What type of AI user are you?

00:00:00,600 --> 00:00:35,370 [eric] [upbeat music] Welcome back to the Token Intelligence Show. AI is changing the way we work, and here on the Token Intelligence Show, we want to show you what the state of the art is, cut through the noise and the hype, and help you use wisdom to be a great leader in this new era that we're in. One of the topics that we've covered a lot, especially in the last couple shows, John, 00:00:36,480 --> 00:00:50,740 [eric] is that the default out-of-the-box pathway that using AI takes you down, uh, in the context of, you know, sort of knowledge work supported by or outsourced to AI, 00:00:51,780 --> 00:00:53,490 [eric] is that it can actually 00:00:54,760 --> 00:00:56,890 [eric] degrade your core skill set, 00:00:58,160 --> 00:01:00,160 [eric] which is concerning. 00:01:00,160 --> 00:01:00,510 [john] Yeah. 00:01:00,510 --> 00:01:13,780 [eric] Um [chuckles] but I think that we're seeing, I think we're seeing evidence of that more and more. Um, let's actually start this episode by defining the term AI slop. 00:01:13,780 --> 00:01:14,280 [john] Okay. 00:01:14,280 --> 00:01:19,680 [eric] It gets thrown around a lot, but I'm interested in what your definition of AI slop is. 00:01:21,220 --> 00:01:23,380 [john] Yeah, I think it's a relative definition, right? 00:01:23,380 --> 00:01:23,560 [eric] Mm-hmm. 00:01:24,580 --> 00:01:32,760 [john] Like, it, it's a junk drawer term for something that was produced to, by AI that doesn't measure up to somebody else's standard, is what I would say. 00:01:32,760 --> 00:01:39,660 [eric] Okay, that's good. Okay. Th- that's interesting. That's... I- my definition would be spicier. 00:01:39,660 --> 00:01:40,440 [john] Okay. I'm ready. 00:01:41,740 --> 00:01:48,760 [eric] The, uh, I would say that AI slop is... 00:01:50,300 --> 00:01:53,020 [eric] It, it is output 00:01:54,240 --> 00:01:54,720 [eric] that 00:01:57,040 --> 00:02:02,900 [eric] bears the very clear markings of being machine-produced, not human-produced. 00:02:02,900 --> 00:02:03,060 [john] Hmm. 00:02:04,620 --> 00:02:05,040 [eric] And 00:02:06,940 --> 00:02:10,639 [eric] what's interesting is I think that definition and your definition- 00:02:10,640 --> 00:02:11,320 [john] Mm-hmm 00:02:11,320 --> 00:02:15,400 [eric] ... apply across different modalities. 00:02:15,400 --> 00:02:16,000 [john] Yeah. 00:02:16,000 --> 00:02:29,780 [eric] So code, uh, we could, you know, writing, um, images. I think it s- can sort of be applied, you know, across the spectrum, which is really interesting, but it is, it is highly subjective. 00:02:29,780 --> 00:02:31,380 [john] Yeah, because if I bring 00:02:32,400 --> 00:02:34,660 [john] a management consulting firm, McKinsey, in- 00:02:34,660 --> 00:02:34,700 [eric] Mm-hmm 00:02:34,700 --> 00:02:39,460 [john] ... and say, "Hey, we need a modernization roadmap for whatever"- 00:02:39,460 --> 00:02:39,549 [eric] Yep 00:02:39,549 --> 00:02:49,600 [john] ... versus I have, you know, a junior level employee like, "Hey," like, "can you put together a roadmap for, for what you think would be..." Like, the expectation of the output is different. 00:02:49,600 --> 00:02:49,790 [eric] Yes. 00:02:49,790 --> 00:02:59,320 [john] Right? And, and I think, well, probably rightly so, if a similar output was delivered from both people regardless of how it happened, and say AI was involved in both- 00:02:59,320 --> 00:02:59,820 [eric] Mm-hmm 00:02:59,820 --> 00:03:11,560 [john] ... and, and, and some amount of human steering, it doesn't really matter. The... For sure, if McKinsey delivered something that didn't meet some sort of standard, it would be labeled AI slop- 00:03:11,560 --> 00:03:11,840 [eric] Mm-hmm 00:03:11,840 --> 00:03:13,740 [john] ... even if they did maybe have a human working on it. 00:03:13,740 --> 00:03:13,890 [eric] Mm-hmm. 00:03:13,890 --> 00:03:20,440 [john] And then vice versa, maybe a junior person kind of outperforms, like, uses AI, and it's like, "Oh, it looks pretty good." 00:03:20,440 --> 00:03:20,900 [eric] Yeah, yeah. 00:03:20,900 --> 00:03:20,960 [john] Um- 00:03:20,960 --> 00:03:21,920 [eric] Absolutely. 00:03:21,920 --> 00:03:22,840 [john] So. 00:03:22,840 --> 00:03:28,530 [eric] Okay. The reason I wanted to try to define AI slop was, one, to just level set on that. 00:03:28,530 --> 00:03:29,240 [john] Mm-hmm. 00:03:29,240 --> 00:03:40,180 [eric] But two, it's the first stop on our three-stop journey today, talking about archetypes of people who, uh, use AI, and this comes from- 00:03:40,180 --> 00:03:40,340 [john] Yeah 00:03:40,340 --> 00:03:44,780 [eric] ... an article that you ran across in The Atlantic. So give us a- 00:03:44,780 --> 00:03:45,000 [john] Yeah 00:03:45,000 --> 00:03:45,420 [eric] ... quick rundown. 00:03:45,420 --> 00:04:01,380 [john] So I have to shout out, um, AI Daily Brief. It's actually, it's one of the top AI podcasts right now. They do, um, just an awesome job. It's like 20 minutes and do headlines and then a, a little bit of a, a deep dive on an article or topic- 00:04:01,380 --> 00:04:01,530 [eric] Mm-hmm 00:04:01,530 --> 00:04:07,740 [john] ... um, every day. And so they brought up and then I found this article, um, by David Brooks from The Atlantic. 00:04:07,740 --> 00:04:08,200 [eric] Yep. 00:04:08,200 --> 00:04:24,220 [john] Um, and he's going over these three archetypes, and I just thought it was really interesting. Um, 'cause anytime you, you bring archetypes into the conversation, you're, you're, you're kind of exaggerating things that are going on, but it's really helpful to, to dig deep into, like, one aspect of how, you know, AI is affecting the way we work. 00:04:24,220 --> 00:04:37,700 [eric] Absolutely. And David Brooks is a very seasoned thinker and writer who has published many wonderful pieces over the years in The Atlantic. Very thought-provoking. Um, so he gets a thumb- 00:04:37,700 --> 00:04:37,940 [john] Yeah, I'm excited. 00:04:37,940 --> 00:04:39,500 [eric] He gets a thumbs up in my book- 00:04:39,500 --> 00:04:39,740 [john] Awesome 00:04:39,740 --> 00:04:43,360 [eric] ... you know, of people who write things that make you think really hard- 00:04:43,360 --> 00:04:43,420 [john] Mm-hmm 00:04:43,420 --> 00:04:44,390 [eric] ... which I think is good, you know- 00:04:44,390 --> 00:04:44,390 [john] Yep 00:04:44,390 --> 00:04:45,580 [eric] ... whether you agree with it or not. 00:04:45,620 --> 00:04:46,160 [john] Yep. 00:04:46,160 --> 00:04:49,630 [eric] Um, so okay, let's start- 00:04:49,630 --> 00:04:50,000 [john] Start with the first one? 00:04:50,000 --> 00:04:51,120 [eric] Let's start with the first one. 00:04:51,120 --> 00:04:51,140 [john] All right. 00:04:51,140 --> 00:04:52,500 [eric] Give me the first archetype. 00:04:52,500 --> 00:04:57,380 [john] All right. So the... And I like this one. Um, it's the productive passenger. 00:04:57,380 --> 00:04:58,100 [eric] Okay. 00:04:58,100 --> 00:05:01,660 [john] Um, it reminds me, do you know what a passenger princess is? 00:05:01,660 --> 00:05:02,260 [eric] No. 00:05:02,260 --> 00:05:03,030 [john] You never heard that term? 00:05:03,030 --> 00:05:04,760 [eric] Passenger princess, this is great. 00:05:04,760 --> 00:05:05,479 [john] [laughs] Um- 00:05:05,480 --> 00:05:06,840 [eric] I'm so excited about this 00:05:06,840 --> 00:05:07,840 [john] ... it's a funny term. I- 00:05:07,840 --> 00:05:07,850 [eric] Yeah 00:05:07,850 --> 00:05:23,000 [john] ... I hadn't heard it until, like, recently. But the term is somebody, like, they're just along for the ride. They're not in charge. They're just, like, hanging out, like, uh, you know, imagine you're on a road trip and you're just riding shotgun doing your own thing and you're, you're not navigating, you're not doing anything, like- 00:05:23,060 --> 00:05:23,400 [eric] Yeah 00:05:23,400 --> 00:05:25,020 [john] ... just, you know, just here for the ride. 00:05:25,020 --> 00:05:25,780 [eric] Exactly. 00:05:25,780 --> 00:05:25,820 [john] That's how I would describe it. 00:05:25,820 --> 00:05:29,200 [eric] So this is my daughters in the backseat of the car- 00:05:29,200 --> 00:05:29,570 [john] Yeah. [laughs] 00:05:29,570 --> 00:05:31,820 [eric] ... listening to K-pop demon hunters. 00:05:31,820 --> 00:05:34,680 [john] Sure, yeah. [laughs] Right. Yeah. 00:05:34,680 --> 00:05:34,870 [eric] Yeah. 00:05:34,870 --> 00:05:37,170 [john] Yeah. No, like, completely clueless of, like, where we're going or what's happening. 00:05:37,170 --> 00:05:38,340 [eric] Yeah, they could be anywhere. 00:05:38,340 --> 00:05:38,900 [john] Yep. 00:05:38,900 --> 00:05:39,350 [eric] They could be- 00:05:39,350 --> 00:05:39,350 [john] Yeah 00:05:39,350 --> 00:05:40,039 [eric] ... anywhere. Yeah. 00:05:40,040 --> 00:05:48,040 [john] So anyways, I, I think when we talk about the productive passenger, that's what, what comes to mind a little bit. And then, uh, productive, 00:05:49,240 --> 00:05:51,490 [john] we'll see. We'll see if that's a fair term- 00:05:51,490 --> 00:05:51,490 [eric] Mm-hmm 00:05:51,490 --> 00:06:04,016 [john] ... termina- like, term here. But I think the idea is people that are using AI with fairly low cognitive effort and then getting an, uh, an output Which is de- pretty much determined by the, the AI. 00:06:04,016 --> 00:06:04,656 [eric] Mm-hmm. 00:06:04,656 --> 00:06:06,176 [john] With variable results. 00:06:06,176 --> 00:06:06,796 [eric] Yep. 00:06:06,796 --> 00:06:11,856 [john] And I don't think it's always ne- And sometimes it works [laughs] and sometimes it doesn't, right? 00:06:11,856 --> 00:06:12,155 [eric] Mm-hmm. 00:06:12,156 --> 00:06:17,376 [john] Sometimes it's kinda that AI slop category, and sometimes, like, it's actually pretty good. 00:06:17,376 --> 00:06:17,856 [eric] Yeah. 00:06:17,856 --> 00:06:20,676 [john] But there's gonna be for sure variability. 00:06:20,676 --> 00:06:20,995 [eric] Yep. 00:06:20,996 --> 00:06:21,136 [john] Um... 00:06:22,256 --> 00:06:28,996 [eric] The, this is what's interesting to me about the, quote-unquote, "productive passenger"- 00:06:28,996 --> 00:06:29,136 [john] Mm 00:06:29,136 --> 00:06:41,016 [eric] ... and interesting in general with AI, and I do have a conclusion here for people who would categorize themselves as a productive passenger. Or, or maybe a, an admonition maybe is a good [laughs] 00:06:41,016 --> 00:06:42,736 [john] An admonition. All right. 00:06:42,736 --> 00:06:42,956 [eric] Uh, 00:06:44,296 --> 00:06:56,436 [eric] the... Okay. I, I think it's easy to default to thinking about the productive passenger as more of a beginner or a novice- 00:06:56,436 --> 00:06:56,446 [john] Okay 00:06:56,446 --> 00:06:57,145 [eric] ... type profile. 00:06:57,145 --> 00:06:57,176 [john] Yeah. 00:06:57,176 --> 00:06:59,216 [eric] Right? Like, okay, it's the junior- 00:06:59,216 --> 00:06:59,636 [john] Mm-hmm 00:06:59,636 --> 00:07:01,316 [eric] ... you know, who's producing this report, 00:07:02,556 --> 00:07:11,356 [eric] you know, whatever it is. Um, but just this week, Brian Chesky, the CEO of Airbnb- 00:07:11,356 --> 00:07:11,676 [john] Okay 00:07:11,676 --> 00:07:14,076 [eric] ... which is a phenomenally successful company- 00:07:14,076 --> 00:07:15,076 [john] Mm-hmm 00:07:15,076 --> 00:07:28,316 [eric] ... uh, got a lot of heat for posting a thread, posting on X and, uh, making a post on X, and then adding threaded comments as people do when they wanna share- 00:07:28,316 --> 00:07:28,326 [john] Mm-hmm 00:07:28,326 --> 00:07:29,316 [eric] ... a longer form thought- 00:07:29,316 --> 00:07:29,656 [john] Right 00:07:29,656 --> 00:07:31,026 [eric] ... that's not an article. 00:07:31,026 --> 00:07:31,616 [john] Right. 00:07:31,616 --> 00:07:33,536 [eric] They'll, they'll sort of post like an introductory thought- 00:07:33,536 --> 00:07:33,976 [john] Mm-hmm 00:07:33,976 --> 00:07:37,235 [eric] ... and then follow up with comments that sorta articulate the different points that they wanna make. 00:07:38,336 --> 00:07:45,996 [eric] And it got deleted because he got a ton of heat for this being very clearly 00:07:47,156 --> 00:07:49,256 [eric] AI-generated. Um- 00:07:49,256 --> 00:07:53,436 [john] The, uh, the int- the initial post and the, like, follow-up? 00:07:53,436 --> 00:07:53,786 [eric] Yes. 00:07:53,786 --> 00:07:53,816 [john] Okay. 00:07:53,816 --> 00:07:55,536 [eric] Especially the follow-up ones- 00:07:55,536 --> 00:07:55,546 [john] Okay 00:07:55,546 --> 00:08:16,185 [eric] ... where it was just, I mean, it, it had the clear markings of being AI-generated, right? What's tricky about this is that the, um, you know, what are the hallmarks of AI-generated? It's not necessarily easy to tell that, right? 00:08:16,185 --> 00:08:16,935 [john] Right. 00:08:16,936 --> 00:08:21,386 [eric] But I think people who have followed Brian Chesky for a long time- 00:08:21,386 --> 00:08:21,566 [john] Mm-hmm 00:08:21,566 --> 00:08:23,156 [eric] ... had an innate sense that- 00:08:23,156 --> 00:08:23,316 [john] Right 00:08:23,316 --> 00:08:26,836 [eric] ... he didn't, he didn't write that, right? They m- 00:08:26,836 --> 00:08:26,916 [john] Right 00:08:26,916 --> 00:08:32,646 [eric] ... you know, he may have originated the thoughts, but, um, but he didn't write it, so he caught a lot of heat, right? And you're talking- 00:08:32,646 --> 00:08:32,646 [john] Sure 00:08:32,646 --> 00:08:39,116 [eric] ... about someone who is an incredibly successful, incredibly articulate- 00:08:39,116 --> 00:08:39,276 [john] Right 00:08:39,276 --> 00:08:39,956 [eric] ... person- 00:08:39,956 --> 00:08:40,146 [john] Right 00:08:40,146 --> 00:08:43,336 [eric] ... uh, who to- who 00:08:44,416 --> 00:08:47,025 [eric] got out of the driver's seat, got into the- 00:08:47,025 --> 00:08:47,086 [john] Yeah 00:08:47,086 --> 00:08:49,826 [eric] ... passenger seat, and became [laughs] a productive passenger, right? 00:08:49,826 --> 00:09:03,706 [john] Well, and, and to be f- fair to him, and when, when we go through these archetypes, it's not all or nothing. You may, you may be a phenomenal executive that's a productive passenger in social media only. 00:09:03,706 --> 00:09:03,716 [eric] Yep. 00:09:03,716 --> 00:09:07,896 [john] That's probably a bad idea, but [laughs] just where we are right now. 00:09:07,896 --> 00:09:07,956 [eric] Right. 00:09:07,956 --> 00:09:13,656 [john] But you could be that and be fully engaged in a different level in other, you know, aspects. And I- 00:09:13,656 --> 00:09:13,716 [eric] Yep 00:09:13,716 --> 00:09:23,216 [john] ... and I think that's why this is tricky. Just 'cause in this case, like, everybody's gonna kind of judge and, and assume that, oh, well, like, you know, he doesn't know what he's doing. You know, blah, blah, blah. 00:09:23,216 --> 00:09:23,876 [eric] Yep. 00:09:23,876 --> 00:09:24,406 [john] Um... 00:09:24,406 --> 00:09:32,416 [eric] Uh, I think one of the other things that makes it tricky... So I- it's easy to default to like, oh, well, you know, more of a beginner, less experienced person- 00:09:32,416 --> 00:09:32,426 [john] Mm 00:09:32,426 --> 00:09:35,456 [eric] ... is gonna be the productive passenger where they just sort of outsource it, right? 00:09:35,456 --> 00:09:35,656 [john] Yes. 00:09:35,656 --> 00:09:37,016 [eric] Well, that's not necessarily the case- 00:09:37,016 --> 00:09:37,236 [john] Yeah 00:09:37,236 --> 00:09:37,836 [eric] ... as we're seeing. 00:09:37,836 --> 00:09:38,776 [john] Right. 00:09:38,776 --> 00:09:50,496 [eric] I think the other really tricky thing is that you have the, these different axes of starting point and task. And so 00:09:51,796 --> 00:10:00,636 [eric] what I mean by that is the, uh, sorta core skill set that someone starts with to do a particular task- 00:10:00,636 --> 00:10:01,256 [john] Mm-hmm 00:10:01,256 --> 00:10:05,056 [eric] ... matters a lot when you think about being a productive passenger. 00:10:05,056 --> 00:10:05,256 [john] Right. 00:10:05,256 --> 00:10:12,226 [eric] So for example, uh, let's say... I, I think about my dad all the time because he's starting to use AI. 00:10:12,226 --> 00:10:12,275 [john] Mm-hmm. 00:10:12,276 --> 00:10:14,536 [eric] He's 70 years old. He has his own business. 00:10:14,536 --> 00:10:15,016 [john] Mm-hmm. 00:10:15,016 --> 00:10:18,336 [eric] And they've never really done any digital marketing. 00:10:18,336 --> 00:10:18,556 [john] Yeah. 00:10:18,556 --> 00:10:20,465 [eric] So he's starting from zero, and so he's- 00:10:20,465 --> 00:10:20,465 [john] Sure 00:10:20,465 --> 00:10:27,356 [eric] ... talked about, he's told me, "I'm using Claude, you know, to generate, like, marketing emails," right? 00:10:28,396 --> 00:10:31,146 [eric] And so going from zero to what Claude produces 00:10:32,216 --> 00:10:33,276 [eric] is probably- 00:10:33,276 --> 00:10:33,286 [john] Yeah 00:10:33,286 --> 00:10:34,696 [eric] ... a win for his business. 00:10:34,696 --> 00:10:34,726 [john] Right. 00:10:34,726 --> 00:10:35,086 [eric] Right? 00:10:35,086 --> 00:10:35,175 [john] Right. For sure. 00:10:35,176 --> 00:10:39,736 [eric] Um, and is that AI slop, right? This is the, this- 00:10:39,736 --> 00:10:39,786 [john] Right 00:10:39,786 --> 00:10:41,096 [eric] ... that's like the difficult, subjective- 00:10:41,096 --> 00:10:41,146 [john] Mm-hmm 00:10:41,146 --> 00:10:41,816 [eric] ... part of it, right? 00:10:42,836 --> 00:10:54,295 [eric] Um, but also the task of a marketing email, right, or sort of the output or that axis, is much lower consequence than making a critical code change- 00:10:54,296 --> 00:10:54,406 [john] Sure 00:10:54,406 --> 00:10:57,596 [eric] ... or in your world, performing a data, a data analysis- 00:10:57,596 --> 00:10:57,636 [john] Right 00:10:57,636 --> 00:10:59,876 [eric] ... that's gonna help someone make a strategic decision. 00:10:59,876 --> 00:11:00,126 [john] Right. 00:11:00,126 --> 00:11:05,415 [eric] Right? And so, um, the threshold for the consequence relative to slop- 00:11:05,416 --> 00:11:05,596 [john] Right 00:11:05,596 --> 00:11:07,336 [eric] ... you know, sorta depends. But 00:11:08,496 --> 00:11:21,576 [eric] all that to say, my admonition around the productive passenger is to spend as little time in the passenger seat as you can. 00:11:21,576 --> 00:11:22,016 [john] Yep. 00:11:22,016 --> 00:11:25,356 [eric] Because I think that it is, 00:11:26,936 --> 00:11:37,216 [eric] it is totally fair game for certain things, but for many things, especially anything where you're producing some sort of output, 00:11:38,336 --> 00:11:38,836 [eric] I think you 00:11:39,856 --> 00:11:47,316 [eric] can get dangerously close to this line where you erode your actual human- 00:11:47,316 --> 00:11:47,696 [john] Mm-hmm 00:11:47,696 --> 00:11:48,296 [eric] ... input. 00:11:49,436 --> 00:11:53,436 [eric] You, you erode the human quality of the output by- 00:11:53,436 --> 00:11:53,496 [john] Yeah 00:11:53,496 --> 00:11:57,626 [eric] ... giving input and then, you know, having AI sorta produce the output. 00:11:57,626 --> 00:11:57,626 [john] Yeah. 00:11:57,626 --> 00:12:08,814 [eric] Right? Uh, and so I think that any time you feel like you're getting into the passenger seat is a really good gut check of like- Hmm. 00:12:08,814 --> 00:12:08,844 [john] Right 00:12:08,844 --> 00:12:20,044 [eric] Like, am I outsourcing this in a way that is going to long term build my skill set? You know, let, let the, let humanity shine through whatever the output is, right? 00:12:20,044 --> 00:12:20,764 [john] Yep. 00:12:20,764 --> 00:12:22,874 [eric] But I do also... I, I do wanna qualify that by saying, 00:12:24,024 --> 00:12:39,424 [eric] I think there's a whole category of stuff like research or, uh, finding information, or summarizing, you know, 28 differently formatted PDFs or those sorts of things where it's like, that [laughs] 00:12:39,424 --> 00:12:40,303 [john] Right. 00:12:40,303 --> 00:12:42,014 [eric] Outsource all of that, right? Like we don't- 00:12:42,014 --> 00:12:42,014 [john] Right 00:12:42,014 --> 00:12:55,674 [eric] ... wanna go back to a world where, um, you know, you don't have the ability to, to leverage AI to automate really menial tasks like that, that weren't valuable for you to- 00:12:55,674 --> 00:12:55,674 [john] Right 00:12:55,674 --> 00:12:56,824 [eric] ... do in many cases anyways. 00:12:56,824 --> 00:12:57,184 [john] Right. 00:12:57,184 --> 00:13:00,204 [eric] Not always, but I think in general it can be helpful. 00:13:00,204 --> 00:13:05,204 [john] Yeah. Yeah, I think this leads us into our next one, which is the reluctant optimizer. 00:13:05,204 --> 00:13:06,384 [eric] Hmm. 00:13:06,384 --> 00:13:06,844 [john] And- 00:13:06,844 --> 00:13:08,844 [eric] Why reluctant? This one's interesting to me. 00:13:08,844 --> 00:13:12,164 [john] Yeah. I, I think, I think the reluctancy 00:13:13,484 --> 00:13:20,984 [john] here... And not necessarily reluctance to, um, adopt AI particularly. There could be some of that. 00:13:20,984 --> 00:13:21,944 [eric] Mm-hmm. 00:13:21,944 --> 00:13:22,114 [john] But 00:13:23,204 --> 00:13:23,564 [john] I think, 00:13:24,704 --> 00:13:34,044 [john] I think it's this- I think this is where most people are, where like, okay, my boss wants, my boss wants me to use AI, so maybe there's a little reluctance there. 00:13:34,044 --> 00:13:34,084 [eric] Yep. 00:13:34,084 --> 00:13:46,784 [john] And you do it, and you use it, and you're like, "Oh, well this is nice." Um, but you find yourself being constantly pulled, like, into almost like pulled down to average, right? 00:13:46,784 --> 00:13:46,814 [eric] Hmm. 00:13:46,814 --> 00:13:55,014 [john] 'Cause that's what I, that's what I feel like one of the primary functions of AI is, pulling you into average, or maybe a little below average depending on what you're doing. 00:13:55,014 --> 00:13:55,064 [eric] Mm-hmm. 00:13:55,064 --> 00:14:15,284 [john] And there's just like this convenience and like, well it's like, ah, I have so many things going on. K- kind of the like... We just did a show on multitasking. This like multitasking mode combined with convenience, combined with this pull to the average that, like, this person essentially kind of oscillates in and out of the passenger seat, I would say. 00:14:15,284 --> 00:14:16,084 [eric] Hmm. 00:14:16,084 --> 00:14:22,923 [john] And I think some of these people have this reluctance to... 'Cause they have a little, I think they have a l- a little bit of, um, recognition of this. 00:14:22,924 --> 00:14:23,284 [eric] Hmm. 00:14:23,284 --> 00:14:41,044 [john] So they think there's a bit of a reluctance to adopt this, 'cause like I'm on like, I don't know, like I th- think maybe my work was a little better before, but if my work goes down 20%, or b- let's call it 15% in quality and I can produce five times as much, like maybe it makes sense. Like, I think that's the weird middle ground. 00:14:41,044 --> 00:14:47,424 [eric] Yeah. It is. Here's a specific example of that. The... If you 00:14:49,244 --> 00:14:55,924 [eric] ask... If you have two separate people who are performing the same task and they use AI to do it- 00:14:55,924 --> 00:14:58,173 [john] Mm-hmm 00:14:58,173 --> 00:15:03,284 [eric] ... when you say sort of digress to the average, right? Um, 00:15:05,464 --> 00:15:08,274 [eric] the... Or get pulled towards the average. 00:15:08,274 --> 00:15:08,304 [john] Hmm. 00:15:08,304 --> 00:15:16,964 [eric] I think one way a lot of people experience that is, one, more conformity in how your craft is performed- 00:15:16,964 --> 00:15:17,224 [john] Yeah 00:15:17,224 --> 00:15:19,964 [eric] ... because everyone's using the same tools, right? 00:15:19,964 --> 00:15:23,464 [john] 'Cause you're using a word calculator, and the word calculator [laughs] pulls you toward the average. 00:15:23,464 --> 00:15:23,704 [eric] Right. 00:15:23,704 --> 00:15:29,044 [john] But the average is going up over time, but asymmetrically, right? 00:15:29,044 --> 00:15:29,054 [eric] Uh, 00:15:30,224 --> 00:15:31,644 [eric] I don't know if I would agree with that. 00:15:31,644 --> 00:15:38,124 [john] With model, like various model improvements? Um, I think we see asymmetric increase in- 00:15:39,844 --> 00:15:47,164 [eric] I think it depends on... I think the, I think what you're working on, I think is highly variable depending on what you're working on. 00:15:47,164 --> 00:15:48,044 [john] Yeah. 00:15:48,044 --> 00:15:52,024 [eric] And I think the less subjective it is, the more I would agree with you. 00:15:52,024 --> 00:15:52,664 [john] Yeah. 00:15:52,664 --> 00:15:53,574 [eric] So for example, 00:15:55,544 --> 00:15:59,304 [eric] data analysis, you know, code generation. 00:16:00,384 --> 00:16:02,214 [john] Mm-hmm. 00:16:02,214 --> 00:16:04,634 [eric] Um, I think a lot of... One, one 00:16:05,744 --> 00:16:07,404 [eric] heuristic I've used for this 00:16:08,544 --> 00:16:11,124 [eric] in my own thinking that's been helpful is, 00:16:13,824 --> 00:16:19,404 [eric] if what is being produced is a means to another end, 00:16:21,304 --> 00:16:28,224 [eric] then I, I would say generally I agree that the average is going up. 00:16:28,224 --> 00:16:29,143 [john] Mm-hmm. 00:16:29,143 --> 00:16:30,263 [eric] Right? Because 00:16:31,744 --> 00:16:41,504 [eric] if you are writing code, if you are performing a data analysis, if you are proofreading a document for spelling- 00:16:41,504 --> 00:16:41,684 [john] Sure 00:16:41,684 --> 00:16:42,244 [eric] ... right? 00:16:42,244 --> 00:16:43,264 [john] Mm-hmm. 00:16:43,264 --> 00:16:43,783 [eric] LLMs- 00:16:43,784 --> 00:16:44,384 [john] Or grammar 00:16:44,384 --> 00:16:47,814 [eric] ... or, or grammar. Grammar starts to cross the line a little bit, be- 00:16:47,814 --> 00:16:47,814 [john] Sure 00:16:47,814 --> 00:16:48,764 [eric] ... which we can talk about. 00:16:48,764 --> 00:16:50,403 [john] Yeah, 'cause there's some subjectivity in it. 00:16:50,404 --> 00:16:50,924 [eric] Um, 00:16:52,124 --> 00:16:56,504 [eric] but like it's, you know, an LLM is way better at proofreading for spelling than I am. 00:16:56,504 --> 00:16:56,914 [john] For sure. 00:16:56,914 --> 00:16:58,474 [eric] 'Cause it doesn't get fatigued. 00:16:58,474 --> 00:16:58,484 [john] Right. 00:16:58,484 --> 00:16:59,984 [eric] It's, it has a perfect dictionary. 00:16:59,984 --> 00:17:00,344 [john] Right. 00:17:00,344 --> 00:17:01,704 [eric] It can do it faster. 00:17:01,704 --> 00:17:02,324 [john] Mm-hmm. 00:17:02,324 --> 00:17:09,664 [eric] Right? Um, you can write tests for all of the, um, you know, all of the code that you're putting in production- 00:17:09,664 --> 00:17:09,674 [john] Yeah, right 00:17:09,674 --> 00:17:10,664 [eric] ... for a data analysis. 00:17:10,664 --> 00:17:11,454 [john] Right. 00:17:11,454 --> 00:17:20,544 [eric] Right? Um, and it can run those tests, and so it will be less technically error-prone than a human in those particular areas. 00:17:20,544 --> 00:17:21,663 [john] Right. 00:17:21,664 --> 00:17:29,144 [eric] I think where it gets very tricky, and circling back to the AI slop question, is when the output is the product itself. And this is- 00:17:29,144 --> 00:17:29,153 [john] Right 00:17:29,153 --> 00:17:34,144 [eric] ... squarely in the, in the arena of the challenge that I face every day- 00:17:34,144 --> 00:17:34,424 [john] Right 00:17:34,424 --> 00:17:36,414 [eric] ... working on a team of writers who are- 00:17:36,414 --> 00:17:36,414 [john] Right 00:17:36,414 --> 00:17:40,124 [eric] ... y- you know, learning to use AI in this new world that we live in, right? 00:17:40,124 --> 00:17:40,343 [john] Right. 00:17:40,344 --> 00:17:45,624 [eric] We're not generating code that ultimately becomes a product that someone uses- 00:17:45,624 --> 00:17:46,184 [john] Right 00:17:46,184 --> 00:17:54,804 [eric] ... in a different interface or... Th- they consume, they're not consuming the code, right? They're consuming the product in a different way, through a different surface area. 00:17:55,844 --> 00:18:07,044 [eric] And so when the, when the output from AI is the product, that's where it gets really tricky. Because as a human you're giving it your input, and it is generating the output. 00:18:07,044 --> 00:18:07,484 [john] Right. 00:18:07,484 --> 00:18:09,884 [eric] Right? And that is what someone is consuming. 00:18:09,884 --> 00:18:10,254 [john] Yeah. 00:18:10,254 --> 00:18:23,232 [eric] And so that's kind of why I say if you have two people doing the same task- They're sort of like giving their human input, and the output will be different, but it's also creating almost more conformity in the way that that happens. 00:18:23,232 --> 00:18:23,252 [john] Right. 00:18:23,252 --> 00:18:31,332 [eric] Which is tricky. And so I definitely sense the r- like reluctant nature of people to just outsource, right? Because- 00:18:31,332 --> 00:18:31,792 [john] Right 00:18:31,792 --> 00:18:34,032 [eric] ... mileage varies a ton on what you get on the other end. 00:18:34,032 --> 00:18:51,492 [john] Well, and it depends on how, how well can you define the output that you want. And in two ways, and, and this could be, let's call it like a visualization or some kind of interpretation or analysis of data. It could be writing. Like, is it something you can tightly template, think le- legal? 00:18:51,492 --> 00:18:52,012 [eric] Yep. 00:18:52,012 --> 00:18:52,632 [john] Um- 00:18:52,672 --> 00:18:54,092 [eric] Standardization for sure. 00:18:54,092 --> 00:18:55,992 [john] Yeah. Standardization, tightly template, 00:18:57,232 --> 00:19:00,102 [john] where, where literally the final result is 00:19:01,112 --> 00:19:03,892 [john] a document with blanks and guidance on what should be- 00:19:03,892 --> 00:19:03,912 [eric] Mm-hmm 00:19:03,912 --> 00:19:04,832 [john] ... in each blank. 00:19:04,832 --> 00:19:04,841 [eric] Mm-hmm. 00:19:04,841 --> 00:19:11,172 [john] And that's not just legal. There's other wa- there's other things that, like, I think would be less intuitive that you could do in that way where you'd get really good results. 00:19:11,172 --> 00:19:11,872 [eric] Yep. 00:19:11,872 --> 00:19:24,931 [john] Um, but it, but if kind of the, kind of part of the point of it is creativity and is, you know, pulling for the right analogies and explaining and stuff, I, yeah, I, I think that's a lot harder to get good results. 00:19:24,932 --> 00:19:31,752 [eric] We have a, we have a mutual friend who has been extremely successful in real estate, right? 00:19:31,752 --> 00:19:32,212 [john] Mm-hmm. 00:19:32,212 --> 00:19:37,402 [eric] And, um, you know, they're less day-to-day now, but are still, like, a very active- 00:19:37,402 --> 00:19:37,402 [john] Mm-hmm 00:19:37,402 --> 00:19:38,552 [eric] ... part of the firm that they built. 00:19:39,572 --> 00:19:43,972 [eric] And [chuckles] we actually talked to them about b- building an agent, uh- 00:19:43,972 --> 00:19:44,512 [john] Yeah. Yeah, yeah 00:19:44,512 --> 00:19:46,192 [eric] ... to analyze contracts, right? 00:19:46,192 --> 00:19:46,352 [john] Mm-hmm. 00:19:46,352 --> 00:19:56,552 [eric] And, like, one of the promises there from this person's perspective is that they, there is enough technical structure- 00:19:56,552 --> 00:19:56,812 [john] Hmm 00:19:56,812 --> 00:19:58,412 [eric] ... to a real estate contract 00:19:59,712 --> 00:20:05,472 [eric] as represented by, or as, like, manifestations of actual law, right? 00:20:05,472 --> 00:20:06,692 [john] Yeah, yeah. 00:20:06,692 --> 00:20:09,571 [eric] Um, of contractual law- 00:20:09,572 --> 00:20:09,912 [john] Mm-hmm 00:20:09,912 --> 00:20:13,992 [eric] ... rel- related to real estate transactions. Now, what's interesting is that you can, you can, 00:20:15,152 --> 00:20:17,332 [eric] you can create a very creative 00:20:18,652 --> 00:20:19,992 [eric] contract that is fully- 00:20:19,992 --> 00:20:20,852 [john] Yeah 00:20:20,852 --> 00:20:21,432 [eric] ... legal, right? 00:20:21,432 --> 00:20:21,931 [john] Right. Right. 00:20:21,931 --> 00:20:22,492 [eric] Um- 00:20:22,492 --> 00:20:22,632 [john] Right 00:20:22,632 --> 00:20:25,252 [eric] ... which this guy is incredibly good at, actually. 00:20:25,252 --> 00:20:25,372 [john] Mm-hmm. 00:20:25,372 --> 00:20:27,952 [eric] So one of the skillsets is sort of thinking out of the box. But- 00:20:27,952 --> 00:20:28,072 [john] Right 00:20:28,072 --> 00:20:28,072 [eric] ... 00:20:29,352 --> 00:20:39,932 [eric] I think that's a great example of where you can get a lot of leverage on, like, the technical side of it and sort of automate the pieces that, you know, were- 00:20:39,932 --> 00:20:40,332 [john] Right 00:20:40,332 --> 00:20:42,612 [eric] ... pretty time-consuming and, and required a lot of- 00:20:42,612 --> 00:20:42,702 [john] Right 00:20:42,702 --> 00:20:49,661 [eric] ... accuracy previously, but still have room for, you know, sort of creativity. So that's, it's a kind of an interesting one- 00:20:49,661 --> 00:20:49,661 [john] Yeah 00:20:49,661 --> 00:20:51,492 [eric] ... that's, like, you know, in the middle. 00:20:51,492 --> 00:20:53,512 [john] It is. All right, our last one. 00:20:53,512 --> 00:20:53,932 [eric] Last one. 00:20:53,932 --> 00:20:58,892 [john] Ready? This is the, um, the marathoner. The mental marathoner. 00:20:58,892 --> 00:21:01,452 [eric] Mental marathoner. Ooh, I l- I already like the sound of this. 00:21:01,452 --> 00:21:05,702 [john] [laughs] All right. So this is a person that I, I think... I don't think anybody... 00:21:07,072 --> 00:21:14,852 [john] Obviously n- nobody perfectly fits in these categories. Um, uh, and following up on our last episode about AI burnout, I think this [chuckles] is the group- 00:21:14,852 --> 00:21:15,072 [eric] Mm-hmm 00:21:15,072 --> 00:21:28,792 [john] ... that is most likely to get burnout. Um, but the idea is, like, they're in the driver's seat the whole time. They're able to accomplish more than they ever had, and they just have this kind of natural ability to, 00:21:30,012 --> 00:21:42,572 [john] to drive [chuckles] for, like, long periods of time without kind of getting sucked in, pulled down to the average or pulled down to, like, the, the, um, output that, that the model wants- 00:21:42,572 --> 00:21:42,732 [eric] Yep 00:21:42,732 --> 00:21:43,172 [john] ... per se. 00:21:45,372 --> 00:21:47,912 [eric] I have a, I have a spicy take on this one. 00:21:47,912 --> 00:21:48,132 [john] Okay. 00:21:49,702 --> 00:21:52,712 [eric] I've written about this before on my blog. So 00:21:54,812 --> 00:22:01,452 [eric] I think it's, I think it's easy to believe 00:22:02,592 --> 00:22:10,471 [eric] that... And, and I would argue even it, it may be one of the more natural things to believe about AI. 00:22:10,532 --> 00:22:10,802 [john] Right. 00:22:10,802 --> 00:22:11,332 [eric] And it's this: 00:22:12,852 --> 00:22:14,012 [eric] that it will 00:22:15,372 --> 00:22:25,692 [eric] generally, it, it will create a rising tide that sort of lifts, lifts the level of knowledge work- 00:22:25,692 --> 00:22:25,732 [john] Mm-hmm 00:22:25,732 --> 00:22:28,352 [eric] ... right, in quality and quantity- 00:22:28,352 --> 00:22:28,612 [john] Mm-hmm 00:22:28,612 --> 00:22:31,152 [eric] ... ubiquitously across the board. 00:22:31,152 --> 00:22:31,161 [john] Yeah. 00:22:31,161 --> 00:22:32,632 [eric] Which I don't believe to be true. 00:22:35,132 --> 00:22:37,672 [eric] I, what I see 00:22:38,792 --> 00:22:50,632 [eric] is that there is actually a greater divide between people who already had an extremely strong preexisting skill set- 00:22:50,632 --> 00:22:50,852 [john] Mm-hmm 00:22:50,852 --> 00:22:50,852 [eric] ... 00:22:52,252 --> 00:22:59,432 [eric] most often combined with a strong, like, written or verbal skill set- 00:22:59,432 --> 00:23:00,392 [john] Yep 00:23:00,392 --> 00:23:06,952 [eric] ... for whom AI is an unbelievable multiplier. And when you combine that with a mental marathoner- 00:23:06,952 --> 00:23:07,372 [john] Right 00:23:07,372 --> 00:23:09,952 [eric] ... who has a lot of endurance- 00:23:09,952 --> 00:23:10,292 [john] Mm-hmm 00:23:10,292 --> 00:23:15,792 [eric] ... to go with that, you have these people who become extraordinarily productive. 00:23:15,792 --> 00:23:16,372 [john] Mm-hmm. 00:23:16,372 --> 00:23:16,852 [eric] Um, 00:23:19,812 --> 00:23:29,972 [eric] adding, like, giving an average person AI doesn't even come close to producing that result, and one of the interesting reasons for this, I believe, actually, 00:23:31,172 --> 00:23:31,492 [eric] is 00:23:32,812 --> 00:23:40,672 [eric] that because the way that you interface with AI is through written or spoken word- 00:23:40,672 --> 00:23:40,952 [john] Mm-hmm 00:23:40,952 --> 00:23:51,992 [eric] ... people who are, who have a core skill set of being articulate are naturally predisposed to being power users of AI and getting far more leverage. 00:23:51,992 --> 00:23:52,051 [john] Yeah. 00:23:52,052 --> 00:24:00,132 [eric] One of the scary things is that the literacy rate in our country is extremely low, and it's not getting better. 00:24:01,232 --> 00:24:01,922 [eric] And so you have- 00:24:01,922 --> 00:24:03,152 [john] What, what do you mean by that? 00:24:03,152 --> 00:24:13,832 [eric] So if you look at, um... So reading comprehension. So I, I actually looked up the research on this and, like, downloaded PDFs from, you know, organizations that- 00:24:13,832 --> 00:24:14,132 [john] Yeah 00:24:14,132 --> 00:24:15,392 [eric] ... you know, do peer-reviewed studies on this. 00:24:15,392 --> 00:24:15,672 [john] Okay, cool. 00:24:16,976 --> 00:24:20,136 [eric] It's hard to track, like, ri- I don't think there's a, a- 00:24:20,136 --> 00:24:20,536 [john] Right 00:24:20,536 --> 00:24:22,646 [eric] ... measure for writing skills specifically- 00:24:22,646 --> 00:24:22,646 [john] I- 00:24:22,646 --> 00:24:23,036 [eric] ... at least that I could find 00:24:23,036 --> 00:24:28,176 [john] ... my best, my best proxy, like some- like verbal SAT scores or something. But- 00:24:28,176 --> 00:24:31,576 [eric] Reading comprehension generally, that was what I found. 00:24:31,576 --> 00:24:32,195 [john] Right. Okay. 00:24:32,196 --> 00:24:36,086 [eric] Um, you know, this was not a full academic... You know, I didn't- 00:24:36,086 --> 00:24:36,086 [john] [laughs] 00:24:36,086 --> 00:24:37,586 [eric] ... deep dive into a full academic study. 00:24:37,586 --> 00:24:37,616 [john] Sure. Right. 00:24:37,616 --> 00:24:40,716 [eric] But these are sort of peer-reviewed research. I'll link to the blog in the show notes- 00:24:40,716 --> 00:24:40,746 [john] Yeah 00:24:40,746 --> 00:24:53,616 [eric] ... 'cause there's, I link to the actual study. But reading comprehension is generally viewed as a proxy for sort of w- we would say, like, literacy. 00:24:53,616 --> 00:24:53,776 [john] Mm-hmm. 00:24:53,776 --> 00:24:59,776 [eric] Right? So competency in, um, like literary skills- 00:24:59,776 --> 00:24:59,786 [john] Yeah 00:24:59,786 --> 00:25:02,076 [eric] ... which would be reading comprehension and writing. 00:25:02,076 --> 00:25:02,296 [john] Mm-hmm. 00:25:03,436 --> 00:25:14,196 [eric] And the more than half of the country are less than 2.5 on a five-point scale, so like 50%. 00:25:14,196 --> 00:25:15,366 [john] Okay. Mm-hmm. 00:25:15,366 --> 00:25:15,936 [eric] Right? Um, 00:25:17,596 --> 00:25:22,036 [eric] and so anyways, all that to say, like, what's interesting is that a lot of people 00:25:23,536 --> 00:25:28,036 [eric] unfortunately lack the core skill set that they need to even- 00:25:28,096 --> 00:25:28,616 [john] Hmm 00:25:28,616 --> 00:25:33,716 [eric] ... at an advanced level instruct AI in a, in a super articulate way, right? 00:25:33,716 --> 00:25:34,436 [john] Right. 00:25:34,436 --> 00:25:42,266 [eric] Now, over time, will, will they figure out affordances for that, you know, to, to sort of translate human intent? 00:25:42,266 --> 00:25:42,316 [john] Mm-hmm. 00:25:42,316 --> 00:25:42,816 [eric] I don't know. 00:25:43,876 --> 00:25:50,556 [eric] But all that to say, I think the mental marathoner is actually more rare. I think that's, like, very rare- 00:25:50,556 --> 00:25:50,916 [john] Right 00:25:50,916 --> 00:25:54,156 [eric] ... uh, in terms of the distribution of these three archetypes. 00:25:54,156 --> 00:26:00,876 [john] Well, it's an interesting framing because it is essentially a vast amount of reading comprehension. [laughs] 00:26:00,876 --> 00:26:02,446 [eric] Yeah. It is. It is. 00:26:02,446 --> 00:26:03,456 [john] It's really interesting framing. 00:26:03,456 --> 00:26:11,176 [eric] Yeah, yeah, yeah. But the problem, as we talked about last time, is that the mental marathoners are at a huge risk of burning out because- 00:26:11,176 --> 00:26:11,826 [john] Yeah 00:26:11,826 --> 00:26:15,876 [eric] ... the leverage that they can get in terms of output is material. 00:26:15,876 --> 00:26:16,376 [john] Right. 00:26:16,376 --> 00:26:19,126 [eric] Uh, but that also increases expectations, right? 00:26:19,126 --> 00:26:19,516 [john] Right. Yeah. 00:26:19,516 --> 00:26:24,396 [eric] Both for yourself, like from your employer, in whatever context you're in. Um, 00:26:26,436 --> 00:26:28,546 [eric] the other spicy take... I'm just full of spicy takes today. 00:26:28,546 --> 00:26:30,416 [john] Yeah. That's the third one today. 00:26:30,416 --> 00:26:40,716 [eric] I know. Okay, so last spicy take on this. I think for the mental marathoners, the risk... And I'm kinda s- I'm kinda speaking in many ways to you and I here- 00:26:40,716 --> 00:26:40,986 [john] Mm-hmm 00:26:40,986 --> 00:26:44,736 [eric] ... because I, I would put us in the mental marathoner category. 00:26:44,736 --> 00:26:45,516 [john] On a good day. 00:26:45,516 --> 00:26:46,176 [eric] On a good day. 00:26:46,176 --> 00:26:46,276 [john] [laughs] 00:26:46,276 --> 00:26:47,496 [eric] Yes. On a good day. 00:26:47,496 --> 00:26:50,196 [john] On a, on a, on a bad day, it's, it, it is easy to get- 00:26:50,196 --> 00:26:50,916 [eric] Oh, for sure 00:26:50,916 --> 00:26:53,916 [john] ... sucked into... Yeah. But- 00:26:53,916 --> 00:26:54,175 [eric] For sure 00:26:54,175 --> 00:26:54,916 [john] ... we aspire. 00:26:54,916 --> 00:26:56,506 [eric] We, we aspire to be- 00:26:56,506 --> 00:26:56,606 [john] [laughs] 00:26:56,606 --> 00:26:58,096 [eric] ... considered mental marathoners- 00:26:58,096 --> 00:26:58,216 [john] Right 00:26:58,216 --> 00:26:59,106 [eric] ... I guess I would say. 00:26:59,106 --> 00:26:59,176 [john] Right. 00:26:59,176 --> 00:27:02,136 [eric] Maybe that was a little bit too presumptuous. [laughs] 00:27:02,136 --> 00:27:02,636 [john] Well. 00:27:02,636 --> 00:27:03,865 [eric] I've seen you use AI, though- 00:27:03,865 --> 00:27:03,865 [john] Right, right 00:27:03,865 --> 00:27:10,916 [eric] ... and you're definitely a mental marathoner. [laughs] But I think that the risk for, for the mental marathoner 00:27:12,096 --> 00:27:16,816 [eric] is that because you run at cognitive red line- 00:27:16,816 --> 00:27:17,096 [john] Mm-hmm 00:27:17,096 --> 00:27:20,836 [eric] ... it is so tempting to slip into the other archetypes- 00:27:20,836 --> 00:27:20,846 [john] Yeah 00:27:20,846 --> 00:27:24,326 [eric] ... which actually begins to erode y- the, the core skill set- 00:27:24,326 --> 00:27:24,365 [john] The... Yeah, yeah 00:27:24,365 --> 00:27:25,996 [eric] ... that allowed you to be a mental- 00:27:25,996 --> 00:27:26,046 [john] Just, yeah, yeah 00:27:26,046 --> 00:27:27,996 [eric] ... marathoner in the first place, right? 00:27:27,996 --> 00:27:28,636 [john] Yeah. 00:27:28,636 --> 00:27:28,936 [eric] So 00:27:30,016 --> 00:27:34,616 [eric] I guess probably the conclusion there is that balance and restraint- 00:27:34,616 --> 00:27:34,636 [john] Right 00:27:34,636 --> 00:27:38,276 [eric] ... and, like, boundaries, uh, are probably a good idea- 00:27:38,276 --> 00:27:38,286 [john] Yeah 00:27:38,286 --> 00:27:39,456 [eric] ... for the mental marathoner. 00:27:39,456 --> 00:27:43,836 [john] Yeah. The biggest thing that I've... And I keep coming back to this, and I'll probably say it every single show, 00:27:44,956 --> 00:27:54,456 [john] that I've found to be useful, is finding, uh, is not getting sucked into the, um, chat dialogue cycle. 00:27:54,456 --> 00:27:55,856 [eric] Yes. Yep. 00:27:55,856 --> 00:28:08,016 [john] I just can't... I don't think I can overemphasize how important it is to abstract, to, to spend extra time up front on defining exactly what you want and how to validate it's what you want, number one. And then two, 00:28:09,276 --> 00:28:22,596 [john] launching a long-term process through whatever tools you're using, um, to do whatever you want, taking space, and then evaluating the result as if it came from one of your employees. 00:28:22,596 --> 00:28:23,476 [eric] Yep. 00:28:23,476 --> 00:28:38,476 [john] Like, that is what's gonna work, and it does two things for you. One, I think you, you can get better results in general. But two, then you're set up to iterate and improving on the context and the prompting and, like, all of the components that go into the AI 00:28:39,716 --> 00:28:42,396 [john] and compounding over time. Whereas if you just work 00:28:43,716 --> 00:28:55,336 [john] synchronous with it in a session, get the thing, deliver the thing, that thing disappears, and you go do it again and again and again. Like, that, I mean, that's, that's really exhausting, and it's not compounding. 00:28:55,336 --> 00:28:57,346 [eric] Yep. I agree. I agree. 00:28:57,346 --> 00:29:21,796 [john] And, and there's still a really... And I think I see some of the tooling changing around, but, but all of the tooling right now is, is essentially set up for, "Hey, what do you wanna get done right now?" And then you get sucked into doing the thing and, um, and you go back and forth and, and you're like, "Oh, like, I think I got a thing," and then that thing disappears and you go back a month later to do that same thing and you're like, "I don't remember where that was," and you start over. I mean, it's ki- you're kinda set up for that right now. 00:29:21,796 --> 00:29:22,456 [eric] Yeah, for sure. 00:29:22,456 --> 00:29:22,866 [john] Um. 00:29:22,866 --> 00:29:29,376 [eric] I think another, another really important thing to remember is that there, you know, the, um, 00:29:32,516 --> 00:29:35,795 [eric] it- user engagement rates are a real thing that real people- 00:29:35,796 --> 00:29:35,896 [john] Yeah 00:29:35,896 --> 00:29:36,756 [eric] ... are accountable for. 00:29:36,756 --> 00:29:43,126 [john] Mm-hmm. Well, the people... Yeah, I mean, that's, that's the type of information you report to your investors, for example. 00:29:43,126 --> 00:29:51,816 [eric] Yeah. And, and I don't say that in any way as a, you know, like conspiracy theory, you know- 00:29:51,816 --> 00:29:52,116 [john] Right 00:29:52,116 --> 00:30:02,456 [eric] ... anything of that nature, but it's just the way that, it's just the way that these businesses are set up, right? And I think it's healthy to acknowledge- 00:30:02,456 --> 00:30:02,506 [john] Mm-hmm 00:30:02,506 --> 00:30:06,106 [eric] ... that that's not always necessarily the best outcome, right? And I think- 00:30:06,106 --> 00:30:06,106 [john] Right 00:30:06,106 --> 00:30:13,236 [eric] ... you know, if we look at things like social media, right, and sort of the, like, some of the psychological problems- 00:30:13,236 --> 00:30:13,246 [john] Right 00:30:13,246 --> 00:30:16,716 [eric] ... that are associated with that, if, you know, if sort of taken to the extreme. 00:30:17,776 --> 00:30:20,116 [eric] You know, an algorithm that's designed to keep your attention, 00:30:21,256 --> 00:30:23,926 [eric] uh, will keep your attention, you know, if you let it- 00:30:23,926 --> 00:30:23,926 [john] Yeah 00:30:23,926 --> 00:30:40,356 [eric] ... over, you know, over time. Um, and so one of my takeaways from what you said is, you know, show restraint and, like, use the tool with intention, um, as opposed to just following the path of least resistance. 00:30:40,356 --> 00:30:41,436 [john] Yeah. 00:30:41,436 --> 00:30:44,736 [eric] All right. Well, thanks for joining us on the Token Intelligence Show, and we'll catch you on the next one. 00:30:49,096 --> 00:30:53,786 [eric] [outro music]