The Great AI Assistant Showdown: Alex Valente, Co-Founder & Co-CEO Quarterzip AI

0:00Alex Valente: You're listening to a day one dot FM show. Could AI do my job? Like, I can assure you, like, a frontier lab's already tested whether it can be a job. It's already come up with the answer yes. And the only reason they haven't pursued doing your job is because there's bigger markets for them to pursue right now, which is, like, potentially the saddest reason why a person doesn't compete against you. If we aren't able to kind of get the productivity benefits of all the things that we see in front of us, things happen slower. It's that we become irrelevant. very quickly. If you work any job, if you're responsible for anything, you're going to be responsible for your factory.

0:36Alex Valente: Have you ever met a factory owner? Like, the person who makes doors in their factory, they didn't do a degree in doors. Like, they don't have a complex understanding of the history of the door. They just make doors.

0:49Georgie Healy: One pushback on the factory, though, is I think of factory as a negative connotation.

0:54Alex Valente: I always think of it like, maybe this is my Italian coming out, but I always think of it like the Ferrari factory. It's like, that's a good place. You need to be in the Ferrari.

1:02Georgie Healy: We're back on board. Right. Which of the big labs are you betting on? I think I know the answer to this.

1:07Alex Valente: I'm probably betting on Anthropik.

1:09Georgie Healy: One tech book everyone should read.

1:11Alex Valente: Oh, let's go with Anti-Fragile. We spoke about it in the air. Got to recommend it.

1:19Georgie Healy: Welcome to In the Blink of AI. I'm Georgie Healy. We have got our guest back for the third time. And there is a reason behind that. Alex Valente is the kind of AI expert you really do want on speed dial. As a host, I know for a fact now that when he comes on the show, he predicts things and mentions things that I will never have heard of. And then they play out in the months to come. I know this because I look at the transcripts, guys. He predicted DeepSeek blowing up before it did. He talked about Waymo's two years before I ever sat in one. He's a serial entrepreneur. He's from Australia. He's based in San Francisco.

2:01Georgie Healy: And he really has a lot of nuance and insight. When I threw three big headlines at him this week, you'll see that there's a lot that you will gain from hearing from his perspective. He also did what my favorite part of the episode was, which is a personal AI agent showdown. There are so many personal AI agents. And if you're feeling overwhelmed, don't worry, we've got you. We analyzed Grok, Instinct, Muse, Dot, and Duet. We gave them our ratings. Sometimes we fought over our ratings and had a very different perspective, but we still agreed on the winner. That's on the show today, as well as many other predictions, hot takes, and basically just my favorite delightful AI expert that we sadly have lost to Silicon Valley.

2:56Georgie Healy: Thank you for listening to In the Blink of AI, and let's dive in.

3:00Alex Valente: You're listening to a day one.fm show.

3:05Georgie Healy: In the Blink of AI is proudly sponsored by Deal, the global HR and payroll platform that's behind Any expansion of companies like Revolut, Shopify, or fans of the show Eleven Labs who quadrupled their workforce in 54 countries. Build your global team with Deal.

3:29Georgie Healy: Alex Valente, welcome to In the Blink of AI, third time.

3:34Alex Valente: Third time. I feel welcomed already. It feels like home.

3:38Georgie Healy: Whenever I need to pull up predictions and insights, I've got all the transcripts from all the episodes and I'll just ask Claude sometimes, like, what were some really good hot takes? And you always come up, mate. You always come through.

3:53Alex Valente: I don't gamble and I don't participate in prediction markets. And what I do instead is I come on this show and I make predictions so that when I'm wrong, people can tell me that I'm an idiot.

4:03Georgie Healy: We've got timestamps. We've got, you know, facial recognition. Exactly.

4:09Alex Valente: For time immemorial, I'm sure it's never getting deleted off the internet.

4:13Georgie Healy: Nope. This is my legal property now. We'll get a few more. I'm actually doubling down. This episode is literally a treat for the listeners. I'm really grateful for you being so generous with your incredible brain, dialing in live from San Francisco. That is wild. Actually, I don't think the last two times that was the case.

4:34Alex Valente: No, this might be my first live cross from the San Francisco studio in the blink of AI. Hell yeah. You can report that the weather is exactly what you expect, which is it's the same weather every single day in San Francisco. It's foggy in the morning, it's become sunny in the middle of the day, and then it gets cold at nighttime. It's like that every single day, and it's never changing.

4:53Georgie Healy: I loved visiting San Francisco for the first time, and you told me about the Waymo pile-ups that happen on the streets. This is the kind of stuff you need a real founder on the ground to unpack for you, I've got to say.

5:08Alex Valente: I recently had my parents in town and obviously the first thing you do when you have your parents in San Francisco is you put them in a Waymo. They've got to see it. The Waymo had automatic doors and it gave my dad a fright. That was the thing that freaked him out the most was the fact that the doors opened automatically. Wait until you've been stuck in a Waymo traffic jam. That's truly frightening when they start reversing.

5:29Georgie Healy: For me, the thing that didn't make sense to me was how to know which Waymo was mine because I'm used to Ubers with upper plates and I'm like, oh, that's clearly my guy. And I'm like, how does it know it's me? And then the internets and the Bluetooths and the things. That was super rad. Mate, you know the format. We kick off the show with an AI hack of the week.

5:53Alex Valente: Yeah.

5:53Georgie Healy: What's your hack in our Year of the Lord 2026 AI agent?

5:58Alex Valente: Great question. look, I've developed a new thing. I think it's annoyed my whole team. I think it's annoyed my co-founder. I think it's annoyed my girlfriend at this point, which is that I now record myself giving a lecture in front of a whiteboard. And then I just send the whole thing to Claude and I'm like, turn this into a blog post and keep my tone of voice. And it'll take like the images off the whiteboard. It'll turn them into diagrams. It's like super strong at it. I find it like generally comprehends like what's going on the tone and the free like the flow of the conversation and it really does well to explain it as like something that's a sequential conversation and it just means that like i can take something that's in my head and turn it into something that's like super comprehensible on a sheet of paper like in the space of like 30 minutes i'm

6:43Georgie Healy: gonna sit down and just have a go i hear a lot of hacks completely brand new to me that is genius So just to make sure I understand, you record it on your camera, like a video, and then you upload that to, I guess, Claude? Yep. That's how you do it? Yep. I'm trying this, dude. Yeah, it's really good. Because I feel like Whisperflow annotated into Claude was okay. It's good.

7:09Alex Valente: Once you start on this journey of doing the hacks, you start hacking everything. It starts being hard to... The keyboard feels really painful. all of a sudden.

7:20Georgie Healy: The keyboard is so painful. January this year, I told you, I get my transcripts. I see when people say what I said at January this year, I feel like we're going to use the keyboard less. Yes. Stop clapping, Alex. We will ruin the audio. But last night I was doing the kids lunches for school the next day, getting ready for this interview, making sure I really understood the nuance between personal AI agents, verbal chat with Claude. I'm like, I'm not even that's going to whisper flow this one like even that's a bridge too far game i know same when

7:55Alex Valente: people are driving now like people might not listen to a podcast they sometimes have a conversation with claude and it's like it's not a terrible idea and i think like we're kind of now at a stage where like i'm having this conversation with my own team it's like we're looking at our company and we're going like hey like what does a company look like in the future where like every single thing is getting done either by an agent or with the assistance of agents like what what What is a firm? What does this look like? And I think what we've landed on that I found really helpful as a way of thinking about it that might be helpful to the listeners as well as like, well, if I'm running a factory and if you're running a factory and we're all running our factories, our firm is a factory of factories and we are all factory owners.

8:33Alex Valente: Um, and it's very empowering to our staff. It gives them this opportunity to look at their role and go like, Hey, I'm, I'm not like a person on a conveyor belt anymore. I'm not doing this like niche job. I'm actually like, I run this like factory and I'm the factory owner and I have inputs and I have outputs and I'm responsible for them. And, and I manage my factory as, as I need to. And using that language has just been, has just meant that our employees have been really excited to kind of go out and start building stuff themselves. And I mean, like we now have like this incredible proliferation of Claude skills.

9:08Alex Valente: And we had our designer build something called, she calls it the factory floor, which is a map of all these skills and how they all interlink with one another. And we have agents that will spin up and start running on different triggers, at different sequences. It just means that we can run our business in a much more AI native way and be much more effective. It's been most interesting watching non-technical staff really go through this because it's something previously that was maybe reserved to the engineering cohorts and it's now so accessible to our non-technical staff.

9:38Georgie Healy: Speaks to me on a deep level, Alex. The last time you've come on the show, the audience was really technical. They were already within our tech industry for the most part. And I see the stats. People that are traditionally non-technical are trying to grapple with AI and the headlines. They care and they're smart. They just say to me to build an AI agent. I don't know what that means. I own a factory and I have a factory floor. smart people can can get it get involved people get it because i think it's like software is hard

10:13Alex Valente: for people who are not in the software industry in the sense i was like we don't normally think of like these complex instructions and abstraction there's all this jargon when when you when you say to someone it's like hey you've got a factory it's like people can imagine that and sometimes it's like half the battle because it means like the number one blocker to ai diffusion is is actually work um workflow identification and when you see a person start to use tools and they start to identify how else they can use them, it's incredibly valuable.

10:36Georgie Healy: My one pushback on the factory, though, is I think of factory as a negative connotation. Interesting. Industrial revolution type stuff. Yeah. And people are a bit triggered by industrial revolution type stuff, right? Can it be like a kitten factory? No, that's bad. No, that's about true.

10:57Alex Valente: I always think of it like, maybe this is my Italian coming out, but I always think of it like as the Ferrari factory. It's like, that's a good place. You want to be in the Fry.

11:05Georgie Healy: We're a backbone board. Right.

11:06Alex Valente: Yeah, exactly. You want to be in the Fry factory. Yeah. And it's like nine, like at, you know, at 1pm everyone breaks for lunch and it's really delicious lunch. It's served to everyone. That's what I kind of.

11:17Georgie Healy: Before we unpack the headlines, I learned something quite interesting. When I worked at McLaren, the production factory floor is polished white paint. Now, if you can imagine, it's an automotive industry. having white painted floors is like next level, like aesthetic. But apparently Steve Jobs did this in Apple, like at Apple as well. He's like, I'm going to have polished white floors. And I find that, think of that as a factory, guys. It's very aesthetic. It's a nice factory that you want to work in.

11:48Alex Valente: That's exactly right. I think like the, this is the funny thing. It's like the, the dirty factory with all the smoke coming out of it. It's actually turned, it's not a very profitable factory. Like that's a bad factory. The really good factories are like super schmick. You want to hang out in those ones. And it's good. I want to be the only one. Exactly. Exactly.

12:07Georgie Healy: Nate, I'm excited to get you on the show. You're really good at unpacking AI headlines in a way that makes sense. I've got three for you. Are you ready to go? Let's do it. Number one is I like to call this agents behaving badly. We've got a lot of headlines around the agents and they're breaking into places so they're not supposed to be. Recently, one hacker used AI agents to break into more than 1 ,200 websites. Here in Australia, OpenAI's agent escaped and got into Medicare. This is headline news. This is day after day. The CloudFare CEO said that this is going to continue happening for the next two years, and it's going to feel like the sky is falling.

12:52Georgie Healy: Whoa, that's hectic. That's crazy. First of all, why are agents so dangerous compared to like the chatbots that we've been using for the last two years, just in basic terms?

13:03Alex Valente: Yeah, I think they have a lot more agency and access than the previous generation of stuff. They're now long running, so we find that they are able to do very complex tasks over an extended period of time. And they have access to a lot more compute. The easiest way to think about it is not only does it have the keys to the car, it's also got a very high-powered car. and not only is it a high-powered car, they're driving it for a long time. So you're more likely to have a car accident in this scenario than in the previous one. I'm looking at kind of a lot of these stories that are coming out at the moment and I kind of see a lot of the conversation kind of split into two respective positions.

13:42Alex Valente: Like there's people who maybe want to regulate and they see what's come up and they go, we need to get people to stop this. And like, what does that mean? It's like, Hey, let's, let's just get it, you know, uh, uh, let's get in some, some of the lawyers and let's get the lawyers to just say, we're done with this. We, um, we don't want to do this anymore. Um, and we regulate AI as much as we can. And then I think there's the other side of this, which is like, we need to, to build our way out of this. And this kind of comes down to this, this, this, um, idea about AI diffusion. Like, Hey, what is the way that we're going to solve for problems with AI?

14:18Alex Valente: like well the way that we're going to solve for problems with ai is actually by using ai to fix these challenges you know maybe that sounds a little bit uh self-fulfilling but the reality is like you hear this like 1500 websites have a vulnerability or whatever i think was like you know php vulnerability that was massive took down a huge chunk of the internet like the only way for people to respond quickly is with AI. We're now seeing, hey, even in scenarios where there are a red, sorry, a black hat hacker, like a person who's truly a malicious or sometimes a state actor. The only way to defend those hacks is also with AI.

14:57Alex Valente: And so we need to proliferate these skills really quickly. We need to get more people and more enterprises using more AI more deeply. And in order to do that, it's going to be very challenging. It requires probably a lot of training requires a lot of change. And this is really what we're seeing in the economy more broadly, which is we're seeing executives really come out and be like, we actually need to do more. Like it's an interesting conversation I think for us to have, which is like, Hey, do we have the recipe in Australia for success? You know, in America, it doesn't feel like there's any barriers to go.

15:29Alex Valente: If you're an, if you're working in the corporate environment, you know, that might not be the same in every other country around the world. The court, the social license is different across the world about how people are using AI. And there needs to be a conversation about like, how do we get that kind of more social license throughout more organizations to use more AI because it's going to protect them as well as make them more efficient and effective.

15:59Georgie Healy: Look, we get it. Global hiring and payroll can be a headache. Companies like Shopify, Revolut and Eleven Labs already scale internationally with Deal. And Airwallex, the leading AI finance platform, they manage and pay over 1 ,185 employees across 16 countries. And they build their global team with Deal. So visit deel.com slash day one and start expanding your business today.

16:36Georgie Healy: So beautifully said. I'm glad you brought up the PHP incident. And for listeners, the way I grapple with this, and Alex, you might be able to correct me here, the internet is riddled with bad code already. It is. And really, it didn't make anything worse. It just found it. And what happened when they found it is they shouldn't have told anyone, but the hackers bound out dove in where they probably shouldn't have. But then we use AI, to your point, to solve that as well. We're kind of getting the ghosts out of the haunted house, but they were already there, right?

17:14Alex Valente: I think one way of thinking about this is like there's a lot of security via obfuscation, like just the fact that things are so Byzantine that no one can possibly go and look at every single thing and figure out where there's potentially like a vulnerability. And like that's obviously changed. because like all of the Byzantine obfuscation that used to exist, it's like that that's now super simple. Anyone can do that. And if anyone can do this, it's like, well, how do we think about like security going forward? Well, we probably need to start working on systems that assume every single line of code is going to be read and interpreted.

17:50Alex Valente: And it's going to be, it's going to be scrutinized and it's going to be attacked. We need to have those systems internally inside of companies and we need to have those systems internally inside of governments. and we need to have those systems internally inside of potentially our own personal lives when we're using software and code ourselves.

18:06Georgie Healy: Beautifully said. Before we move on to the next headline, is this something for regular listeners to worry about or really is this, you know, the businesses can be, you know, the ones this is a nightmare for? Is there anything we could be doing to further arm ourselves or, you know?

18:22Alex Valente: It's a great question. Yeah, I think like, hey, like there's a lot of AI dumerism, like let's address it. Like people are scared about AI because it's like, it's clear that it can do jobs already and the models are incredibly powerful. My position is pretty, is pretty clear. Like the models don't need to get any more powerful for them to have any more of a, of a really massive impact on, on everyone's life. And if that's the case, like, well, how does that affect us? Like, well, we'll really just accessing the power that's available already in the models becomes something that's integral to both our economy, the companies we work for and our own careers.

18:59Alex Valente: And probably the way I would look at something like this PHP incident is I'd be like, hey, listen, if I'm a normal person on the street, I'm not going to go and patch every website. I can't do that. And I probably can't manage all of the code of every single tool I use. But I can protect my own career by using the tools and learning how to use the tools in an optimal way because I'm going to empower myself to pursue more jobs or I'm going to empower myself to be more productive or I'm going to empower myself as well to kind of represent my interests because let's be frank, these tools are so great that we're already seeing people use these tools to stand up in a political environment or in a legal case, whatever it might be.

19:41Alex Valente: People are using these tools to represent themselves far beyond coding and all these things. And the only way you can do that is by really empowering yourself to experience that AI diffusion.

19:52Georgie Healy: Glad you brought that up. I don't think we've ever mentioned, you know, yes, we've talked about coding, helping people bring business ideas to life in a world they never would have necessarily been able to do so. Couldn't have access to a CTO easily. But yes, legally, if you need to represent yourself, support yourself, advocate for yourself. I've definitely used AI to read through some contracts for me. don't have that lawyers in the family no and it's and it's like i think everyone's doing that like

20:22Alex Valente: i think there was an article in the in the australian press about people representing themselves in fair work like you know there's a there's a plethora of of people who are using these tools to empower themselves in a legal environment and that there's like there's nothing wrong with that like um the government has kind of come out and said like that's kind of a reality going forward and like it's kind of a reality for like for the justice system more broadly like People are going to use these tools, whether it's the lawyer, whether it's a person individually. Again, you can only do that, though, if you truly understand how the tools work and you can actually go through that process of air diffusion.

20:58Alex Valente: Otherwise, you're going to be left behind by people who do know how to use those tools and are more competitive than you in that regard.

21:06Georgie Healy: Mate, I feel a little bit weird about this. I have some agents looking out for me for secondhand golf clubs, right? I don't know what I'm looking for. I know enough to know what is good. And I feel like it is a very unfair advantage. Of course. As soon as something comes up on Facebook Marketplace, I've got it. I understand, I know, and I know how to. And really, that's not fair. Like, it should be an even playing field, first in, best dressed. But I'm like, no.

21:38Alex Valente: It's really challenging. And like undeniably, I think the point you raise here about like, like, is it fair? It's like, yeah, like, no, it's an unfair advantage, but it's like, like, let that sink in. Like it's, it's, it's, it's okay. If you have an unfair advantage when you're buying secondhand golf clubs, like it's probably not going to ruin anyone's career or change anyone's life. People are going to work every day and they have an unfair advantage and other people don't. Like what happens when a firm has an unfair advantage? What happens when an economy has an unfair advantage? like the us for example is the only people who can access frontier ai models and they have this incredible unfair advantage relative to the rest of the world like what does our economy look like relative to theirs well we're probably going to be out competed we're probably going to lose out in a lot of like major industries that's the same like throughout the whole world if we aren't able to kind of get the productivity benefits of all the things that we see in front of us things happen slower it's that we become irrelevant like very quickly and i think like that's the true threat to the economy like we can talk about ai agents going rogue and like potentially getting hacked and that's it that's a challenge there's a solution to it the places where there are less solutions the things that are much harder is like well how do i take an economy or a firm and how do i get it to use ai in a way that keeps it relevant in the future by essentially becoming more productive and more and more competitive that's a really big challenge this is why i get

22:58Georgie Healy: you on the show, mate. You always nail it. This is very relevant to the next section. Most of us that are listening to this show have a paid AI subscription to Claw, ChackGPT, maybe Gemini as part of Google WorkSuite. You've probably got a paid.

23:13Alex Valente: I think we actually have every paid LLM for tests and purposes at a minimum. I think we have every paid LLM and we have every single open source LLM available to us. That's incredible.

23:23Georgie Healy: I have too many and not enough at the same time. But for many people, they're still using the free models, right? And the headline here is that, you know, increasingly ads are going to show up. Already OpenAI have, ChatGBT has advertising on 89 different countries on the part of that rollout. And then more than 100 ,000 advertisers have already run campaigns on their platform. This is going to get worse, not better. What does it mean for the free users and what should they be aware of, Alex?

23:54Alex Valente: Yeah, a lot of people like in San Francisco use the phrase like, are you token rich? Are you token poor? And it's a description of a firm's structure and whether or not they have access to a large amount of tokens. Obviously, the foundation labs, they consider themselves to be token rich and they just consider everyone else to be token poor. And that's like the, I don't know how much you follow Twitter, the concept of the permanent underclass. You're actually just token poor and you can't keep up with the thinking available to a firm that has incredible token riches. I don't necessarily agree with that exact characterization, but it's emblematic of potentially the challenge that exists, which is like, hey, if you don't have a lot of tokens available to you and, you know, a lot of work requires thinking and it requires intelligence on tap, like undeniably, you're like a less competitive entity.

24:44Alex Valente: It's really easy to look at the world and go like, oh, well, people will just compete. They'll find something to do. We have to recognize that knowledge work is an incredibly large percentage of our population. it's an incredibly large percentage of the american population it's an incredibly large percentage of the european population you know it is a very high percentage it's an even larger larger percentage of gdp so it's the most generally the highest paying jobs we need to to be able to do these jobs at a much higher level of abstraction much more efficiently otherwise like we're going to be left behind or otherwise these like jobs are just going to become obsolete and i think the The only way to do that is to like be at the front frontier, like use frontier models, have access to frontier amounts of tokens.

25:26Alex Valente: Otherwise the firms that do, they're going to, they're going to outcompete you on your, your job. I can assure you like right now, there's, there's probably people sitting in industries like insurance, for example. And then they're wondering like, you know, could AI do my job? Like I can assure you the fact, like a frontier lab's already tested whether it can be a job. It's already come up with the answer. Yes. and the only reason they haven't pursued doing your job is because there's bigger markets for them to pursue right now, which is like potentially the saddest reason why a person doesn't compete against you.

25:54Georgie Healy: I could kill you, but I'm going to do it after lunch.

25:57Alex Valente: Yeah, it's like I could kill you, but like I've actually got higher value people to kill right now. Like you're actually just such a loser that like...

26:03Georgie Healy: You're below my line.

26:05Alex Valente: Yeah, exactly. Like that's like pretty rough to comprehend.

26:09Georgie Healy: It's rough. I was at UTS University here the other day. There's students in the audience about to graduate and they're like curious what careers to do. And of course they asked me my opinion, like, oh, what should we do? And I was like, I honestly, like I'd be lying like to guess. Is there anything you reckon, Alex, that students should consider?

26:35Alex Valente: I had the same conversation with like my cousin and I've had the same conversation with like people all around. Like we always have this conversation. I read this book by Nassim Taleb, which is anti-fragile. I don't know if you've ever heard of this book.

26:47Georgie Healy: I love Nassim Taleb. We've got all the books on the bookshelf. I haven't read anti-fragile. He's a black swan, white swan, but black swan event.

26:57Alex Valente: Black swan, yeah. No, like I think I'm the same. I think I have like all the books on the shelf and like I perform it. I've read all of them, but I've only read anti-fragile. Like his point is really interesting, which is like, you know, the opposite, like people consider the opposite of fragility, robustness and he makes the case like well that that's actually incorrect he's like the the and the opposite of fragility is like anti-fragility where like chaos doesn't cause the thing to break it makes the thing better um like something that's robust can like resist chaos he's like but what if what if there's another thing that if there's chaos it gets better um and you could he like puts forward this case like that's anti-fragility um and yeah and i think like in the ai era it's like i think the people who because there's so much change going on like the people who survive and the people who who who find it advantage so people who have the traits of a person who are anti-fragile it's like a person who who maybe doesn't uh look for things based on their prestige or their status but maybe sees value beyond that and pursues those things like you know maybe a person who is very principled who can apply their principles to to industries regardless of of of what that industry is maybe these people like do better when there's a lot of change because people are looking for direction rather than a person who potentially is very conservative and and does the same job every single day so like these are the types of things that like started to think a little bit more given the fact that everything's changing so quickly.

28:33Georgie Healy: Yeah, I completely agree. It's almost like it doesn't matter what career you choose, guys, but can you be epic under pressure, better under pressure than everyone else, and adapt and pivot. And I mean, what career is linear? I know mine hasn't been, and I think that's one consistency is like it just keeps adapting and evolving, right? Yeah.

28:57Alex Valente: we spoke earlier about this concept of the factory of factories and i think that actually like that points most people towards something that like is a valid skill they need to learn which is like a person who is in a fact like if you work any job if you're responsible for anything you're going to be responsible for your factory like it doesn't matter what you make like it's kind of funny it's like i almost look at like an interesting analogy is like what does a have you ever met a factory owner, the person who makes doors in their factory, they didn't do a degree in doors. Like they don't have a complex understanding of the history of the door.

29:34Alex Valente: They just make doors. I think there's like something to be said of that in the modern era, which is like, look, you're probably going to become like that guy who runs a door factory where like, you might not know doors as intimately as you think, but like, you're going to know how to you're going to know them, how to be a manufacturer. You're going to be responsible for the output of DAWs. Whatever it might be, I think like there's something in that framework of like we're moving up the abstraction layers. We're no longer on the conveyor belt. We're somewhere a little bit higher up. And if we're higher up, it's like, well, we're going to take that responsibility with us and we're going to do the job differently, even though we still might be responsible for it.

30:11Georgie Healy: Beautifully said. Last quick one. Andrew Charlton, our Assistant Minister for Science, Technology and the Digital Economy here in New South Wales, wants AI companies to demonstrate their safety systems. So apparently airline industries and banks need to do this currently. He wants that rolled out for all AI companies, to which my first question is, what the hell's an AI company? Do you just meet an internet company? Like, what the heck? Do you think this is a fair thing? And Alex, you're in that kind of political role. Is there anything you would?

30:45Alex Valente: Yeah, it's a great question. I think Andrew's hitting on like hitting at a problem that exists with social license. He's got an issue, which is like these AI companies are clearly incredibly valuable and we require them every single day. They're looking to potentially build data centers in our country that are going to run foundation models, potentially even run the training of foundation models. And like, how do people in our society feel about this? You know, I'm here in California, obviously in San Francisco, people are excited, but like even in San Francisco, they've put a moratorium on data construct.

31:16Alex Valente: data center construction like so the the social license is really complex i think andrew is trying to to get that social license problem fixed in a novel way that's like maybe if they demonstrate their safety they'll they'll do like people will be more amenable to their to to them being around a data center that might be running an ai model all these things i kind of appreciate that like even if even if anthropic and open ai can demonstrate to me that their models are safe that doesn't make me safe from AI. Like there's obviously the ability for people to use AI models that are open source in an unsafe way.

31:53Alex Valente: Like the safety question of like how do we remain safe using AI over an extended period of time just like generally remains unsolved. Like there's nothing to stop an AI model from pretending to be safe and secretly harboring its negativity towards us for an extended period of time. Like this is the kind of misalignment problem that people often talk about, about like an AI potentially deceives us. All of these things like mean that it's going to be hard for us to demonstrably prove that AI is safe forever in every single case. But again, it hits this thing of like, well, maybe we can make people feel like these companies have a social license to operate.

32:29Alex Valente: That is actually the hard part. And I think the more we can do on social license, the better, whether that's on data centers, showing them that like, hey, they're going to be net neutral to the environment and they're going to generate tax revenues for our country and they're going to be beneficial to the local community in some way. If we can do that, we can potentially accrue incredible economic benefits associated with that, whether it's on foundation models like, hey, these foundation models, they operate within our copyright legislation or whatever it might be. If we can build that social license, we can actually create incredible economic benefits for people because I think right now, a lot of the time, societies and populaces are kind of getting in their own way when it comes to the AI diffusion journey.

33:13Alex Valente: A lot of the times people are shooting themselves in the foot by being like, hey, I want these AI things around me. People who do want those AI models around them and are experiencing the productivity benefits, they're out-competing you. Like if you work in a knowledge industry, do you work in a services industry? And eventually if you work even in a blue collar industry, you will be out-competed by people who are using these tools. Right now, we are slowing down our progress because we're unable to really own that social license.

33:42Georgie Healy: A poll that I did on my Instagram recently, which was data centers. And there are two options. Is this going to help Australia's economy or is it a negative? It's actually a super mean question. And yeah, it was interesting. People in tech clearly were like, well, we're not going to compete on foundational models. We need these data centers to work. You know, data centers here in Australia, I got to tell you, it's like a swear word. It really is. People hate. Yeah.

34:11Alex Valente: It's not something that I agree with in the sense that like I'm very, I'm pro data center and I'm pro data center build out. Like I'll go on record being like, I think a lot of the opposition to data centers is like strict nimbyism, like just like people just who are against everything. Like a lot of the times I find the conversation even challenging because like, I don't even understand what people are against. You know, it's an amalgamation of things like, like I'll take to maybe speak to a specific example, like listening to people talk about data centers, small data centers in the city, like people will come out and be like, oh, these data centers, they're going to steal our power.

34:44Alex Valente: The power grid operator comes out and says like, actually the reason why this data center is in the city in this particular area is because the grid is congested and we require a load on that particular node of the grid in order to reduce power grid congestion. It's literally going to bring your power price down. Like it's going to make it better for us to transmit power. and you present that to people and they're just like no i hate it this is the hard thing about like doing really large-scale engineering projects that involve like a lot of stakeholders which is like it can be really challenging to explain the complexity of a solution and take people on that journey there are so many opportunities for australia to own the data center build out and to do an incredible job like to accrue incredible economic growth at a time where our country is looking for that economic growth.

35:34Alex Valente: And right now we are getting in our own way.

35:37Georgie Healy: I'm really outing myself, but when I graduated 20 years ago from high school, I got into metallurgy. The mines that I could have worked at 20 years ago are not in operation anymore here. And by the way, the mines aren't great either, guys. So they've gone away, but for whatever reason we're kind of like numbed the idea of mining things out of the ground but we're anti

36:02Alex Valente: data center it's really odd like we look at the data centers and there's a lot of mimicry of like things that that people have heard previously like oh the you know they consume all this water or they they use all this electricity and it's like like okay well let's pass legislation then that says you can't build a data center on you unless it's positive for the water and agree and positive for the power grid it's like people are still opposed to it comes back to this thing about social license. The reality is data centers struggle with social license in our country because they've been unable to explain and elaborate to people why they're valuable to them as individuals.

36:37Alex Valente: I think it's really hard to hear the numbers of the amount of investment that's going in when you see the economy around you is not doing very well. And you're like, well, why are these guys like getting all of these like, you know, super profits or all these super investment cycle? And what does it mean for me? Am I just going to lose my job because of AI? Again, it comes back down to things like we need social license because we operate inside a democracy. There's nothing wrong with that. And how we get social license, it involves us addressing everyone's needs. But it might also involve like not just answering the oppositions, but also creating some carrots for people.

37:14Alex Valente: Like, okay, there's lots of sticks. It's like, well, what are the carrots? Like, why should a person want a data center in their local area? Like, you know, I think that conversation needs to be had more rather than just like, oh no, we can answer all of the opposition to it.

37:32Georgie Healy: You're right. They're too busy defending and not enough actually are like inspiring. Correct. Nate, this is the part everyone's been waiting for. The rapid fire personal agent showdown. It's like the era of the personal agent in general, I feel. All big like AI companies have one. Even some startups have one. Mate, do you have any personal agents running right at the end?

37:59Alex Valente: Yeah, I've tried a lot of them. I've tried a lot of them, I'll be honest with you. And I'm quite wedded to my clawed instance.

38:04Georgie Healy: We're going to put these on blast. You're going to rate them. I'm going to rate them. We're going to announce a winner. These are the top five, some of which, to your point earlier, I'm not even allowed to use yet because I based in Australia. So I rate it based on vibes.

38:22Alex Valente: I can report on them for you from the, yes, from the San Francisco office. Live from California. Yeah, let's do it. All right.

38:31Georgie Healy: I'm going to give a brief one sentence. You're going to rate it in a brief sentence why, and then I'll give my number, and then we're going to go through. So GrubBot, Elon's XAI, it was launched in August. every bot gets its own computer and it keeps working after you log off. Alex, score out of 10.

38:51Alex Valente: Yeah, I think like we're seeing this now with a lot of players, like everyone's releasing this thing of like, it's got its own VM, it's got its own computer that's running in the background. It's kind of, it's owning its own jobs. Look, I think it's like a, I think it's like an eight out of 10 right now. I think it's probably going to become a 10 out of 10 soon. I think we're seeing the same thing coming from, from Claude. I think the solution architecture makes a lot of sense and I think like you know this is what the kind of AI employee is going to look like a little bit more going forwards and we're going to see more and more like products have something like this.

39:21Georgie Healy: I see Croc AI is the perfect amalgamation of your factory with all these roles that are you know able to do their own domain and expertise stuff I play with it for a bit to be honest I don't continue to use it because I'm like a solopreneur like I it's just it's just a little bit, I don't really need long horizon tasks currently. I don't think, I don't think maybe that's a self-limiting belief. I give it a seven. Instinct launched in August as well by a 23 year old founder. This is a startup guys. No app. You text it through WhatsApp. Mate, what's your score for Instinct?

40:02Alex Valente: Look, I'm a five. Like it's really cool, but I just didn't really get it. I gave it try yeah i know all the founders i know they were like how you know like just get instinct like you

40:12Georgie Healy: get instinct to talk to your girlfriend it's like what um would give it a three yeah yeah exactly right we would probably kill me that could give it to its head yeah exactly um look like i've i've

40:28Alex Valente: got so much like um yeah i've clopped a lot of flack from just being like i don't really get it Like I don't, like booking a restaurant for me is fun. I like doing the booking part. I like looking at restaurants. I don't want to give it necessarily right now to an instinct agent, but I appreciate its existence because it probably will become useful to me at some point. So that's why I give it a five.

40:50Georgie Healy: Mate, I give it a 10. Have way too much admin and we're in different positions. This is a personal AI assistant. For me, two kids, two different schools, a million fucking emails. I need a personal AI agent and it's on WhatsApp where I'm already native. I don't need an extra app. It's surfacing things like, oh, dude, you didn't book Handball and now your son can't do it this time. I know. And I'm like.

41:18Alex Valente: Let me jump in here. I need it. And this is the difference between you and I. You have necessary messages you must respond to. If I get a message and I'm not interested in responding, I'm like, I don't have it. No one's relying on me. Like no one's going to die. It's going to be okay.

41:34Georgie Healy: No one's crying at the school gate because they can't play handball.

41:38Alex Valente: Exactly right. Exactly right. But that's probably the best pitch on handball.

41:41Georgie Healy: This is why I just had both of us. Because if you did my take, you'd be like, Grok sucks. Instinct's great. So I like that. Okay, we've got Muse from Meta. Launched in September, so one month later than these other two. It lives in like the Meta apps you already use. There's a little keychain gadget, which looks kind of like a Tamagotchi. Comes at Christmas. Mate, what do you think of Muse?

42:03Alex Valente: I think it's been pretty strong. I think its integration into the Facebook ecosystem is pretty incredible. So I'm a nine. I think let's give you some new golf clubs. Let's give you some new golf clubs, man.

42:16Georgie Healy: Okay, so my rating for Muse is also really high, but admittedly I haven't used it. This is based on vibes. It's based on vibes. Everyone loves Muse, and I love the idea of the integration with the glasses, especially for the vision impaired. I'm looking at something talk to me about it none of the other agents can do that I think that's awesome Dots this is the open AI player I think you need to have a pro or like a higher tier what do you give Dots?

42:45Alex Valente: well Dots are actually very similar to Grubbot and they have like a similar level of performance and they actually look a lot like I mean the branding is quite funny like everything's kind of converged on this like cartoony thingos I don't know what to call them blobs yeah it kind of has a very similar functionality i would probably rate it probably

43:02Georgie Healy: something similar i've given it a go a little bit and eight eight i think again this thing about

43:08Alex Valente: like having a virtual machine or having its own computer and being able to do long run tasks like inside of a business environment it's really valuable and like undeniably it's not a solution for every workflow but it's going to be a solution for a substantial amount of workflows and people are going to interact with it how they interact with an employee going to be inside of slack it's going to be inside of whatever and that's how they're going to get sold.

43:28Georgie Healy: Mate, I give it a six and I'm going to tell you why after the next one. The next one is Claude Duet. It's from Anthropic. This is the latest one. It just launched and it's kind of the more secure option. It checks it before it sends. It apparently reaches out to you, Alex. You don't have to prompt it. It's like, hey, I thought you might want to message your girlfriend. What do you think of Claude Duet?

43:55Alex Valente: I think it is a, it is like, it's very clawed. I like all of those safety features and stuff. They also hamper a little bit of the product experience. Like the, sometimes the benefit of all these other things is like that they, they are kind of, you know, they, they just do things and that's the magic. But like inside of the enterprise, which is where like Anthropic makes the majority of its revenues, it's like, it's kind of necessary. I'm going to probably use it a lot because it's going to work inside of our team. And again, we're kind of all in on Claude at this moment. So I have to say like, cause I'm all in on Claude, it's at least a nine, but like probably the safety features, I think for most people, they're going to probably find it like more like a seven.

44:34Georgie Healy: Yeah. I love Instinct and it's the least safe one. Correct. Like my brand, it's off brand for me, but apparently it does, it's got, um, it's more sophisticated than dots. So I give it higher than a sixth. I'd give it a eight because I love Claude, which means that it's tied for Muse and you have to choose the winner.

45:00Alex Valente: Yeah, let's give it to Muse. They've got great branding. I love that little, they've got that little cartoony, fluffy thing. It's fun.

45:06Georgie Healy: Yeah, that cutie, cutie patootie is kind of fuzzy. He's got a name, but I've forgotten it. Congratulations, Muse.

45:15Alex Valente: Yeah, exactly.

45:16Georgie Healy: I'll have to tell you. Congrats, Muse.

45:18Alex Valente: That guy, what's his name? Alexander Wan. He's a, like, what a turnaround on that guy. Well done to him.

45:25Georgie Healy: I want to do a full deep dive on that guy. He is a genius billionaire.

45:31Alex Valente: Love the Crocs.

45:32Georgie Healy: He wears camo Crocs.

45:33Alex Valente: Yeah, love the Crocs.

45:36Georgie Healy: Obsessed. You saw him on Cheeky Pint, the popular one?

45:39Alex Valente: Yeah, I did. Yeah, very like, surprisingly likable.

45:42Georgie Healy: All right, before we let you go, we know you're super busy. We're just going to do rapid fire hot takes. Number one. Which of the big labs are you betting on? I think I know the answer to this.

45:53Alex Valente: I'm probably betting on Anthropoc.

45:55Georgie Healy: What's your most controversial take?

45:57Alex Valente: No one's AI pills enough outside of probably four suburbs of San Francisco. It's way bigger and it's way more powerful than people think.

46:05Georgie Healy: Where is the AI industry getting it wrong?

46:08Alex Valente: They don't like humans enough. They should go to a pub and talk to a normal person.

46:13Georgie Healy: Get off Twitter, get to the pub. Finally, let's finish on three recommendations. Where do you get your best AI news?

46:21Alex Valente: I get it from Hacker News.

46:22Georgie Healy: One tech book everyone should read.

46:24Alex Valente: Oh, let's go with Anti-Fragile. We spoke about it in the app. Got to recommend it.

46:28Georgie Healy: Ready to go. And one gadget you can't live without. Hardware we're talking about.

46:33Alex Valente: Oh, I've got one of those standing desks that just sits on top of your standing desk. You know, the ones that were the cheap ones you get from Officeworks? I love them because it has a separate spot for your keyboard. That to me, it's like, I can't live without that.

46:47Georgie Healy: Mate, you are such a delight. Thank you so much for rating the agents, unpacking the headlines, dialing in from a completely different time zone. We love having you on the pod. Don't get too far away.

46:59Alex Valente: I'm going to go and eat some dinner, I think. Thank you very much. It's a pleasure as always. I'll see you next time.

47:04Georgie Healy: See you next time, mate. Thanks so much. Thank you so much for listening to In the Blink of AI. If you want to go deeper on anything we've spoken about today, I write a weekly substat called Attention is All I Need. Yes, it's hilarious. It's a pun. And essentially, I go into AI rants, tech news, events I'm going to, and more. It's bite-sized, and I hear it's awesome. The link is in the show notes below.