Episode 17 - AI FOMO Is Making Us Work More

Welcome back to another episode of the Not Brothers Podcast. Episode 17, we'll be talking about the AI efficiency paradox. It's this theory of why do we feel compelled to work so much harder and longer when AI is supposed to make our lives easier and faster and less disjointed? and somehow it seems to be the opposite of that. So, Ryan, what are your what are your thoughts on this? We were talking pre hitting the record button and we've both we both feel and have experienced this first hand and there's a lot of data that goes along with this this paradox as well.

It's an interesting one. you know, the whole idea of AI is like you can utilize AI to to do things more efficiently or or with s with less effort than you ordinarily would, or you can automate things or whatever, which is great. and we've certainly used it in a ton of places where that is the case. Like we can we can have AI do daily reports or knock out some piece of work that's like really easy to kind of bowl down the center lane. The the interesting conundrum with that is while It should make it more possible for you to spend less time. Like theoretically, you could utilize AI and spend less time working, right? Like that would be the easy sort of trade-off. And what what I've found out winds up happening is actually the exact opposite, is you wind up spending more time working and almost having this like FOMO moment when you don't have agents running, because it is possible for agents to run 24/7. So then you've sort of like are fabricating reasons for them to run or are almost like, you know, am I not working hard enough or am I missing out on something? Right. especially when you see like all these bullshit clickbait articles that people put out of how they have these, you know, 15 agents that run their whole company and do all these great things and When you really dig into it, you know, the only thing they're actually doing is creating things that need to be maintained and the person is actually just doing a whole shitload of maintenance and not actually accomplishing anything. But it raises the question for at least a moment of like, shit, what am I missing out on? And and why don't I have fifteen agents running twenty-four hours a day doing things that I don't even know what to put them to work on on. it it's it's a real challenge and it's an interesting one. The other one that I've faced that's maybe a little more tangible is the speed at which it can move is a lot faster. you know, I have like we have our repositories or our our Kanban boards of things for our projects and all the features and bugs and whatever. And it's possible to just turn an agent loose and go, you know, tackle every one of these, or turn a series of agents loose and tell them to tackle them all. And now they're all moving in parallel. And you have code that's stacking up and meeting reviewed and functions that you know might be finished or might not, they might be completely fucking wrong. but it can move faster now than I can. So like now at at this point, code review is the bottleneck. For me and a lot of our products, because I still believe in reviewing code and I still, you know, throw away a lot of AI generated code and wind up solving it a different way. But it is a difficult conundrum and paradox to to solve for. I think you and I were talking about kind of what spurred this earlier this week was you know I've started to get back in the gym and one of the ways that I convinced myself but that was okay because I'll wind up just sitting here working until forever. It's not super healthy. is I was like, All right, well I'm gonna stack up these things I need to do. I'm gonna kick off three or four AI agents and then I'm gonna go to the gym. And then they sat and ran in the background while I was gone and I could even still check in from my phone because I have it all set up with, you know, remote access and things. But by and large, I just kind of checked out for an hour and why I came back, those were finished, and you know, that work that I was working on had progressed in some tangible way and was useful.

It's almost like the You know how the this whole conundrum of work life balance used to be a thing? And then I I think it's now well, I I think it's now more accepted that there's no such thing as work life balance, at least not in a professional setting. It's like work-life integration and you almost have to be okay with a little bit of work intruding on your your

Always.

Home life and a little bit of home life intruding on your work life, and there's some trade-off there, and you get positives and negatives on both ends. And as long as you book in both in some way loosely, it's okay. I almost feel like there's like a there's a a similar conundrum happening with AI right now. And you just described it perfectly. It's all right, well, if I can be working while I'm not working, should I be? How much of that is how much of that do I need to pre-plan and and feel bad about? not having agents run, you know, to your point 247 or for the hour that you're at the gym where you're quote unquote losing productivity, does that even matter? Right. And is it is it creating f cycles in the human brain that are just disruptive to natural creative thought because productivity, quote unquote, is about token usage as opposed to actually accomplishing something meaningful. And sometimes that downtime is actually more important without having anything running, without having any of those things behind the scenes, you know, contributing to some end goal because it allows you to reset and it allows you to come back and solve problems differently and challenges differently. I don't know, just kind of thinking about it more abstractly and bigger. There's we've been here before and we'll probably be here again with some some future evolution. But this feels like a natural extension of the work life balance challenge.

Yeah, I mean it's the same. Inherently it's probably the same exact thing where, you know, for years we used to talk about that challenge where you know I've always described the the initial interaction sort of you know, who's busiest dick measuring contest that happens a lot in in settings where you know you meet somebody for first time or you haven't talked to in a while and first thing you do is is talk about who's busiest and how busy you are. And they you know, I even engaged in that for a long time and then there was a day that like a light bulb kind of went off on my head and I was like, if I'm actually that fucking busy all the time And my life is that chaotic all the time professionally. I'm doing something wrong. And the achievement is not actually how busy you are, the achievement is honestly how much free time I have. So, you know, I started re repositioning and delegating things and, you know, approaching everything I did differently. And I think had the that was the year that I I decreed that my goal was to make myself as useless as possible. And the outcome of that was I had a lot more free time to invest in bigger picture thinking, strategic thinking for the organization, and just a lot of things that didn't keep me glued to a computer or or, you know, feeling busy, but not actually contributing in the ways that I could have the maximum impact. And think that's a similar problem to now is like I do agree with you that the the need the need for those resets are still important. So like there is an im there is an importance in just kind of like setting all the shit down and going and doing something that's not work. And for me that's really difficult because I enjoy I enjoy tinkering. I enjoy technology. So what I do for work is kind of what I do for fun. So it's super, super easy to find yourself spending all day tinkering with With things because we're getting paid to do it on behalf of clients, and that's what we do for a living. And then immediately flip over in the evenings and kind of just like keep doing the same thing because I enjoy it from that aspect. And then potentially find yourself burn out at some point where you're like, you know what? I think I'm just gonna become a fucking farmer or something and never touch a piece of technology again. Although now they're getting automated tractors and shit, which is like the coolest thing that I've seen in a long, long time.

I actually I mean what you just said is actually a a real a real conundrum too. You have technologists that people that are actually really close to the ground of things that we talk about on this podcast that that do want to do exactly that. I was just I just met some guy at lunch th just this week and he was he works in IT and he works for the government and he he literally said those words to me. He's like, dude, like my my aspiration in life right now is to just go unplug and freaking be a farmer.

Yeah.

And he's been working in IT for his whole career.

I'm I'm pretty sure there's I'm pretty sure there's like a whole swath of homesteaders that's coming. Like it's it's gonna be a huge thing that when when this generation of like the older the senior generation of technologists today, when they retire, because of all the shit that we've had to go through with evolutions and technology and just like all the changes, I'm pretty sure a lot of them are just gonna be like, you know what?

Yeah.

I just wanna live off grid. Like and you have the intelligence to do it, right? We have the ability to be like, I'm gonna set up my own solar system and like manage my own power grid and shit, but I'm just gonna be in the middle of nowhere with no people interaction, no internet, or like Starlink or something, just so you can, you know, watch TV and play video games, but like never touch a computer again or a server or whatever.

Yeah. It's a real thing, right? I mean, we could have very we could very easily retitle this episode and call it something like AI overload or, you know, AI overworking and and whatever. But it's a it's a real challenge. And I do think that, you know, as much as we embrace technology, I think the biggest balance that we've tried to strike, and you've said this many times, is that

Yeah.

Just because you're using AI to produce things is still your thing, which means that it needs to be good and it needs to be right and it needs to be accurate. And just using tokens for using tokens' sake or max or maxing out your, you know, your subscription plan for whatever time period is is a false sense of productivity. And that false sense of productivity feels good in the moment, but it in reality it actually checks no boxes. It actually just creates further rework and more rework and more rework. And that creates further and further headaches, which means it turns an eight hour day into a twelve hour day. And then you feel like you're behind the next day because you didn't get to the thing that you thought you should get to. And and that that continues that cycle continues to to progress itself. And I I think there's I think you're right. I think there is a tremendous amount of burnout currently happening. And I think that's gonna continue to happen. now that all that said, I do think bigger organizations and probably just groups at large are starting to to have the rec the recognition that AI is not a full human replacement layer and probably never will be. And so there's there's some walk back happening, not just among big tech companies, but also, you know, smaller, not smaller, but like big co's that have big technology leans that aren't aren't taking that angle any longer. They're walking back the idea of full human capital replacement and they're and they're really leaning leaning into where we've always been, which is this is about work augmentation and about freeing up humans to be able to do less busy work and do more really meaningful work. use AI for what it's good for, not as a human replacement, but as a as a as a buddy.

But think the the challenge that we've seen in a lot of those organizations, at least lately, is they got kind of like a taste of things and got drunk on it. And you know, in many organizations now I think some of them have stopped, but like the measurement of success, there's always this desire within organizations, every organization, we've even tried it ourselves, to like apply a metric to productivity. And we know firsthand it's incredibly fucking difficult to apply a productivity metric. And it ultimately leads to like whatever you measure is what you'll get more of, right? And that isn't necessarily a good thing. In development, for the longest time, people measured productivity with lines of code. So what you wind up with is just like a whole shitload of code being written. That maybe isn't actually necessary and actually just increases your code debt and the complexity of your product when in reality, you know, two or three really good lines of code could accomplish the same thing that this 50 lines of code garbage does. But from a productivity perspective, if I if I do that, even though that's correct, that's more correct. I'll get dinged on my performance review and all those other things. We see the same thing with like the token maxing that goes on right now, where organizations said, hey, whoever's using the most AI, they're doing the best. So they set up, you know, leaderboards and those sorts of things. So well, fuck yeah, we can use the shit out of some tokens. Are we actually producing anything with them? Don't know, it doesn't matter. We're not measuring that. We're not measuring the outputs, we're just measuring how many tokens we used and who's using the most, who's AIing the most and the best. And I think the next evolution of that'll be, you know, organizations going, shit, now we're just spending a lot of money on A on AI tokens. We act this is not sustainable. We can't we can't provide everybody unlimited budgets of so now you're gonna have these like AI token budgets that exist. And you can spend that budget, but then there's like some efficiency metric that comes along with it of like producing a certain amount of outputs based on the tokens it took to to do it. Like I'm not exactly sure how all of it's gonna work out, but there's some that'll be like the next layer, it'll still fail, but it's like the next way that we're gonna try to measure how this stuff works is like who can who can generate the same output. with fewer tokens used because then they're effectively more efficient at AI usage than other people. It's gonna be strange.

Yeah, there's there's a weird balance there. And it gets even trickier because now you have models that are coming out that are like they use four to ten times more token usage than we've ever used with any model that's ever been out before. And they're supposed to be quote unquote better, but you don't really know if they're really that much better. And so there's there's this this the this economy of tokens for productivity that I think is gonna have to balance itself out somewhere somehow. be beyond just like people usage, right? I'm talking just the economics to your point of companies expecting certain a certain level of output for each token utilized and that math just not mathing. And I think we're we're seeing that start to happen now, especially in in in bigger companies and even even in our company with some of this stuff. It's like, well, is it worth, you know, throwing AI at these things or do we just is it even worth doing at all? Right. So those questions that

I mean that's the that's the other one that's difficult too, right? Like I've I've had that conversation with you know some folks when they're like, you know, I can see them prompting an AI agent to do some stuff. I'm like, hey, you know, you still edit fucking files, right? Like you could just type the change that you're trying to get it to make versus being like rename this file to this and it renames it something r you know completely wrong. And so you've just you've just burnt minutes. Of your life that you'll never get back. And tokens, I don't even care about the tokens at that point. It's just they, you know, it's very easy to get into the the minds. And I've certainly done it too. I've just been like, you know, rename all this stuff and package it up. Like, I do know how to run the rename command and compress a zip. But I had the agent do it anyways. But now tarballing something, different story. Still no, still don't know that command.

Yep.

We'll never know that fucking command. i it's It's interesting to see sort of the overuse patterns that emerge there. and I think in part, you know, kind of the the the basis of what we're talking about, that that that fear that like I'm missing out on something, that somebody knows how to do this better than me, that I don't have enough agents running, or I'm not using I'm not getting maximum efficiency out of it, I think is all is ultimately born out of the the Inherent fear of like being passed up a replacement because of whatever what the demands are are seeking or what we think the demands are seeking. and we still don't even know where that's gonna land, right? We still don't know exactly where all of the markets are gonna land. you've got people like Dario from Anthropic who've been saying for two years. that you know developers are not gonna be needed at all. So then you have you know I've had discussions with people thinking about becoming developers and they're like, well, should I learn how to write code? Like seems like nobody's gonna need to do that anymore. I'm like, I don't think he's right. I think he's wrong. do you have people on the complete opposite angle where they're where they would swear up and down that AI can never write write proficient code. And I'm like, you're probably wrong too.

in and I think Mm-hmm. There's there is that balance somewhere. And I think you have almost every, I'll call them celebrity technologist, for lack of a better phrase, that are starting to walk back all of that, right? They're they have all been proponents of, well, it's going to replace all marketing jobs, it's going to replace all creative jobs, it's going to replace all development jobs, it's going to replace all information jobs. It's like, mm-hmm, it's not. It's gonna, it's gonna remove and has remove some stupidity. From where humans were doing things that they shouldn't be doing, because it's just a a massive waste of human capital and human energy. It'll remove a lot of that in the same way that using a calculator removed the abil or the the need to hand do calculations, right? It's it's an unnecessary thing to get to the same output. It's it's it's a different tool with a very different level of complexity than a calculator, obviously. However, it's the exact same idea. It's removing some of the things that humans just don't need to be doing. and in adding to the I can all use the word productivity, using adding to the better outcome, better output. At least that's the idea of it, right? I mean, most people are not using it that way now, and you know, probably myself inclusive in some times of, well, that was trash. I'm gonna throw that away. That was a massive waste of energy. But it's an experiment, it allows you to experiment with things that you may not otherwise.

Yeah. I I definitely think like from a a tangible deliverable perspective, one of the things that we always have on website projects, right? We build websites for clients all the time. migrations from their existing website to a new website is always a pain in the ass. And it's always that way because the the content is never never fits one to one. There's always like a little finagling that needs to happen. and there could be a lot of finagling that needs to happen. And that's an area where we've found that like we can use we can use AI agents to help do that finagling where previously you just basically had to Mechanical Turk, right? You're you're just you're getting some whoever draws the short straw gets to enter a whole bunch of content. And by and large, it's a lot of copy and paste. And in some cases, there's some reformatting or some decision making that has to happen. But that has allowed us to provide better end products for our clients. and in some cases, save them some money on that that aspect of the the project. But really, it allows us to deliver a more complete and quality product. without an exorbitant more increase in effort.

Right. And and I also think that's a difference too. We were talking before the show about a a study that just came out for agencies in particular. And it talks about this idea of what clients expect from AI utilization versus what agencies are interpreting that the clients want as part of AI utilization. And really what clients want is transparency that you're using AI. So disclosure of some kind of what's going on, how you're using it, why you're using it, et cetera. And then the second is better, better outcomes. Better ideas, better thoughts, better thinking. They don't like in fact, in the even in the same study, it says that most agencies that are leaning heaviest into AI are actually the ones that their pipeline is struggling the most because you can tell it doesn't pass the sniff test. Right. So they're leaning into these AI expectations and that, you know, we can do more for less and cheaper and and whatever. And that's not what clients want. They don't expect to spend less. They actually expect to send spend the same amount but get a better product. Because you're you're offsetting some of your overall tooling with something else. And I I think that's probably true not just in agencies, but certainly all professional services and probably most organizations that are usually utilizing AI in general. At least that's they that's how things should equate, is it's not about lowering costs in and of themselves. That may be a a derivative outcome in the end for some situations. But what we should be doing is actually leaning in and doing more work with it. In fact, I think Jensen Wang from from NVIDIA is actually the one that said this, completely indirectly related to agencies. But he said, you know, organizations that are actually doing the cost cutting deserve to fail. Because what they should be doing is taking those same investments and going into tangential industries, finding new markets, finding new products, finding new, new ways to reinvest those same dollars, not just throwing it to the bottom line. That's ridiculous. Is in you can see NVIDIA actually is living is living proof of that. I mean, by way of saving dollars o across a bunch of different stuff, they've entered into massive amounts of new markets in the last three years. And so it's been in really interesting to see a big company like that eat their own dog food of AI is a a productivity enhancer, not a human replacer.

Well and you see companies that went the other way that are now doubling back on that that stance, right? Salesforce is the one that comes to mind immediately where they were early adopters of AI and and I'm sure somebody in you know senior VP executive level was like, man, we can get rid of all of our staff now and just replace them with AI. And all of those savings go straight to the bottom line. So they decreed that Agentforce is the future and you know, laid off a whole bunch of staff because they didn't need them anymore. Only to find out that Agentforce is not the future and they need all those people back. yeah. And you know the net of that's gonna be it's gonna cost them more money, right? It would have been actually probably cheaper to deploy both and tan, like keep your people, deploy Agentforce in tandem, and probably find.

Imagine that.

some balance that actually exists and and some tangential products or market share that you could capitalize on that you're not currently capitalizing on. But instead they were so focused, laser focused on just like how can we cut everything. That you forget, you almost forget how you got there in the first place, right? The reason that companies like Salesforce exist are because they had really fucking smart people working on stuff, pitching ideas internally that turn into actual products that they go sell to people. If you get rid of all the smart people, now you just have I mean you certainly have a product, but you can basically ride that product until it's dead. You got rid of everybody else who's gonna advance it.

Yep, that's true. So what else can we say about this?

I think acknowledgement is a big thing, right? Like the the idea that this exists is not is not probably not new. It's it's something that's always existed. I think it's probably compounded by It's compounded by the volume. There's a there's a high volume of activity out there that's talking about things that will make you feel like you're missing out. There's a graphic that I saw on on X yesterday that, you know, reminded me even that the majority of people have not even used AI at all. So if you're finding yourself in a position where you're like, you know what, I use I use Claude, I use OpenCode, I use whatever. but I'm not using it in these ways that you know this clickbait article is saying they are and I'm don't have things running 24-7, that's okay. That's like the the point zero zero zero zero zero zero zero one percent of people are operating in that fashion. And there's n yet to be any productivity proof come out of them yet. there are certainly tons of ways that can be used and there's tons of experiments and I love seeing all the experiments that people were performing. One of them, you know, I was messing with this week because I burnt up a bunch of my Fable usage right before Anthropic extended the deadline again. And I was like, shit, how do I get the most out of the remainder of Fable that I have to work with on my subscription? And, you know, so we've started playing with the idea of using Fable as the orchestrator. So Fable just tells Codex what to do, which uses a lot less of of Fable. And got me a little bit of extra. Is that more efficient? Is that something I would continue forever? I don't know. I've I've done some, you know, kind of side-by-side tests, and there's some marginal differences, but I'm not sure if it's something that's that much better than just saying, hey, codex, go do this. so maybe I maybe I would, maybe I wouldn't. But I think that recognition that, you know, this this AI FOMO idea is kind of here to stay and just checking yourself every once in a while to make sure that you're not being you're not falling victim to it and are taking the time to make sure that what you're doing is actually yielding fruit and being willing to throw something away and say, you know what, I've spent a bunch of time on this 15 agent architecture that is actually not adding any fucking value. Let me just throw it away. and deploy, you know, a single OpenClaw instance or a single Hermes agent. I did that for one of mine. I had like five agents and it was just gross and I found that just one was far more efficient.

Yeah, I think from my vantage point, it goes back to this convergence of of work life balance and knowing knowing when to put it down, knowing when to walk away, knowing that it's fine if you have some periods of time that aren't being filled with quote unquote productivity. And also remembering that this agent, these agents are literally programmed to continue to get you to continue to use them. That is what they're made to do. And so they will give you the the false accolades and they will run you in circles over and over again, whether it's intentional or not on the on the maker's behalf, that's how they that's how they behave. And so just knowing when to stop that pro that that loop and stop the the the chaotic nature of how it kind of will lead you down a path is probably maybe my number one takeaway from this conversation of just how how how to stop that FOMO.

Well, till next time.

Till next time.