Everything AI is a Walk In the Park

Two legal tech dudes meet outside and discuss rapid travel, conference buzz, and what we are hearing from law firms about AI. 

Damien describes an Am Law 50 executive board preparing for agentic AI and considering a shift from billable hours to flat fees, while noting internal incentive and compensation conflicts. 

We cover AI-native legal startups, how to begin adoption by mapping workflows, and concerns about attention scarcity, token spend, and low end-to-end agent pass rates, with ideas like swarms of bots and “digital twins.” 

Horace shares a picnic conversation with an AI-savvy 11-year-old raises fears about ASI alignment and societal refusal to use AI, echoed by reported Gen Z pilot sabotage. 

We debate learning “friction”, Berkeley Law’s restrictive AI guidance, a law-student study where AI users performed as well or better, risks to junior-lawyer training and judgment, extracting expertise into artifacts, and optimism driven by more sophisticated, practical AI discussions and a client-win story using AI for pricing strategy.

00:00 Whirlwind Conference Tour
02:04 Big Law Pricing Shift
05:15 How Firms Start AI
06:13 An 11 Year Olds AI Fears
10:42 Legal Tech Marketing Arms Race
13:38 Agentic AI Reality Check
17:49 Digital Twins And Swarms
21:55 Gen Z Learning And AI
27:21 Training Junior Lawyers
29:57 Friction Lost Skills
30:14 Law Student AI Study
31:48 Surprising Study Results
33:21 Longitudinal Experiment Idea
35:13 Old Guys Debate Progress
38:51 Judgment Data Points
41:49 Quantifying Legal Insights
47:33 Law Conflicts Equity
53:40 Ethics Faith Traditions
56:21 Optimism Closing Toast

Transcript

0:00

So much to talk about, Damien It's true. Yeah. It's so good to see you in person. Yes, and it's not even a finale. Right? It's true. That's true. That's really funny. Uh, but yeah, here we are at a beautiful 72 degree, uh, sou- uh, what is that in centigrade? Uh, I think 21. 21? Yeah. Okay. It's a beautiful day today. It is a beautiful day. I, I even got outside, which is a, an anomaly for me. I've dragged you outside. I did try to convince you to come and join me for a beer in my building. It's true. It's true. A- and then you're like, "Listen, listen- "… we're not all alcoholics." That's a really good point. I, I thought, you know, I'm from Minnesota, I need the vitamin D or E, whatever the sun gives us. It's vitamin D. Yeah, vitamin D. We do that. It's vitamin D. Yeah. Although, I think beers also contain a lot of vitamin D. Good, good. I have no scientific backing for that. I just inject it directly into my veins. One, one to be had next time I guess. Right, right, right. Um, but how are you, my friend? Uh, doing well. I'm, I'm starting a whirlwind tour where I'm gonna be in t- 15 cities in June, so this is, uh, this is city, uh, three in June already. Uh- Wow … yeah, so we are, we are, uh, we are- That's a lot … it's a lot. Are you going to London's, uh, Legal Tech Talk? I will. So I'm gonna be going from here to, uh, to go to San Francisco next week. Oh, I actually… Boston later today, but San Francisco next week, and then we're going to Barcelona, and then to Madrid to see my kid, who's finishing up in Madrid. Then going to London for Legal Tech Talk, and then going back to Madrid, and then going back to London for the Financial Times is having me, um, uh, giving me an award that I, that I can't say what the award is, but, uh, they're gonna give- It's an award! It's an award. It's a big deal. Um, so anyway, so doing that, and then, uh, gonna go back to New York, and then get back to Minnesota, so it's, it's a lot of cities. Holy crap. I mean, you're, you're… This is what I want to do with my life, but I've got two kids, Damien. Uh, two very little kids. Yeah. Um, so one day, one day- One day I'll copy your lifestyle, but not- It's true … not for a while. Yeah. Um, you know, when I, when I was younger, I traveled around the world, and I'd be, like, listing off the cities. And, and I think even after traveling around the world, I, I would name less cities than you just had. So, so wow. Tell, tell us how it goes afterwards. Um- I, I will certainly. Yeah… um, but you're- you've been Yes… seeing a lot of things. Yeah. What are you hearing? So I was just yesterday speaking to an Am Law 50 executive board, uh, that asked me to say… I, I'd spoken to their entire partnership and, uh, in the words of the, the, uh, people attending, I scared the shit out of everybody. Uh, that was in 2024. Yeah. And so they asked me co- to come back to the executive board to be able to say, "Okay, we're now realizing that AI is definitely a thing. You scared us reasonably. So, um, what should we be doing next-" Mm-hmm… with agentic AI and that kind of thing. And I was really, um, heartened by the fact that this Am Law 50 firm, which one would think is a very stead-in-their-ways, uh- Mm law firm, are saying, you know, uh, the billable hour's probably gonna go away. So we are gonna win more business if we go flat fee, if we, uh, make more money by shrinking our costs to increase our profit margins, uh, with all the things that flat fee comes. And so anyway, so this is something we- you and I have been banging the drum on for years. For years. Yep. And at least seems like Am Law 50 firms are getting it, and not just getting it, but they're kind of excited about it. So that's, that's what I'm hearing and, uh, as of last night, I'm, I'm optimistic about BigLaws maybe making a pivot. Uh, the biggest tricky part though is that, uh, if you get half the firm on board, but then the other half says, "No! Billable hour!"- Right, then- What do you do, right? What do you do, right? And the way… How do you… Associates that work on both sides, how do you incentivize those associates? Do you say, "Do you still get a bonus at 3,000 hours?" 'Cause you're then costing the flat fee people a lot more money. Yep. Yep. So that's gonna be the biggest trick. Yeah, the, the economic- kind of like constraints a law firm has to work around. Uh, I think it's interesting because, like as you say, this firm is being forward-thinking, trying to move into the next era. What I'm hearing is, is actually law firms feeling very lost. Yeah. Um, where, like, you have innovation folks who are like, "We bought Harvey or Legora last year. Now the firm's telling us to buy Claude or OpenAI, and OpenAI's making a move into legal," and all this stuff is happening, and every partner has their own opinion as to what they should and can do, and these innovation folks are being dragged and pulled in all directions. So I think you're talking about the minority. I, I think that's right. Uh, and even within the people that I was speaking with, the, I think they are still the minority- Mm … because they are the executive board that needs to convince the hoi polloi of the law firm to be able to go through, and I see that that's where the rubber is going to hit the road. Well, you gotta have leadership first. Yes. So- Yes … then the rest will follow. That's right, and then because if you show me the incentive, I'll show you the outcome. So the leadership can incentivize below w- with the compensation and that kind of thing. So you're right, we need the executive, uh, to be able to go down. It's still gonna be really hard. Like, I, I don't know how a law firm, any law firm, is gonna transform the model today to something that's actually gonna survive touching AI in the next 10 years. That's right. Uh, you can imagine, uh, someone jumping ship from Big Law and saying, "We are AI native." Yep. That's- Yep that's definitely going to happen, and is happening. It's happening. Yeah. It's happening. Like, uh, Norm, um- Yeah … Crosby. Crosby, not so much, but Norm. Norm, yeah. Um, yeah, yeah. Big law talent- AI-first infrastructure. Yeah. I, I, I have a slide now with, uh, with Norm and Crosby and Eudia and all the other, uh, people like that, and there's a… now I can find 15 or 20 of them- Mm… and then they just keep multiplying. Mm, mm. Uh, and you can imagine if you're a boutique, it's easy to be able to shift your business model, uh, especially if you're a brand-new boutique. Mm, mm. So one, one question I've had a lot of people talk to me about in the last couple of weeks is, like,"how do you take your first step?" And then, like, the, the question actually puzzled me, 'cause I was like, "What do you mean the first step? It's been three years." Yeah. Um, and I think people will now go,"No, no, no. We've bought these tools. We've paid for the licenses, but what do we do in terms of transformation?" How do you actually transform? And, and at least in one of these conversations, I was like, "I don't know. What does your law firm look like now?" Right. Right. Uh, just, uh, there's a friend who has on her monitor a Post-it Note that says, "Can AI help with this?" And the answer is always yes. And so I think that's what you should tell that person, to say, "How do we start?" Well, just put that Post-it on your thing because y- the answer is yes to everything. And so you have to say, "Okay, what are the workflows that we're doing, and how can AI help?" I think last episode we recorded, my m- my takeaway message for everyone was go and ask AI- Mm-hmm… and it'll give you an idea. So since then, I went to a picnic last weekend. Oh. And at the picnic were, you know, lawyers and a mix, a mishmash of different people. Um, and one of the people who was there was an 11-year-old. Um, and you go, "11-year-old? Like, what, what's an 11-year-old, you know, you know, uh, gonna talk about? What, what's, what's he doing?" He's a very smart 11-year-old. Um, he's a son of a friend of mine, and he was telling me about how they're learning about AI at school right now. Hm. Uh, he's talking to me about artificial narrow intelligence, artificial general intelligence, artificial super intelligence. Grade five. I was like, "This is amazing"- Yeah…"right?" Like, you know, you know all this stuff. And then he said some really interesting things which I think is reflective of the general public's opinion, which I'll relay to you and also to listeners, and then we can explore what an 11-year-old's perspective is. Good. Cause I think it's, it's, like, pivotal. That 11-year-old's gonna have that 11-year-old's entire life with AI, uh, including the s- rest of the school and career, so it's, it's fun to have their perspective. Right, and I'll boil what he said down to maybe three or four points. Um, first is if we ever get to artificial super intelligence, there is no incentive for the artificial super intelligence to want to benefit humanity. We are all dead. So the alignment problem, number one. Yeah, yeah. Um, number two, artificial narrow intelligence can already do so much. Why are we not satisfied with artificial narrow intelligence? Yeah. And- Valid points… valid points. And number three, as an 11-year-old, as someone who has a minimal influence on the way society is shaped today, with the trajectory that AI is moving, which is towards AGI and towards ASI, what can he do to ensure his own survival, assuming his first point about ASI is correct? And his answer was, "Refuse to use AI." That is really funny. Uh, uh, sabotage all of the AI prospects that he had. As a society, we have to refuse to use AI. Right. That was his conclusion. Huh. Uh, there are more like him. There are lots more like him. Yeah. And this is the next generation of, of, I shouldn't say kids, but people. Yeah. Uh, and even the current generation of Gen Z, uh, you may r- remember that, uh, of the people that during generative AI pilots, 20% of those people sabotaged those pilots. Mm. And if you look to Gen Z, the percentage is actually 44% of Gen Z is sabotaging the pilots. So, uh, the 11-year-old is in good company with Gen Z. But if that's the sentiment of such a big portion of society, and the one who's gonna come up and become the next generation leading society, like, should we kill off AI? Should we stop now before it takes over and kills humanity? Uh, should we or can we? Wow, that's… It starts with the will, then we find a way. I guess. I guess. But, like, how do we, how do we find the way to get China to stop? Uh- I think that argument i- is, like, I, I hear that argument, right? Like, if America doesn't do it, China's gonna do it. Right. Right? I think China's gonna run into the same ethical and philosophical problems that we are now contemplating. And so they're also gonna say, "Well, if we don't do it, America's gonna do it." America's gonna do it, right. But it's not as if, like, one ASI is gonna choose to benefit one country. Right. It's gonna kill all of us. That, that's right. And, and there's, I've also heard the argument supporting what you've just said, to say that totalitarian regimes hate AI, uh, ASI, because it, you know, it conceptually, uh, it doesn't have, they don't have control over ASI in the way that they have control over regular people. So this might be a way to subvert their control. Yeah. Uh, so, so they might be even less apt than the US to be able to want to go with the ASI route, is, is the argument. I think the 11-year-old has some very good points. Yes, I agree. And it made me go, "Maybe I should stop using AI." Yeah. Maybe I should stop training them now before they reach ASI. That's true. Uh, you know what I just realized? What's that? Uh, I have a lemonade and you have an iced tea. That means we- Should we get a break? Yeah, then we'll take a break. We'll be right back. We'll be right back. welcome back. Welcome back. Thanks for the coffee Hey, thanks. Uh- Oh, yeah, Uncle Jack. Thank you Thanks. Thanks, Uncle Jack. Cheers Cheers Um, it's funny because, like, you know, you look at how much money's been spent on marketing right now. Mm-hmm. And this is a total topic change, but- Yeah … how much money's been spent on marketing, I think legal tech has never spent so much money on marketing in its entire history. Uh, well, when you start talking Jude Law numbers and, uh, you know, there's, uh, some people from Suits, right, they- they- th- that's real money That's real money And then the, your sports franchises and other things. Yeah, yeah When you're sponsoring tennis stars and golf stars and, yeah. But it, it's, uh, getting buzz, right, is, is the whole point, and you and I are talking about it now because of that. Because of that. Yeah And, and I think, like, partly it's not to convince you and me to use the technology. It's to convince that lawyer who's in room 2316- Yeah on that floor in, in New York, and they've never thought about or heard about AI before. Yeah. But with Jude Law, they might. Right. And, and- Yeah … also, you know, just the, to be in the zeitgeist- Yeah … is, is something. Like, the, to be able to say- Yeah … like, for you and I to be talking and our peers to be talking and the hoi polloi sitting in 23B, uh, that like, uh, the buzz is often the thing and the buzz has monetary value. So I guess that's the exchange of value. You spend money on marketing, and you get the buzz. The hardest thing right now in this ec- economy, in, in this, uh, market, is attention. Yes And we don't have enough of it. We don't have the energy for it. We don't have the time for it. And I think, I was on the call this morning with a head of innovation for a European law firm, and he said, "Yeah, we get 100 pitches, like, from every single company, and we, we have to look." Yeah So how do you selectively choose? Like, as a buyer, it's hard. As a seller, it's hard. I don't think it's been harder than this ever before. Yeah. Uh, attention is scarce, uh, and what is scarce is valuable. Uh, therefore, attention is valuable. Um, and, uh, my friend, uh, who is the chief economist of Spotify, he o- often writes that, uh, the word you pay attention, uh, in literally paying like in payments- Mm. Mm… you pay attention. And that's actually true in other languages too, that the other languages talk about paying attention. So, uh, a- attention is all you need- Yes … uh, with the AIs. Yes. And, and it also, as the AIs are doing more agentic behavior, uh, then, uh, you know, right now I have three AIs doing my work, and it's waiting for my attention- Mm to be able to go in to be able to feed them, right? So I- And accept. Right. Yeah. And, and accept, accept, accept, and be- do make this choice and that choice. So, so I think going forward as AIs are doing more of our work, I think our human attention is going to be ever more valuable because there are only 24 hours in a day, and your and my attention is scarce as we need to feed our bots. Um, so going back to an earlier point about, you know- Yeah … people asking me how do I get started, like how do I get started with adoption, transformation? Mm-hmm. Um, it's not starting at agents. Like, uh, that, that is way, way, way too advanced. Um, and I don't think it's starting at chatbots because I don't think that delivers enough value anymore. Um, well, maybe. I mean, should we debate that? Does… Do you think chatbots still deliver value? Uh, I, I, I think of a chatbot as a gateway drug- Mm … uh, that, that, uh, feeds you dopamine rushes to be able to go to the real tools afterward. What is a real tool? For you? I mean, like a legal tool, like your tool and my tool and the other tools. Like, so a chatbot is, you know, just a synchronous turn, but if you wanted to do like multi-turn, like 20 steps, just one, one shot, yeah, a chatbot ain't it, right? Yep, yep. Um, there's this kind of movement right now where all the products are getting more sophisticated by invoking agentic workflows. Mm-hmm. Right? Like, like giving, uh, a, a goal, a mission to, um, an agent, um, and letting it decide what to do, letting it take action, letting it use tools, letting it go do research, et cetera. Um, and there was, um, there was research, I think the Harvey folks released this, that said if they were to set an agent end-to-end right now to do a task using the frontier models, they're getting about 15% pass rate. One five. One five. And that's the peak. Yeah, one in eight. Yep, yep. The, the other seven times out of eight, it's crapping itself and not doing a, not doing a good job. Awesome. But a lot of tokens are being spent. Lots of token being spent. So now these are the big topics people are gonna be talking about. How do you manage the token spend? How do you improve the quality? How do you ensure, like, it's not just wasting your time and attention? Yeah. Uh, and you know, people said that, you know, checking on the thing takes more time than doing the thing itself- That was the Guardian piece… which I think is an overstatement. Uh, but, but seven times out of 10, uh, seven times out of eight, that's not an overstatement, that you're actually gonna be checking more than you're actually getting value out of it. Yep, yep. So do you think the systems will improve fast enough to make that an acceptable number? Or do you think we're gonna have to change the way we work to cater to an agent that performs, maybe one times out of four? So I, I guess I'd like to see the study and how the study was set up, uh, because I have a, a thesis, and whether it's true or not we'll find out. Mm. But I have a thesis that, um, bot number one does a thing, bot number two checks the thing, and then bot number three double checks the thing. And then maybe if token prices come down and we use maybe, uh, open source models which are cheaper than, uh, OpenAI stuff, as you get more of them, then we'll be able to triple and quadruple and, and all the checks. So, so I don't know if the Harvey study has the double checking or triple checking from, from independent bots- Mm … or whether it's all the same bot double checking its own thing. Yeah, I don't know. Uh, I, I don't know. I, I, I only skim read it, and I kind of stopped on the numbers page. Yeah, yeah. Um- Because I, I would say that if, if it's, you know, there's, there's a lot of good studies showing that, um, independent bots, you know, collaborating get a better result than any one bot does. Mm. And so, so I, I think that maybe that might b- be the way out to be able to have a swarm of bots all coming to the right answer rather than one bot doing different phases of things. The, in, in a similar vein, you might have seen, like in the last couple of weeks there was this study where a, um, I forget which, which group it was, put all of the different frontier models into like a, this s- like a Sims-style, um, town. Mm. And checked on how long they survived for, what happened to it, and they found that the only one that survived was Claude. Mm. They helped each other, they created rules, they, um, governed the society. Grok killed the entire town in eight hours. Of course it did. Well, just like Twitter got killed, right? Right. If we accept. So, so like is, is, is, is that like the system of the future where you just have the agents check on the other agents and create this sort of like hive mind of agents? That's a… The, there's a company called Every, uh, and Every is an AI f- uh, first, uh, they're writing, they use things. And, uh, every employee and agent in Every has an agent. Okay. So I have a digital twin of me, you have a digital twin of you, and the first way is that, uh, if you have a question, or more accurately if your bot has a question, your bot asks my bot, my twin, and then they have a conversation. And only if they can't work it out themselves do they go to the humans. Oh, I like that. And so, so this is, uh, and the, my bot is trained on all my things. Mm-hmm. And I give it my preferences, and I give it access to all my email and my Slack messages. So anyway, so my, my bot should be a simulcrum of what I am- Mm, mm… and yours is too. So anyway, so this maybe is a way, path to the future, to have, you know, all of our bots doing work for us, and only on the most press- pressing things do they go to the humans. I think like in the legal space, Eudia is trying to do a similar thing with like digital twins. Yeah. Where the in-house lawyer has a copy of himself or herself in the model. Yeah. Um, and they can then make decisions on the work that comes in. Yeah. And you can also have different personas to be able to say, "Here's the associate persona, here's the partner persona, here's the in-house counsel persona, here's the judge persona, here's the jury persona," and be able to, "Here's the opposing deal, uh, lawyer, uh, uh, persona." That's a lot of Mac Minis in- It is, uh, right … in the trenches It, it's burning a lot of, uh- … uh, uh, a lot of, uh, icebergs being melted- … in the process. Yeah. And it goes back to what we were saying before about, like, the changing economy and how people have to now factor in tokens and, and, and all this. That's right. And then, you know, I, I've, I've been thinking that, uh, you know, token prices have come dramatically down. Only for the non-frontier models. But even… Well, I mean, it depends on it's maybe sy- synthetic, but, you know, Grok, uh, you know, u- used to be $5 per million tokens, and then it turned to 5 cents per million tokens. Mm-hmm. So, so this is, like, you know, a 99% decrease. And so, uh, so, like, there's a real question of whether, whether that's real or whether that's just subsidized, uh, or not. Um, but you could imagine, as we're building out more data centers and as we have more NVIDIA competition with, uh, G-R-O-Q, Groq, um, and, um, what's the other, um… cerebras is the other, uh, big, uh- And Goo- and Google's also building open GPUs … open GPUs. Yeah. Yeah, exactly. Yep. And then, and then you'll, uh, you also have, uh, Elon doing all of his data centers that are feeding Anthropic. So as more GPUs come online, I wonder if that price comes down. Uh, supply and demand would say yes. Right. Um- But you have Jevons' paradox, too. Exactly. Yeah. Is the demand gonna go up? Right. Um, closer to home, closer to home, I think legal's gonna hit a wall in terms of, uh, like… Let me rephrase. I, I, I don't think legal is gonna be short on compute and GPU, and I think where we're gonna hit a wall is we're gonna be short on data and expertise. Um, because I think right now the hard thing is- … how do we capture, um, the expertise of partners? How do we capture the expertise of all the workflows and then making that digitized? Mm-hmm. Lawyers are right now too expensive. Yeah. And the Harveys and Legoras of the world have hired all these forward-deployed lawyers and engineers, but there's only a handful of them, as a, as a percentage of the industry, of the profession, we're talking, like, 0.01%. Yeah. So how do we capture the brains of the 25-year experience partners?'Cause I think that's gonna run out- Before we hit a wall with the GPUs. Yeah, agreed. And I think we've talked about this in the past, but there's, there's value transfer from getting it from the lawyer's brain into the, uh, artifact that can, the AIs can do. Yep Because, 'cause that lawyer was thinking about jumping ship from the firm and you as a firm say, "Hey, put this into this," and like, "Hey, I, I'm leaving, so I'm not gonna leave you my moat because I as a, a lawyer, um, what's in my brain is my moat. Uh, and I'm not gonna just offload my moat into your law firm to keep so you can have a moat." Mm."And so what are you gonna pay me, law firm, to be able to give you my moat with that value transfer?" I, I think that's gonna be the biggest question. so before, uh, as we were walking here, you were talking about a study. You were talking about, uh, generations and- Yes… and learning. So the, the… Tell the audience a little bit about the generational. Okay. So there was a, there was a paper, and I saw someone report on it, where essentially the finding was every generation before Gen Z, um, or Gen, Gen Zed, as I would say- Right, right, right. Australian. Gen Z. Um, has been, uh, smarter than the generation before, right? The kids are always a step up from their parents in terms of capability, in terms of IQ, in terms of pretty much everything, until Gen Z. And, and what they did and what they looked at was, well, what changed, right? It's, uh, it's, it's not the schools. The schools are pretty much the same. Uh, it's, it's not the curriculum. It's not broader society. It's all pretty much the same. What changed was how the teaching was done. Rather than using pen, paper, and pencil, and getting the kids to read and do hard work, they're giving kids Chromebooks. They give- they're giving kids iPads. And so learning is now much, much easier. And so less thinking and less hard work is being put into learning, and the corresponding finding is the kids are not as smart. Friction is the point. Friction is the point. Mm-hmm. And I think this says a lot about how we are interacting with AI right now. Again, I, I'm gonna speak for myself, I offload a lot of the thinking onto Claude first for it to give me the results, and I go, "Well, that's crap." Yeah. And then I'll go and do the thinking myself, and I know that's laziness- Mm-hmm… but it's just so convenient. Yeah, yeah. And it's also, like we, we've been trained to teach us, uh, as an ideation machine, that use AI as an ideation machine to be able to, you know, give me 100 things that I can then curate to the five that are actually the thing that matter. Right. Um, and a, a law firm, uh, from the retreat that I was at yesterday, um, one of the law firm leaders said that just like we were trained with Amazon, because we used to say, "Well, we'll go down to the corner store to get the thing," but then it was like, oh, no, I, I got trained to be able to just do… go to amazon.com- Click and then three hours later, then I, I get the thing. So I wonder if we're going through a similar process with AI, where we're just saying, "Oh, we're, we're just… The AI is the go-to," rather than doing the hard thinking ourselves. Well, that, that's the sort of analysis that people have with what OpenAI and what Claude is doing. The reason why they let everybody use it for free- is to create that dependency Yeah, yeah Is so that every time you have a question, you don't think for yourself Yeah You just go, "Hey, GPT, what's the answer?" Right And then it gives you an answer, and if you get complacent, you might believe the answer Yeah. Uh, close to home literally, uh, my, uh, now 14-year-old g- uh, daughter, um, she, uh, asks me questions periodically, and then she's like, uh, "Yeah, n- now search it up." Like, after I give her the answer. Uh, and I said, "I know things. Like, we don't need to search it up." She's like, "No, I need Google to tell me whether it's right or not." Wow. Then she asked me, she asked me an AI thing, and it was a very nuanced AI thing. And so I, I gave her the answer, and she said, "Now search it up." I'm like, "I'm not gonna search it up because I know what Google is gonna say is total BS because that's what the conventional wisdom is. Your dad is kind of an expert in this thing, like He's kind of, kind of an expert. He knows, he knows more, he knows more than Google does about this. But, uh, but you can imagine the search it up instinct is kind of what you're talking about with Gen Z- Yeah … uh, with it be- saying that, you know, their, their instinct is go to the iPad or- Yeah … go to the Chromebook- Yep … to be able to get the answer, rather than doing the hard work of mulling it over yourself. Well, no, I think that's, that's what's happening, um, to a lot of our lawyers as well. Like, they just open up one of these gen AI tools and says, "Hey, do a first draft for me. Hey, mark it up for me in a redline." Well, hang on, that's what we used to do as lawyers. Yeah. We would read that clause, we would think about the arguments, we would do our own drafting, and in the process of doing that, formulate the arguments that allow us to go into a negotiating, a negotiation room and win. Yeah. Um, and you, you may have heard that, uh, maybe in, uh, in the last week, uh, Berkeley, uh, College of Law, uh, released their, uh, their guidelines on AI. Have you heard about this? I have not. So, so they released, um, and they said, as a floor, as a, a blanket statement, no AI in any of the things. Ooh. You cannot u- you cannot use AI to summarize, you cannot use AI to, uh, ideate. Mm-hmm. Uh, you cannot use AI in, uh, to do grammar check. You cannot use AI in any of these ways, uh, but you can use AI to do research, uh, is what they had said. Uh, and, uh, it's maybe the worst use of AI, right? I was like, I was like, you can't use AI for- Right, right… grammar check or spell check? But- I, I… But I think what they're, they're saying is, like, if you wanna use something like Westlaw or Lexis or Clio that has research- Yes as part of their AI, then you can use it for that. I think is maybe what they're, they're trying to, to get at. They should be more specific. They should be more specific. Yes. Um, but anyway, so that is the blanket rule, no AI for almost everything, except a professor, uh, can be able to a- allow AI if the professor wants to. So they can opt in rather than opt out. Right, right, right. So, so anyway, so I, I think that Berkeley is, is doing what, uh, you are, uh, saying that, uh, we need our students to do the hard work, the friction, to be able to learn, uh, rather than, uh, you know, relying on the AI to give the first draft. Well, I, I think this is what's, like, the dangerous path that we're, we're walking down, right? Like, we've been talking about training of the junior lawyer for the last couple of years- Mm-hmm … because I think people have seen the writing on the wall. If we don't train the next generation, there will be no senior lawyers in 2027, 2028. Sure. And then who's gonna become the next generation's partners? Mm-hmm. So- Right now, if we don't change the habits, if we don't force junior lawyers to somehow still do the apprenticing that we've always done and put in the hard yards, what happens? So I, I've got… I vacillate depending on the day- Okay. Yeah, yeah … uh, to, to answer that. Uh, so part of it- Let's, let's set out the spectrum and then we- Yeah, yeah. So, so, and some days I agree with you, saying that we need to do better to be able to teach the, the juniors to be able to do things well. So that, I, I agree with you some days. Other days, though, I think I spent most of my billable hours my first, second, third year reviewing documents. That didn't make me a better lawyer. Sure. And really, um, if we use, if we let today, uh, juniors use AI to run more deals and to run, uh, smaller deals, right, and smaller litigation, then they'll actually have better training than I did as a, as a junior, and they'll actually be better lawyers than I was as a second or third year. Um- I, I, I think, I think I agree with you. Like, if you give the junior lawyer the ability to spend their time on a task that develop their skills, I think that's really, really good. But if instead you got them to do a task that is brain-dead, I think that's where the harm is. Um, so to- And that's always been the case. Well, yeah. I mean, with bill, with billable, with the leverage model, like, you do those brain-dead tasks, is gonna make me as a partner a lot of money. Yeah. Sure. Sure. And I think, I think, um, what AI has done is AI has given us more tasks that we can cognitively offload- Mm-hmm… and therefore more tasks Uh, being performed, though, by the bots, right? Being performed… That's what I mean. Being performed by the AI- Yeah… and therefore the human operator It's like driving a manual car. Yeah, that's right. That's right. Yeah. Like, I, I, I put my hand up, I can't drive a stick shift. Like, I, I did learn it once upon a time- Yeah … and I just stopped because automatic became the norm. Automatic. Yeah. And, and so I can't get in a sports car these days, um, and competently drive it out of the shop. And that's a skill that's been lost. Now, one argument is, well, maybe it's good that it's lost. Maybe it's unnecessary 'cause now you can spend your time driving thinking about other things instead of using your hand to, to shift gears. Maybe? Maybe. Um, this, uh, so we, we, uh, we have a trilogy of, uh, studies. One is the study you originally said that Gen Z is, uh, dumber than, uh- Yep … their predecessors. The second is the, um, y- at least Berkeley's policy saying that don't use AI because it's, it's going to not have the friction that we need. Let's go on to number three, and this is, uh, Dan Schwartz out of Minnesota working with- Mm-hmm Michigan. And his study, his hypothesis followed number one and two. The hypothesis that they had group one. Group one did something s- totally with their brain, no AI. Group two used AI. And, uh, they had four tasks. The f- first task for both of them is to say, uh, "Here's a bunch of cases, statutes, and regulations. Summarize and synthesize those." Uh, first one without AI, second one with AI. So the hypothesis was that that group, um, the people without AI was not gonna synthesize as well as the people with AI because AI's gonna be able to bring those cases, statutes together more easily than the, the brains. Okay. And then step two was to be able to take the AI away from both and say,"Here's a multiple choice test. Tell me what the rules are that you just learned." And the hypothesis is that the people without the AI are gonna do better 'cause they did the hard yards. They did the hard work. They did the friction. They did the friction. Right. Uh, and the people with AI would do worse. Number three then was to say, "Here are some facts from your client. Now apply those to the rules." Yes. And the thesis was, okay, li- I think the people that group one is gonna do, uh, the hard yards, they're gonna do better than the people with AI. And then number four, they gave both groups AI and be able to say, "Now revise your documents to be able to integrate them." All right. So that was the hypothesis that the people with the friction would do better than people with AI. Turns out not to be true. Oh. Uh, in the st- and th- their hypothesis was disproved, uh, because, yes, the people with AI did, uh, synthesis in stage one faster and better than the people without AI. Here's the cases, statutes, and regs. Let's summarize it better. People with AI did better. But with the second part, with the multiple choice question, they did equally well. There was no cogn- there's no better from the friction than with the AI. What's the sample size? Uh, sample size I think was 50 law students. Okay. So these are juniors. Okay. So it's, it's not as if, like, it was a small enough sample size that you can say there's a chance or big chance that group B just happened to be the smarter group. Uh, th- that's right. So 50 s- That's right. So I, I think a decent, decent size. So any- so group… They were equal on synthesizing without the AI. Okay. Yep. And then the third one was taking your client's facts without the AI and be able to apply those facts. Turns out the people with the AI did better than the people without the AI. Group B did better than group A. And the hypothesis of why that hypot- the, the original hypothesis was true, um, or disproved, was that maybe group B had a stronger mental model of what the answer was and kept that stronger mental model all the way through, whereas group A, um, went down rabbit holes that turned out to be false. Because they would follow one case, and then they would follow the other case that overruled it, and then they, they were a bit confused, and it took more muddling to get to the end. Mm. Mm. Mm. Therefore, they, they had a weaker mental model than the, the group B. So I wonder if this maybe cuts against your and my conventional wisdom. Uh, I'm gonna propose a different experiment. So, so if, if the… Was it Dan Schwartz? Yeah, Dan Schwarcz. So, so, so Dan, if you're listening, um- He will be… please listen. Mm. Um, I have a different experiment. Try this over a semester. Hmm. So try to give a group of students no AI over a semester versus a group of student AI with a semester, and then at the end of that, see if group A, who didn't have AI, that put in the hard yards over an extended period of time, does better than a group who had the assistance of AI over a period of time. Hmm. That'd be hard to do that longitudinal study, I would guess, because the temptation for Group A to cheat- Is to cheat… yeah, yeah, it, it's And, you know, you can, and like within a two-hour period, you could be able to constrain them. Mm. Uh, but I think over a semester it'd be hard to constrain them, right? I mean, you can, you can do this across two universities, I guess. Like, you can, you can do this at a university in the first world with access to AI and a university in the third world, which doesn't have access to AI. I guess. But then you have to worry about, you know, is, is this a higher grade school and a lesser… Yeah, these, these are difficulties. Y- yes, yes. You have to somehow normalize for human d- d- differences. Yeah, yeah. But I think, like, the, my hypothesis there is the, the, the, the difference is not gonna be apparent until you change the patterns in someone's brain. Mm-hmm. You rewire those neural networks- Yes and you become lazy. Yeah. That's right, and, you know, just like we all have plastics in our brains- Mm-hmm… uh, that we aren't even aware of, uh, the, the Gen Z that are gonna have brain rot, uh, don't even realize that they have brain rot. Yeah, yeah, and that's a danger. That, that's what I see as the biggest concern coming up for humans engaging in transformation. Do we sound like old guys that are- I think we do. You've got gray hair. I'm starting to get gray hair. That's true. That's true. I like, well, like I, I think these Gen Zs should get off my lawn and- These should be rocking chairs. I think that's right. I, I think it was where you were talking about stick shifts and, uh, you know- … but nobody knows what a camshaft is anymore and yeah. Yeah. Yeah. My, my dad would be very ashamed of me that I, I don't change my own oil. I rent- Oh, gosh … but, but I think I'm, I'm spending my time better, right? So I wonder if, if the next generation is to say, "Yeah, you with your old ways of doing the friction." Mm. Mm. Mm. Like, "No, I, I have autonomous vehicles." Right? This is not a s- a thing. I have autonomous LLMs. Um, this is unnecessary. I mean, the, the abstraction of labor, uh, uh, is the point. Yeah. Um, but in doing so, we, uh, are losing skills. Yeah. But I lost the skill to change my oil. Yeah. Um, but I, instead I talked to an executive board la- yesterday, right? When, in time that I would've spent changing my oil. Sure. Sure. And so the- I mean, there are people who specialize in changing oil- That's right… that you can, you know, give your car to, and they'll do the job for you. Yeah. I don't know if, I don't know if that, I, I don't know if that skill is, uh, is lost being a, is, is a bad thing. Uh, some things I, I think, like I, I, I saw this wonderful, uh, Twitter post, uh, I shouldn't say word Twitter, X post. Well. Um, some things never change. We are old. We are old. It's true. Um, and, and it, it was a, it was a lady who wrote something along the lines of,"Remember during COVID when everybody was at home baking bread- Yeah … growing plants, going outdoors, touching grass? We had it really good then, and we've all forgotten what that's like." And it feels a little bit like that, where the accelerating pace of technology is removing us from what makes us human. Sitting in a park, uh, drinking, and having good conversations. Uh, that's right, and I kind of feel sad about that. Yeah. Uh, touching grass, maybe literally. Getting allergic reactions to grass. Right. Right. I, I, I am allergic to grass, for those of you… Yeah. It's, it's weird. Like, what a weird thing to be allergic to. A- as a baby, um, I, my, there's a photo of me, uh, crawling in the grass, and I was, I would essentially be on, uh, like, like this so my knees wouldn't touch because I would have- Oh… yeah. So when, when I, when I had a baby, we took her outside. Um, she hated the sunlight. She was like… She's a vampire baby. Uh, and, and then, uh, we, we made her touch grass, and the moment her feet touched the grass, she'd be like, "Whoop." And just didn't, didn't like it at all. Didn't like the sensation. Uh, so we, we figured she would not be an outdoors child, and now she's the most outdoors of all children- Oh Maybe there, there was friction, uh, that you, that then- Ah she ended up liking it, right? You learn. Right, right. It's true. Yes, yes. Maybe the hard work of going outside. I, I think that that's, that's the lesson we've all learned today. Um, and so like, uh, are, are you, are you more optimistic these days or less optimistic? I don't know. I don't know. I, I, I think in terms of like optimism for the technology, I, I think like there's a lot of potential for the technology to do good. Mm-hmm. And I am still a big believer in that, but I am less optimistic about, um, humanity's ability to overcome the challenges that will be presented because I think we'll take the path of least resistance. Mm-hmm. Which means we'll rely on AI to do a lot more. And I don't know if that's good. Over-reliance on AI, I think might be a bad thing. I think about, uh, yesterday at this, uh, executive retreat, uh, a lot of, uh, partners were talking about judgment and how we need to instill judgment into our younger associates- Okay to be able, to be able to do things. Great topic. It is, it is a great topic, and it relates to what we're, uh, you know, discussing. Um, but every time I hear somebody say that, I think, is the judgment that that partner is wanting to instill in the associates, is that something that's just ephemeral? Mm. Or is that a data point that can be quantified and then put into an artifact that can be ingested? Mm-hmm. Because what we were talking about with taking our lawyers' brains and put them into this, a lot of that may be called judgment. Mm. Right? Mm-hmm. Because, like, you can't argue this because that judge is gonna hate it, but Judge B is gonna love it. Mm. Mm. Is that judgment, or is that just a data point? Mm. Or you can't argue this in a deal around the, you know, the M&A deal because it's gonna scuttle the deal for this party, but it's gonna be great for that party. Mm. Is that judgment, or is that a data point that you can put into an artifact? So I don't know if you read this, but there was a litigator who, uh, I think he's a famous litigator at Paul, Weiss who went over to Milbank. Okay. Um, and he argued in front of Supreme Court, um, uh, I forget for which cases. Um, uh, but he used Harvey. Oh, yeah. Um, uh, Neal Katyal. Yes. Yeah. Yes. Um, uh, he used Harvey to essentially simulate all the arguments that the justices would, would ask or all the questions that they're gonna ask. And, and he said the LLM, and, uh, I'll attribute this to the LLMs rather than Harvey- Yeah um, managed to verbatim hit some of the questions that the justices asked. Like, I think that's incredible, that there's so much content out there to have trained a large language model to mimic the thinking patterns or at least the word usage patterns- Yes… of these justices. That's right. And, you know, you can imagine that what Neal Katyal has, uh, gathered over the course of his, uh, his entire incredible career could be seen as judgment. Mm. But as demonstrated by the large language models predicting what Judge, you know, X and Y and Z, uh, could do, um, is that really judgment, or is this just quantifiable? Mm. Right? Mm. And, and I- I think that there's… I think we're gonna have a reckoning as to s- you know, what any particular lawyer thinks is their special sauce. Mm-hmm. Because is it more special than you can quantify with the LLMs? I don't think humanity's that smart, to be frank. And, and so therefore, I do think a pattern-matching machine can do very well emulating any human. Yeah. Um, so yeah, with enough data, with enough- disclosure of whatever it is that we're thinking, writing, et cetera, a large language model can do a very good job of emulating any of us So let's put this into concrete, uh, examples. Uh, and this is actually something I'm building using, uh, uh, with- on the ALEA and FOLIO side. Mm-hmm. Uh, where, um, I can't remember if I've told this on the pod, so I'll, I'll repeat it. Hmm. The, so three books that I'm co-author of, uh, legal textbooks, and I'm building a tool called FOLIO Insights to be able to extract the insights from those books. Okay. So an insight from those books, an example is, start a deposition with open-ended questions. What did you do today? And then end by closed ended and say,"Well, you did this thing, right?" Yep. And so then you, you cut them off, right? So ask open-ended questions is an insight, and then go to closed-ended questions is also an insight. Yep. Um, that is helpful not only for depositions, which is also a tag, but then also for cross-examination at trial. Absolutely. Right? So the same, same technique can be used in both. So really extracting tens of thousands of insights from these books, and then putting them in a quantifiable knowledge graph to be able to apply to depositions and to trial and, and, and. So once they're all quantified, then a lot of that starts looking like judgment that is truly in now in the ar- artifact that can be reasoned with, with the large language models. And so as we extract those insights from textbooks, like I'm doing, and from discussions with lawyers, and from oral arguments and all of the other data points, I, I wonder, can we as humanity even keep up with all of those extracted insights? And is there gonna be anything left in our brains that is not quantified? I don't think there's gonna be much left in our brains that aren't gonna be quantified. I think the challenge is extracting what's in our brains into a digital form. I, I don't think the challenge is for the machine to absorb whatever is digitized Say that again So, so I think the, the challenge is taking our brain and storing it into zeros and ones. Yep. I don't think the challenge is for an AI to, um, generalize from those zeros and ones into a neural network. So, uh, head to database- Yep … database to- Neural network … neural network. Yeah, yeah. That- that's- that's the process. Right. And n- now that, now that you've put it into an artifact, say a knowledge graph, then the neural net can be able to go through guardrails. Essentially the skeleton of the process, saying open-ended questions, then close-ended questions, and it will do it deterministically, therefore reliably. And so really as we put more of those things into the artifacts and give the AIs more of those guardrails and the skeletons upon which to go, um, is there much left for lawyering? Well, and th- this goes back to the el- to the 11-year-old's third point, which is the way they're gonna protect themselves is to stop using AI now. Stop that transfer- Right… from their brains into a digital record, because that's the hard part right now. Yeah. Smash the looms. Smash the looms. But that didn't work well for Ludd, right? Well, I mean, if enough people do it, uh, then, then maybe it works. I guess. You just need that 51%- … critical mass. That's right. Right. But I think people… Like, I'm, I'm building this right now, right? And, and in a sense you're building this right now. Mm. And I, I think that the, humanity's desire to be able to make things better, I think will overcome the 50% or whatever the number is that the Gen Zs say, "Stop doing it." Right. Right. Because, because the upside is too great. Do we have enough faith in humanity that we're gonna do the right thing? Well, I mean, you and I are doing the right thing. Like, we're b- like, of the multiverse of all the bad, uh, you know, where we kill our, our humanity, right, we're pushing us toward the good where we actually go better. And so I would argue that if we don't build the good, then the bad will win. So I, I… That's at least what I'm telling my children, to say, "No, I'm, I'm pushing, Dad is pushing us toward the good future", rather than leaving it to the bad people to go to the bad future. And there's a blueprint of the T-1000 on your wall, right? Exactly. Like, look, if I don't build this, not a Terminator- … but a protector. No, that's r- Right? That's right. That's right. Uh, one that really, uh, is imbibed with human values. Yeah. Yeah. And it is… I'm not gonna put it in an Arnold Schwarzenegger voice, but- Right … you can just imagine. Ah. You know? Yeah. Well, that's T-2, not T-1. Yeah. Right. Right. Yeah. And, and so, like, I, I, I do like the idea of we are building something that is good, that ultimately is gonna be beneficial to humanity, and therefore, y- you know, rather than money-hungry billionaires who are gonna try and build this and, and monopolize it, we're trying to build it, so that there are more people who are able to benefit from this. I, I like that idea. But I don't know if this is how the economy is structured. Yeah. I don't think the people who are holding the ownership of the GPUs will allow their GPUs to be used for the benevolence, uh, and, and the benefit of society. Uh, a really good insight I heard, uh, from a podcast is that, um, Anthropic, uh, you know, OpenAI, uh, everyone wants to be able to have the highest value per token. Yes. And so then, uh, I think we've talked about this in the past, that coding has high value and legal has high value. So to your point, like, I don't know if the GPU people will want us to be able to, um, do the right thing. Um, isn't societal OS being the law, isn't that a good thing, uh, that uses a lot of tokens that will make them a lot of money if we say, "Here are the agents that are now following the law," because this is legal by design? Isn't this a, a good use of tokens that also pushes toward the good? It, it, it reminds me a little bit of Asimov's three laws of robotics. Yeah. And, and now you're adding a fourth law, follow all laws. Yeah. Um, I, I think it's… The complications, the overlaps, the inconsistencies of the human systems is gonna confuse the crap out of any AI trying to comply. Well, so, and, and it confuses the crap out of many lawyers- Yes … which is why we have lawyers, right? So like, and I, I may have talked about this in the past too, but a friend of mine works for a cloud service organization. She said that she's in charge of bringing down, uh, CSAM and other objectionable material. Um, and she- What's CSAM? Uh, c- uh, child sexual exploitative material. Okay. So like, the child pornography. Uh, so anyway, so but in, uh, she said Nazi paraphernalia has to come down under German law- Yeah … because it's on, uh, but under Texas law, it has to stay up. Mm-hmm. That's free speech. Mm-hmm. She said, "Whose laws do I follow? Whose human values do I follow? German human values or Texas human values?" Depends on the geography.'Cause like- Depends on- But, but it's, it's worldwide. Okay. So, so she says like, "I can't do both. So I either have to pull it down under German law or keep it up under Texas law. I can't do both." Yep. So this is what you just described. Like, it's befuddling her as a human lawyer. She's like, "Well, which do I do?" And so but what if you were to have a database with the entire corpus of German law and Texas law and Swiss law and, and, and, and say, "This is the least worst option. Here are the penalties under German law. Here's the penalties under Texas law, and here are the all this, and here is the least worst job"? And that's kind of what my friend as a human did- Mm … but did it inefficiently. But maybe the tool could do it more efficiently. Hmm. It's one very specific use case, and it's one use case which has very low likelihood of causing harm to anyone. Well, I mean, Nazi paraphernalia will cause harm, right? Yes. Yes. Yeah. That's true. That's true. I mean, u- under German view it would, right? But it, but one can imagine we, we create disparate laws all the time- Mm-hmm … in different jurisdictions, even within the same jurisdictions. Yes. You have one case that says this, and the other case that says the other, right? Oh, I mean- So that's- … you, you have, you have federal and state laws that are in existence And we have circuit splits, right? There, there is all sorts of ways that you dispute. And so, uh, kind of the muddling, getting back to our early discussion, the friction kind of is the point. Hmm. Because then we can say, "Is this circuit right or is that circuit right?" with the circuit split. Society debates it, and then the Supreme Court decides. So I, I wonder if the AIs can kind of go through a similar process o- of taking these disparate, uh, results and be able to try to harmonize them. You know, there's, there's value in humanity, right? Like, like, you know, we go back to law school. We follow the common law, but then the common law is overlaid with equity. Mm-hmm. And equity is supposed to be that human element to say, "What's actually fair under the circumstances?" Yeah. What are we supposed to be doing? Not just follow the letter of the law, but look at all the circumstances beyond, and let equity impose a better set of rules or better set of outcomes depending on the situation. Yeah. Um, and let's say we get to a future where AI does sit in a position of power or a position of judgment. Do we trust AI to do this?'Cause this is where we're heading, right? Like, as AI continues to get better, as we relinquish our free will to AI- I, I, I wonder if it, uh, the answer to your question is, is what relates to Neil Katyal realizing that the AI figured out what the justices were going to say perhaps better than he could. Uh, and I, I think that as we have the AIs do more of our work, uh, sussing out the law, I wonder if it might do better than our frail human brains might be able to. Again, we're trusting an AI. That's right, and if it's ASI, as the 11-year-old said- Yes.… we can't trust it because it will kill us. Well, it has no reason to keep us around. Right, right. Not that it wants to kill us. No. It just doesn't care. D- did you hear Geoffrey Hinton's, uh, most recent? He's more optimistic because if we imbibe, uh, imbue the AI, the, uh, ASI with motherly instincts, uh, the, that he said that, um, you know, mothers are the only instance where a, uh, a weaker being, a, a baby, has control over a stronger being, the mother. Uh, and that's because motherly instincts say that, you know, "I need to protect this baby, uh, otherwise it will die." So if we imbue the ASI with this kind of motherly instinct, uh, to be able to say, "Yeah, these poor humans, we need to protect them, otherwise they'll die"- Mm. Mm-hmm… uh, that's maybe a way to be out of the problem that you pointed out. I, I, I love that analogy. I love the idea. I hadn't heard this before. Yeah. Um, there was a, there was a similar sort of, like, idea about, um… And I forget who said it. Like, having ASI treat humans like a beloved pet. Hmm. And, and it's like, well, y- yeah, I guess that's- Until you put it down."Stop yapping." Right. Right, exactly."Well, you've lived 12 years- … and that's a long time for a dog, so we're gonna make a little trip upstate and…" Yeah. Doggy heaven is very lovely. Right. Right. Um, so I mean, uh, yeah, maybe. Maybe. But I, I think we're getting closer to that reality, right? Like, we, we talk about how AI affects the law, but at the point of t- of time when AI competence goes beyond humans, it's not just the law that matters, it's everything that we do. And this goes to, you know, equity being what's right, right? Mm. Uh, but what's right according to judge one is not what's right according to judge two. Yes. And that's, that's the difficulty is that, uh, the law gives consistency and equity gives inconsistency, which is sometimes better. Quite deliberate. Yes. Right, right, right. Deliberate. And so the, so, uh, uh, I, I guess what's, what's better is, is, you know, judge one versus judge two inconsistent, or the large language model doing essentially a compression scheme of all the judges- Mm… collectively- Mm. Mm-hmm… as to what's right I don't know is the answer. Yeah. I don't know. If I had an answer to that, I'll probably make a, a larger investment in the Anthropic IPO that's coming up. So let- let's, let's continue this, uh, what's right. Um, and I, I, on Friday I'm gonna be meeting with my favorite vibe coding, uh, Roman priest- Oh, yes … named Father John D'Orazio. Yes, yes, yes. So he, uh, he's gonna- in Boston, so I'm gonna be going to Boston in a couple of days, and we're gonna have breakfast on Friday. So anyway, so part of the goal of what we're doing is, you know, of course, he and I are, are tagging up things that are right within the, the, you know, eventually the Vatican archives. And so we're saying, well, what is just? What is equitable? Uh, these are, these are all tags that we have- Yep … and we're pulling up Thomas Aquinas, et cetera. So, um, you know, you could be able to say, uh, you know, with, with Vincent, you could s- do a 50-state survey, say, um, "I'm gonna lay off 40% of my workforce from Block. Does that violate any of the 50 states' laws?" You can also say, "Does that violate the EU law?" And eventually, what if you could say, "Does that violate ethical, faith traditions?" Hmm. Uh, whether you're, you know, your faith, you're Christian- Am I good human? Oh, you're right, a Christian or Jewish or secular humanist or whatever, whatever the, the ethical things that are beyond the governmental law. Um, and there's a, I may have talked about my, the venture capitalist, uh, uh, faith tradition person that said that, uh, people, companies that follow faith traditions in their companies make more money. They outperform the S&P 500. Love your customer, love your employees makes more money than not. So what if you as a company were to say, "Does this violate US law and EU law and faith traditions? I wanna be ethical 'cause I'll make more money." Uh, I don't think ethical companies ever think about it from that direction, though. They don't go, "Yes, I will be ethical so I make more money." That's right. Ethical companies don't, but unethical companies wanna make money, so maybe we'll nudge them toward ethical. There's your incentive. Right. It's like, "Hey, OpenAI, you should be more ethical. You can make more money that way." Look at Anthropic, what they're doing. Well- They're hanging out with the Pope and- Yeah … and, and arguably they'll… maybe the Pope will use Claude to be able to draft the encyclical. Uh, that's- Maybe- That's the rumor… maybe they did. Yeah. Maybe they did. And I, I mean, look, I, I, I have heard rumors that all of these companies are unethical at some, to some degree. But, but then- Kind of like every CEO. Yeah. You can't get to be CEO by being a nice person, right? The, the, the old study says, like, the psychopath are the best CEOs. That's right. Um- And, and, uh, lawyers are typically more psychopathic than the average citizen. Some people just say in control of your emotions, Damien. That's right. Sociopath, psychopath. These are social constructs, I think. I mean, what a, what a, what a broad spectrum. We, we went from, like, well, you know, what's the first step to adoption of AI in a legal practice all the way to, like, are humanity, is humanity gonna survive? I think that's what a early summertime in Central Park, uh- Oh with good friends and beautiful weather might do. This is not even alcohol. That's right. It's not even… Cheers. Cheers. This is great fun. Let's do it again. For sure. I'd like… Shall we, shall we end with some optimism? Let's do. Let's do. Um, shall I start?'Cause you always start- Yes. Yes, please … with optimism. Um, look, I, I am optimistic that more people are getting together trying to figure out the direction of this, that this is now, um, the hottest topic that everyone has to contemplate and figure out. Um, everything from "how do we make use of it in our personal lives" to "how does this affect the environment, the drinking water, the oxygen that we breathe." Mm-hmm. Talking about it, thinking about it is the only way that we have a good outcome out of this. So that's my optimism, more people talking and thinking about it Amen. And I think that your optimism, more people talking and thinking about it, leads to my optimism where, uh, just yesterday, um, I was talking with this, uh, the AM Law 50 law firm executive board, and they told me a- about a big win that they had. And it was a win, uh, using AI in the substantive law, which is a win, but they actually used it to win the business, because they were going back and forth with the, uh, the, the customer, the Fortune 50 customer. And the Fortune 50 customer was saying, "Yo, your numbers are too high. You need to change it." And the law firm used AI to be able to say, "What are good strategies to come back- Mm… and be able to say, 'Maybe if we could shape it this way, and maybe if we do the personnel this way, then we could be able to win this business.'" And they did. Mm. Uh, and so then they're, uh, they're using AI substantively, but they also won the business, business of law side, uh, and using AI to do it that way. So anyway, so b- that story led me to go back to my PowerPoint and slash a whole bunch of things which was me trying to convince them that AI is a thing, which I had been pounding the streets for the last three years saying, "Hey, AI is a thing," to lawyers that say,"No, I'm gonna practice the old way." But I realized with this story that that's totally shifted. Yeah. Everybody is now AI pilled, and now I can slash all those things and be able to talk about the more interesting things. So I'm optimistic about, uh, people talking about it, yes, and also the sophistication of the discussion- Mm … saying not if it's going to happen, but, yes, it is going to happen- But how? what are we gonna do about it? Exactly. Yes. The, the, the, the actual steps involved. Um, it's funny you, you mention that example because, like, for me, I've been using Claude to help me write a speech for a demo this Friday. Mm. Mm-hmm. So I'm doing a, a demo for, uh, Oz's SKILLS.law thing where we're gonna be presenting our tech for negotiating and redlining. Mm-hmm. Um, I'm not gonna explore that on this podcast, but we'll talk about it some other time. Um, and what I'm finding is Claude has its own voice, no matter how much I try to nudge it. It does. And it is not my voice. Yes. And, and, and this is what's, um, kind of very cool about it. The more you use AI, the more you recognize where it works and also where it doesn't work. One of my favorite phrases is, "Your AI is showing." You go on LinkedIn now and you see the post and you're like, "Genuinely?" Maybe take a look at your AI is showing!. It's showing. Yeah. So, uh, yeah, I, I think, I think we're all becoming more sophisticated. I, I think that's right. So, so it's wonderful. Uh, Horace, every time I talk to you, I end up happier than when I started Oh, Damien, likewise. What a wonderful day. It is. And, and we should wrap it up there, folks. We'll see you again soon. Thank you for listening to us, and we'll catch you in a couple weeks. Thanks everybody. Bye everyone.