00:17:39 · Guillaume Moubeche, George Bonaci
[00:17:39] S1: Can you explain what Jevan's paradox is? [00:17:41] S0: Yes. Basically, if something becomes very, very cheap, it doesn't get used less, it gets used more. Like the best example, think the classic example, it's like, oh, you add more lanes to a freeway, that doesn't decrease traffic. That that increases it because all of a sudden more people are gonna start driving. And I think that's very true obviously like with work. Like work used to be a commodity that could only be executed by like human beings or maybe like pre industrial revolution like by animals. And now work can be done by like by agents. And and not not just like manual work, I should say, but like actual, like, intellectual intelligence can now be done by machines. And so when that gets commoditized, it's not like people lose their jobs. It's actually probably like, we probably need more people. We need to hire more people. It's very similar to the the freeway example.
00:18:21 · Guillaume Moubeche
[00:18:21] S1: What's your feeling like personally, you know, because to me, when we started to see, you know, what we could do with, I don't know, like OpenAI, first we chat GPT, then Cload, etcetera, like early on, you know, the thing was, okay, everything's gonna be automated. We're gonna ultimately like work less. However, you can do a lot more, and you kind of feel like you have superpowers, and eventually, you work on so many more stuff yourself that you start like changing context and topic, like very quickly. Like typically like myself, when I work, I always have like at least, I don't know, four or five, like a chat GPT conversation at the same time. So when you think about something, I'm doing something else, or, like, having agents running in the background. So... And eventually, like, you feel like you're switching context all the time and your brain feels like it's working, like, 10 times harder than back in the days. Like, what's your feeling like personally? And do you also see like a similar impact?
00:19:20 · George Bonaci
[00:19:20] S0: I feel exactly the same way. I I was on a con... Or I was on a phone call with a friend yesterday who's who's doing this. And he was like, man, how do you get more than like seven or eight concurrent projects at once that you're managing? Like, how do you keep that straight in your head? Do you have a tool or something? I was like, like, no. That's the same problem. It's like, how how do... Can you really multitask on seven things at once? Like, all the... To your point, it has given people superpowers, but really I think this is a symptom or maybe an indicator of how early we are to like this paradigm shift. I can't imagine one or two years from now, we're going to be having the same conversation. It's probably... We're going to either have like some crazy like orchestrators or the models are going to get so good, the harnesses are gonna get so good that they can kind of orchestrate themselves. Like, I don't actually know what shape the solution is going to take, but this is I think a temporary problem because... Actually it's a temporary problem and probably a symptom of the much larger thing, which is this technology is is so new that we have not captured even like a a small percentage of the potential impact that it's going to have. I saw a great tweet. I can't take credit for this, but I saw a great tweet somewhere comparing this to the the .the.com boom. I think like during the .com boom, everyone knew the internet was gonna be huge. Like, no one debated that, but they didn't know in what form it was going to be huge. I think like people thought that everyone was gonna have their own personal website and communicate through their own personal website and like turned out to not be true. Like aggregators like really took over. I think it's probably similar here. Like everyone knows AI is gonna be huge. It already is. But in what final form it takes and, like, what is the the steady state we approach and how people yet leverage it and what impact it has for companies and individuals, I think is is very TBD.
00:20:54 · Guillaume Moubeche
[00:20:54] S1: Yeah. Yeah. I agree. Same. Like, as you mentioned, like, with the Internet era, like, the social network and all this this kind of, like, social apps, a lot of people are trying to understand, like, what is gonna be happening with AI in a similar way. So it's gonna be interesting to see. I wanna switch a little bit on the on the topic because we're talking about, like, AI a lot, and I wanna go, you know, like, towards hiring because I think, like, hiring has changed dramatically with AI. Like, I personally hired, like, hundreds of people, like, in the last years, but I feel like the level, especially, you know, for business cases, etcetera, like a presentation, whenever you want people to work on something, because everyone's using AI nowadays, it's very hard to actually, you know, like see, you know, the the top performing one, or actually, like, you can see, like, some people who don't make any effort, and it's, like, pure AI slop, etcetera. But I do feel like the bar is is raising whenever, but I also feel like people lack a lot of critical thinking. So what's your view when it comes to hiring? And do you also struggle to find people who use AI and also understand what they're doing and not just give you like a 20 pages Notion doc that looks great on paper, but that's actually meaningless and not helpful at all?
00:22:16 · George Bonaci
[00:22:16] S0: It's a great question. So a couple answers, but I mean, maybe maybe to answer on like the trend of AI slop. Shorter docs are better. Like, one... I think very rarely do you need anything more than, a one pager. If you can't fit it on a one pager, then you probably don't have good clarity of thought. And I think the tendency we're kind of seeing is, like, what is AI? AI is not going to increase the information density of, like, what you're prompting it to do. So if I have some thoughts in my head and I prompt the AI to write a doc on it, it could write it in one page or it could write it in 40 pages, but that's not going to change like the actual amount of useful information in it. It's just changing the information density. And so like in general, I actually think it's a pretty good rule of thumb where you you can use AI as a sparring... When it comes to writing, you can use AI as a sparring partner. You can use it to collect information. You can use it to do research. You could use it to even like set up experiments, but you probably shouldn't use it to like write for you if the audience is a human being. It can maybe help you edit, but like I wouldn't do that. Number one, I think it breaks the social contract between human beings as well. Like, I haven't seen many people talk about this, but I do think there is a bit of a social contract where if like a human being is spending time on something, then another human being feels obligated to at least give it some attention. And if you're not spending time on it, then like, why would you feel obligated to give it any attention? That is true like with total strangers receiving an email. Like if it's just like a... Like this is pre AI as well. It's the same problem. It's like, this is a copy of... Blatantly copy and pasted email. Like, you screwed up the HubSpot token and it says like, hi, first name. It's like, I'm not going to read that. It's the exact same thing. It's just really, really amplified with like AI slops. So anyway, that's that's my answer to that. To answer like your first part though on hiring, I I do think that's the most important part. You need to be hiring people that frankly... I think the laziest heuristic are are they builders? Because if you're able to build something, can think from first principles, you can understand the opportunities, you can get things done.
00:23:58 · George Bonaci, Guillaume Moubeche
Like, all of those are really, really important. The... By far the best interviews that that I have are ones where we're talking about something or some situational question and and the person goes like, let let let me just show you. Can I show you? Can I just like screen share? It's like, that's that's funny right there. [00:24:12] S1: And then [00:24:12] S0: you can go into such level of details. Like, yes, please show me. Like, let's get specific. Those type of people generally do very, very well in general. But I think especially with AI, because they're going to tinker, they're going to understand the potential and the limitations of the technology. Like they themselves are critical thinkers, so they're going to have very high information density in how they communicate. So they're probably less likely to just spin up a 40 page Notion doc with a one sentence prompt, but at least that's that's how I would think about it.