55] Eric: Yep. Yep. [12:04] Eric: Mm-hmm. [12:32] Eric: Right. Right. [12:37] Eric: Mm [12:55] Eric: Yeah Yep. [13:05] Eric: Mm. [13:15] Eric: Interesting. Yeah, yeah, totally. [13:25] Eric: Mm [13:32] Eric: [laughs] We'll have him as a guest then. Live. [laughs] Your performance-... review is gonna be on air. [laughs] [13:48] John: that leads it to burnout, is context switching, is what I would say. 'Cause I think that's- [13:53] Eric: Mm. Okay, say more. Gi- give me a practical example because I am about to heartily agree with you- [14:01] John: [laughs] Depending on what I say [14:02] Eric: ... [laughs] depending on what you say. [14:04] John: I mean, we've known for a long time. So I think of back to Scrum-era IT. Agile Scrum, like 15 years ago, um- I don't know. People... Yes and no. I don't- [14:11] Eric: Yes. [14:15] Eric: Is that over? Okay. We'll, we'll go, we'll go scroll around LinkedIn to see if it's... [laughs] [14:21] John: I don't think anybody has a new, a new version of that. And I think people are already getting kind of burnout on it. I think, to be honest, mostly from misapplied-... versions of it. But that's what happens, right? Like if something hits, like, and then over time the original intention just gets... People get so far from it. And then, yeah. So, okay. Anyways, but like Scrum, like Scrum-era IT-... right? Um, you've, you've got somebody in charge of planning and, and then like they're in charge of like removing impediments and like-... blah, blah, blah. Like, I, I think that type of work, and, and it'll... I think it's just being done by managers, honestly, and directors-... and stuff now. Um, uh, uh, big enough teams still have those people, but thr- that work exploded of like how do you keep-... AI unblocked that can just move like really fast? Um, or how do you keep people using AI unblocked? Because it was already like fairly hard for-... a productive team keeping them unblocked. But yeah, uh, I think that got a lot harder and, and you don't typically have transportable skills of people between let's just grab this dev and like stick them on requirements now. Like that doesn't usually work well. It... Honestly, if you had to pick, you probably want the requirements person with an AI tool versus a dev doing requirements. Usually. [14:28] Eric: Yeah. [14:31] Eric: Yeah, yeah, yeah Yeah. Sure. [14:41] Eric: Yeah, yeah. [14:49] Eric: Mm-hmm [14:57] Eric: Mm-hmm [15:04] Eric: Mm-hmm Mm-hmm. [15:14] Eric: Mm [15:19] Eric: Yeah. [15:25] Eric: Sure Mm-hmm. [15:40] Eric: Yeah. Sure. Sure. [15:50] Eric: Right. Right. Right. Yeah, it is... It's really interesting. I, um, I agree with the context switching, and I think there are two, there are two components to that, that I've both experienced-... personally and that I've heard other people-... talk about. [16:08] John: Yeah Yeah Which, just to interrupt real quick, the context switching being because there's so much more requirements works, it's like switching back and forth between we're doing more projects in parallel, we're doing more issues and planning in parallel. Right. [16:22] Eric: Yes. So I, I definitely agree with more projects in parallel, right? So if you take, let's just say a person who is talented at their work, and they, you know, build a, they build a lot of competency in using AI to accelerate... We did an episode about this, right? Being good and fast. Right? AI is a significant accelerator, right? And so that person can produce more output because they are automating or dramatically compressing the time-... it took to do certain parts of-... what they were doing before, right? And so there's just a lot more that can be happening. And, and in reality, you know, i- in, in theory, you think about that as sequence, right? So like I knock out a project, I move to the next one, I knock out- But that's not how work happens in the real world. Right? It means that you get things done faster, and so that there are more projects happening that you're a part of, and those things tend to happen in parallel. Right? And that is very difficult because you can't have a perfect world where everything is sequenced and you do things faster-... but you knock them out in sequence. You know, and so, so there is a lot more, uh, context switching that has to happen. But the other thing that I've noticed that is really interesting, and this is sort of a, a specific flavor of this, but as you, as you learn to wield AI tools, you tend to, uh, develop skill in, um, asking AI to do more complex tasks. Right? And those tend to take longer, right? [16:43] John: Yes. Yeah. Right. [16:55] John: Yeah Right Right. [17:10] John: Yeah. No. Never. Yeah. [17:22] John: Right. Right. [17:32] John: Yeah Yeah. [17:40] John: Yeah. [18:05] John: Hm. Yep. Mm-hmm. [18:10] John: The-... the AI runs longer Yeah. Agree. Yeah Right. [18:11] Eric: The actual-... the actual job, yeah. The AI workload-... runs longer. Okay? So in, um-Net-net, the amount of time it takes to do the thing is far, far, uh, shorter than it was in the previous world. Um, but the amount of time that the agent runs the initial job, whatever it is, takes longer because you're asking it to do, uh, a more complex task or set of tasks upfront-... right? And, uh, and that could be for anything, right? It could be, you know, a task where you're-... asking it to do some sort of analysis. In my world, it could be, you know, doing research and scaffolding out-... you know, a report that's going to inform or be part of some piece of-... content or, you know, whatever. Scanning through transcripts, like, whatever that is, right? And [18:28] John: Mm-hmm. [18:41] John: Right Right. [18:47] John: Right [18:53] John: Sure [18:56] John: Right Right. [19:04] Eric: before, you were doing all of that work yourself-... and so that time was occupied with, with that sort of manual labor, let's call it-... as part of the, as part of the cognitive, you know, the overall cognitive task. [19:07] John: Right [19:12] John: Mm [19:18] Eric: But now you ask the agent to do that, and you have to wait. And so there's this-... weird-... context switching where, you know, there's this, these moments of downtime that were possible before, 'cause you could-... get distracted and-... you know, all that sort of stuff. But now it's literal downtime where you're waiting on this job to complete, right? So what do you do with that time? Well, you start something else, but let's say the job takes, like, six minutes. [19:22] John: Yeah Right [19:33] John: Right Yeah [19:40] John: Right. [19:46] John: Well, and, and the cycles are indeterminate. Like, I would say they can b- be between, like, a minute or two or maybe, like, 30 minutes. And it's not completely clear upfront what your time window's gonna be. And that's tough. [19:49] Eric: Yes. [19:54] Eric: Yes, exactly. [19:59] Eric: Yep. Yeah. And so I think that that type of context switching is difficult because it requires developing a new skill set, right? How do I plan my day when it's difficult to... 'Cause before it was like, "Okay, I'm gonna have to do this research. Okay, I'm just gonna block off the first half of the day." Right? Okay. Well, now I'm in a world where that, you know, four-hour period, let's say, the work can be compressed into, let's say, a half an hour, right? Depending on-... what the task is, right? And so, but y- how, how is that half an hour distributed over a four-hour period? 'Cause it's not like you can just... It's not like it happens in one discrete-... 30-minute period, right? And so I think that's the other challenge of context switching, which I think can contribute to burnout, is how do I actually manage a... How do I plan out my workday when I have these indeterminate, um, these jobs that don't have a discrete, you know, time allocated to them-... where there's downtime in between which creates a context for context switching. It's just interesting. It's a, it's a totally different way of working. And it's, and it's not really intuitive, actually, um-... when you first start doing it. [20:08] John: Right. [20:18] John: Yeah. Right. [20:30] John: Sure. Yeah [20:44] John: Right Right. [21:03] John: Right [21:10] John: Right. [21:14] John: Right. [21:18] John: No It's not. And, and it also feels like a phase. So-... so I think as, like, we keep progressing through these different stages of AI adoption, it's hard to wanna, like, over-index in changing your life on what, like, is probably just a phase. [21:24] Eric: Mm [21:36] Eric: Yes. Totally. [21:38] John: Because I think everybody's pretty confident the task horizon will just keep expanding-... where the things can run for longer than 30 minutes. Um, its ability for ambiguity, it will get better at asking you questions to clarify-... for example. Um, and, and then there's obviously the, all the experimentation with having it wake up on a loop to, like-... figure out what to do with, like, very broad goals. Um, and whatever level of that we get to is a whole 'nother thing. But even if that is distant future, which really might not be. The question is... And, and that, and that's a fundamental... We've talked about this a lot. That is a fundamental difference when you're, like, delegating responsibility of, like, "Hey, AI employee, you're in charge of marketing. Here are your objectives. Here are your-" "... KPIs. Here are your," whatever framework you use. OKRs. Like, go do it, you know? That's a r- and we're not there yet, but that is a really different framework than, like, "We're working on a project. I need you to complete this particular task. Here's the expected outcome from the task." Like-... you know? So. [21:44] Eric: Yep [21:54] Eric: Mm-hmm [22:04] Eric: Yep [22:07] Eric: Yep. [22:16] Eric: Yep. [22:31] Eric: Mm-hmm. [22:36] Eric: Yeah. Yeah. Yeah. [22:48] Eric: Yep Yep. I, I think that the, um, there's... We meet with a group on Friday mornings-... just to talk about AI. And it's a really interesting, diverse group of people, you know, who, who join the discussion. And one of the guys mentioned, this is weeks ago, um, but he's a software engineer. He's been a software engineer for a long time and a very experienced, seasoned-... full stack software engineer. And he mentioned... You know, we started talking about output and what do you think your actual productivity-... increase is, and-... you know, he said in some areas it's probably 10X, in other areas it's n- and he said, you know, average, it's probably at this rate, you know, 2X and-... increasing in terms of-... the amount of productivity gained. Which is-... which is pretty insane. [laughs] It's definitely astounding. Right. Um, but with... And, and, you know, and then we started talking about, okay, well, what, what's made you more productive? And it's like, you know, I'm not, I'm just not writing as much code anymore. Uh, but then he just, a- almost offhand, he said, "You know, but the big difference is that I'm doing the hardest part of my job-"... "all day long." Whereas previously-... I would, you know, write a bunch of code, um, you know, write a bunch of tests-And then, you know, review it and, you know, sort of go-... through the QA process and whatever. And now he's like, "The... All I do is, is just the most difficult part of the job-"... you know, making sure [laughs] things go well. [22:58] John: Yeah Mm-hmm. [23:12] John: Mm-hmm. [23:16] John: Yeah Yeah. [23:24] John: Right Right [23:36] John: Mm-hmm Right Right. Yeah Astounding. Yeah. Right. [23:54] John: Right. Right. [24:04] John: Yeah Right. Mm-hmm [24:16] John: Yeah [24:23] John: Right I think... And I've heard two versions of that, like, I just do the worst part of my job all day long, or I do the hardest part of my job all day long. Which are different, and I think he likes doing the design stuff. But there are some people that just really enjoy implementation, and they feel like the part of the burnout has to be like, "I do the worst part of my job all day long." Like- [24:30] Eric: Yeah. [24:33] Eric: Sure. [24:41] Eric: Right. Right. You know, uh, but I think that there's also this interesting, interesting question around how... I- if you take someone like the software engineer who is very, who is objectively talented-... and, you know, there are... I, I, I agree with you. It seems like, okay, well, if I had to choose to do more of the architectural thinking than I did before-... I would probably choose more of that, right? But would I choose that to be 95%-... you know, is... Like, I don't know. And at which point are the gains logarithmic? Like, do they actually plane off-... at some point? Because, you know, it's like an engine, right? You can't run it at red line in perpetuity-... without there being- [laughs] Without there being damage, right? Um, or at a minimum, you know, or at a minimum hitting the limit of, you know, the amount of output-... that you can get. [24:56] John: Yeah [25:09] John: Right Yeah. [25:15] John: Right [25:18] John: Right. [25:24] John: Right [25:32] John: Yeah Being damaged. [laughs] Right. [25:42] John: Yeah Yeah. Yeah. Okay, I have some thoughts on your macro point. So that's, like, kind of the individual level. Like, I think we really dug into that. On the, on the macro point, for our world, for SaaS-... like, it's... Pretend like this wasn't AI and it was some other thing. There's just enormous competitive pressure that happens to be coming from AI, that happens to be-... a productivity t- tool for devs-... let's call it, very simplistically. [25:47] Eric: Okay. Yep. [25:56] Eric: Mm-hmm [26:01] Eric: Mm-hmm. [26:07] Eric: Mm Mm-hmm [laughs] Right. [26:12] John: So there's just enorm- enormous, like, competitive pressure. So I, I think it's easy to not think about that and to just talk about, uh, what we spent the majority of the episode on, which is important, but it's just a practical, like, really competitive time. Because there's the real- there's a high perception and a reality of needing less talent to accomplish the same, or if you wanna keep the same talent, to expect to 3x for, like, whatever the x is-... 10x more out of the same people. Um, and the, and the, like, consolidation in the market and the stocks of various things-... you know, from crashing on a regular basis. And, you know, just, just-... you know. [26:15] Eric: Mm-hmm. [26:27] Eric: Yeah. [26:41] Eric: Yep [26:49] Eric: Yep [laughs] Yeah Yeah, yeah, totally. It's a lot. I mean, I do think that the... There is a, um, there is this subtext of, of imagining what it's going to be like to remain competitive in the future. And there being a lot of unknowns as part of that equation. Right? What is software engineering going to look like? What are salaries going to be? You know. Am I learning the tools fast enough? Is it going to be a situation where two people can do the job of 10, and w- how can you ensure that you're gonna be one of the two? You know. And I think one of the other interesting things is that has hit software, that has hit the world of software very, very early, but it's starting to happen in other industries as well. You know, and the widespread nature of that's pretty interesting-... uh, which I think contributes to the burnout because people are, you know... It's a time of great unknown in terms of-... how you stay competitive. [27:12] John: Right. [27:15] John: Right. [27:22] John: Sure. [27:34] John: Yeah, right. [27:45] John: Right. Right [27:53] John: Right Yeah. I'm ready for it to move on. I'm ready for it- [27:58] Eric: [laughs] For it to become normal [28:00] John: ... from, from software. Like- 'Cause, 'cause all of the pro- all the product... I mean, think about this. All the productization from the frontier lo- labs in the last couple years has been really heavy on software. Cloud code, c- um, codecs, uh-... whatever. Um, Cloud code being the, the biggest one, but they've heavily productized and, like, honed in on that use case. I'm sure-... millions, hundreds of millions of dollars have gone into RL training focused on coding. Like, I hope, I kinda, I kinda hope we just get to a point of diminishing return on that-... and, like, somebody else can take a turn-... with the focus of hundreds of-... of millions of dollars on something else. It would be great. Um, I mean, part of me, like, it, like, would be fun to see the, obviously, the coding models continue to improve, but the other part of me is like, "Let's try something else, guys," you know? We've, we've spent a couple years here. Can you slow it down just, just for a bit so everyone can catch up? [28:01] Eric: Oh. [laughs] Yeah. [28:09] Eric: Mm-hmm. [28:14] Eric: Yep Mm-hmm. [28:21] Eric: Yep [28:26] Eric: Mm-hmm. [28:32] Eric: [laughs] Right Hand off the torch Yeah. Yeah, yeah, yeah. [28:49] Eric: Yeah. [laughs] [28:54] Eric: Yeah, sure. Sure, yeah. Yeah, I mean, it is really interesting. I think, you know, early in our careers, I think that, you know, we tried to establish ourselves, um, you know, with business acumen, with technical acumen-... and, you know, sort of establish a career path that seemed very sure and very secure, right? And so much is changing now. And so it's just been a long time, and I don't say this in a, in a prideful way at all-... you know, but once you sort of knock out the first half-decade of your career and you have, you know, established-... you know, references and all that sort of things a- and, and all of those components, you feel a sense of, like... At, at a minimum, you feel a sense of confidence of like, okay, I, I can go out-... and figure out how-... to try to create value and-... you know, I have some track record, right? And it is weird. Like, I, you know, sometimes I go home and it's, you know, my kids ask, you know, "How, how was your day? What, what did-"... "what did you do? What did you work on?" And it's hard to explain to them, you know, we're... Like, I'm, I'm living through a time where me and my peers are watching our very jobs, like, change before our eyes-... and we're having to figure out how to adapt, and it's pretty crazy. [laughs] [29:09] John: Mm-hmm [29:15] John: Yeah. Right. [29:24] John: Right [29:29] John: Right [29:40] John: Right Right Exactly Yeah. Right. [29:53] John: Right [30:07] John: Right [30:12] John: Yeah. We're, we're like, we're like farmers in the field sitting in chairs watching the machines-... like, harvest the crops. [30:17] Eric: [laughs] [30:22] Eric: [laughs] I would like to think that I'm, I'm s- I'm- [30:25] John: With, with, with tiny little remotes in our hands-... that operate the machine that harvests the crops. And we used to spend hours in the sun-... you know, harvesting the crops. I mean, it's weird. [30:28] Eric: [laughs] [30:33] Eric: That's right That's right. Yeah. It is. Yeah, it is, it is strange. I do think... So I, I have a quote for you by the way. Yeah. So this is from the, um, one of the founders of Whisper. [30:41] John: Oh, yes. Good. [30:48] John: The, uh, text to speech thing. Wait, speech to text. [30:50] Eric: Yes. Speech to text. Yes. [laughs] Gosh. Speech to text. Okay, this is really interesting. Wonderful article. We'll, we'll put it in the show notes. Uh, I texted it to you this morning. Yeah. This is the one. Um, I can't believe you hadn't read it yet. It's only like-... it's only like a 30-minute essay. [30:53] John: Yeah, that. [laughs] All right. [31:02] John: Oh, okay. Yeah, yeah. [31:09] John: I haven't read it yet [laughs] Can't wait. [31:13] Eric: Just dump it into Claude and ask for a summary. [laughs] [31:17] John: [laughs] Somebody told me that on a call to... Somebody literally told me that on a call yesterday, and they were like, "I'm gonna send you this long doc." And they're like, "Don't read it, just dump it into AI." [laughs] [31:25] Eric: It's like, then why are you sending it? Like, you can dump that into AI. Okay. This is really interesting. So he talks about the economic impact, what he believes to be the economic impact. Um-... you know, and so he is a firm believer that in many, in many knowledge worker jobs, you know, um, two people will do the work of 20, you know? Um, [31:28] John: Yeah. All right. [31:37] John: Okay [31:50] John: Okay. [31:52] Eric: you know, which I think that, you know, it's yet to be seen if, if the impact will be systemic at that level. But I think you and I both see that, you know, it's-... the, the, the economics are going to be very different-... and things can get more efficient. But I think he brings up this really wonderful point. I feel like this article is probably gonna come up multiple times on future episodes. [31:58] John: Sure. Yeah. It's possible Sure [32:12] Eric: But, uh, I'll just read, I'll just read here. So, um, Sahaj Garg is his name, uh, co-founder and CTO at Whisper, and quote, "The economic disruption is severe, but the identity crisis may be worse. We have a direct historical parallel to deindustrialization of the American Rust Belt. When manufacturing left cities like Detroit, Youngstown, and Flint, the economic damage was obvious, but the deeper wound was to identity." "These were communities where being a steelworker or autoworker wasn't just a job, it was who you were. It defined your family's place in the social order, your sense of contribution, your self-respect. As Case and Deaton wrote in 'Deaths of Despair', 'Jobs are not just the source of money, they are the basis for rituals, customs, and routines of working-class life.'" "Destroy work, and in the end, working-class life cannot survive. The loss of that identity, even more than the loss of income, drove the epidemic of depression, substance abuse, and what they termed the deaths of despair, [32:41] John: Hmm. [33:01] John: Hmm. [33:16] Eric: uh, what they termed deaths of despair. The coming displacement of knowledge workers has the potential to be that dynamic at a much larger scale. A software engineer's identity is as tied to their cognitive ability as a steelworker's was to their trade. 'I'm smart, I solve hard problems, I build things,' is not just a job description, it's a self-concept. When AI can solve harder problems and build things faster, that self-concept shatters. By default, I suspect that mass unemployment of knowledge workers simultaneous with the disillusion of a core part of their identity will lead to a widespread depression and despair." So we'll just end the episode there on a high note for everyone. [laughs] So this is very interesting. I think that is one of the most insightful, [33:54] John: [laughs] [34:00] Eric: uh, pieces that I've read on the impact because-... most people focus on economics-... efficiency, um, but I think the idea- [34:04] John: Yeah Mm-hmm [34:10] John: Or just practical, like job... Yeah. Job numbers. [34:11] Eric: Exactly. Right? And don't get me started on this like, "Here's 10 tips on how to survive," you know-... you know, whatever. Oh my gosh. That's, it's-... it's really bad. What is interesting though, and I actually thought about emailing him, and I may still do this. Uh, [34:18] John: [laughs] That'll be our next episode. [laughs] All right [34:29] John: Oh, you should. [34:31] Eric: but I think that he misses the identity impact on the two people who replace the 20 because that's not free. [34:42] John: Okay. So what's it like? [34:43] Eric: And, and I think, and I think that that's, I think that's a big piece of the burnout. That's what I think about, right? Is that, is that y- you know, I've seen teams who previously would've been 10 be run by two people. Right? [34:56] John: Mm-hmm. [34:59] Eric: But that's not free for those two... It, it's not like the AI just completely removes 80% of it, right? It is... It, it makes the job of the two people actually way more difficult-... because of the context switching, because of, you know, the in- the expectations of output, right? Even just to maintain the-... you know, 10 people's worth of output, right? [35:06] John: Right. [35:11] John: Yeah, for sure [35:19] John: Well- And there's a... The, I like to think of it w- with like ownership, of like o- outcome ownership, and if you have... I don't think there's any AI yet where you're, where if you're using AI to do a thing, you're not at least like a co-owner on the thing. So if you're told to own 10 times more things, and you have like some people that can like... or people and/or AI. Maybe it's like one other person and then like AI. Like the ownership problem is still a problem. [35:38] Eric: Mm-hmm. [35:52] Eric: Yep. Yeah. T- totally. Totally. And so [35:53] John: 'Cause, yeah. [35:58] Eric: it, it's just interesting and, and the reason I thought about emailing him was saying like, okay, uh, yes, the... And I'm not, I mean, I, you know, I'm not trying to compare like different flavors of despair [laughs] here. [36:14] John: Yeah. That's right. [laughs] Yeah. [36:16] Eric: But, but I, I do think the identity problem and, and the despair from mass job displacement is, is a, you know, a real potential-... challenge like for our society, right? [36:25] John: Yeah Oh, for sure. Yeah. [36:28] Eric: But the people who keep the jobs, if they hate their job- That's also-... a major problem, right? Like, that's not good. Um- [36:32] John: Yeah. That's bad, yeah Yeah, yeah. Yeah. You have, you have two... You end up with... You can easily end up with two bad outcomes. Yeah. [36:42] Eric: Exactly. And I think that that is one thing that I have not seen-... written about very much-... is that there's... I, I think there's two things. Number one, I think that there is this sense of inevitability in a lot of these types of pieces. Now, and I would say, I don't agree with everything that he wrote in this piece. It's a marvelous read-... and I think he's very... I think he's thinking very well about a lot of these things. I don't agree with everything-... but there is a sense of inevitability in this piece and a lot of other pieces where it just sort of says, okay, there's gonna be... There's only gonna be a few people who, you know, sort of retain the jobs. And here are the negative consequences for everything that happens to everyone else, right? And you just sort of... It's like this inevitable thing-... but there's not, there's not any thought experiment or cycles, at least that I've seen-... going towards, like, well, if two people are doing 10 people's jobs, that is a [laughs] significant-... difference for those two people, right? And maybe they're think- If there is mass job displa- displacement, you know, which it's yet to be seen how severe that impact-... might be. Um, anyways, that... It just really got me thinking about-... well, the, you know... From what I've seen, there are a lot of people, and circling back to the, the beginning of the episode, [36:47] John: Yeah Agreed [36:58] John: Yeah. [37:01] John: Sure. Right. Yeah [37:08] John: Right. Right [37:20] John: Yep. [37:25] John: Yeah, yeah. Right. Yeah [37:35] John: Right [37:41] John: Right Yeah. [37:50] John: Right [37:54] John: Yeah [38:02] Eric: the people who are retaining those jobs currently don't seem to be raving about their experience. [laughs] Right? [38:08] John: No. Well, and I've seen this in micro, microcosms, like working, like years ago when I worked for a company owned by private equity where they would, um... You know, like the whole move is you're consolidating companies, and then you-... have synergies, and you get rid of people. Um, and the people that stayed there, like, long term, like, that, that made it through, like, quote, "all the cuts"-... and were synergized all the times-... they weren't happy. Like, they were-... like s- the most overworked people-... I'd ever met. Yeah. It's wild. [38:21] Eric: Mm-hmm Mm-hmm. [38:31] Eric: Mm-hmm Mm-hmm No Right Right. Yeah, yeah. By the way, points for using the word synergy-... twice in the same run. [38:43] John: Oh, yeah I have it on my, my bingo card. Yeah. [laughs] [38:47] Eric: It's definitely on your bingo card. Okay. I think that's a wrap. Let's go. Let's go tell the- Tell- [38:50] John: All right. Yeah, it is. Let's go tell the AI-... AI, "Continue." [laughs] [38:56] Eric: Tell the AI to continue. [laughs] We'll catch you on the next one. [39:03] John: Yeah. [outro music]