Ep#107: Revolutionizing Patent Law with AI ft. Evan Zimmerman (CEO & Co-founder, Edge)

[00:00:02] Taylor Kenerson: it. Hello, everyone. My name is Taylor, and I am here with Adil and an amazing guest today, Evan Zimmerman. He is the cofounder of Edge. It's a patent assistant helping attorneys and investors write patents painlessly, and I'm sure he could definitely dive more into this. 00:00:19 Exploring the Influence of AI in the Legal Field [00:00:19] Taylor Kenerson: He's also an investor and has a lot of knowledge in IP space and legal space. Thank you so much for joining us, Evan. [00:00:26] Evan Zimmerman: Thank you so much for having me, Taylor and Adil. I'm excited to, to be here. [00:00:32] Taylor Kenerson: So let's just jump right in let's jump right into some of these challenges that we're seeing. of people are really scared and have a lot of questions around AI and its drastic growth and especially in the past year year couple years. Can you touch on the influence of AI, especially in the legal field, and how that affects companies and how you as a founder can leverage the technology instead of seeing it as a competitive? [00:01:03] Evan Zimmerman: Yeah. Definitely. And, I mean, I'd say there are 2 things that you're touching on. There's the social impact of AI, and then there's as a company. So in terms of the social impact, we're already starting to get research that comes out on this. 00:01:15 The Impact of AI on Efficiency and Equality in Law [00:01:15] Evan Zimmerman: There was actually a paper that came out just last week testing this on lawyers. And what it's finding is that it helps everyone, but it helps the lower performers the most. So a lot of what you're seeing is, yes, a huge amount of efficiency, but also a lot of equalization. And what we see, for example, with our customers is that for the people who are more skilled, it basically handles a lot of the easier tasks that you can do higher order things. So if you think about in our use case, we help people get patents. And so what that means is you don't you know, if you think about the work people do with a patent, 90% of the work is spent on 10% of the value, which is crazy. It's just because patents are so long and there's so many things you have to do. With us, what happens is that you can decrease the time you're spending on, like, the background section, the routine parts of the written description, and so on, and really focus on things like the claims and the prior art, which are really the higher value asks. On a social scale, I actually wrote an article about this last year for an academic publication, the Dakota Digital Review. And we saw this with, like, electricity, the industrial revolution, a lot of other historical technologies where, really, these things take some time, but, ultimately, they end up being a net good for society. They very oftentimes end up creating jobs on net, although they change certain jobs and eliminate others. And a lot of times, what you see is that things that used to be important, they're still around, but they're a smaller part of society. You know, farming in the 18 fifties was over 90% of all jobs. That was just what people did. And today, it's less than 2%. But there are still farmers. There are actually more farming, more farmers than there have ever been in terms of the number of people, and there's also more agricultural output. And that second thing, the agricultural output is actually what's more important. And so I think as founders, what's really important to say is, well, how can I think about the output that I'm giving and using AI to really maximize the value you're creating for your customers? That'll help you actually charge more for doing less and create more opportunities because of that for you to have your own vendors, and that's really how you get the improvement in the economy overall. 00:03:44 Adapting Pricing Strategies for SaaS Companies Using AI [00:03:44] Evan Zimmerman: I will also say as a founder, because you're doing more with less, your customers may be doing so too. And so charging per seat, which is what a lot of SaaS companies do, may stop being the most efficient pricing strategy. So that's another thing to think of too is how does AI not only change your product and change the way that you run your company, it's also going to change potentially, not necessarily, but potentially your pricing model. 00:04:11 The Value of AI in Enhancing Creativity and Efficiency in Pricing Models [00:04:11] Taylor Kenerson: I love that you touched on the pricing model aspect. I think, especially in today's world, you can't just look at 1 KPI against your pricing. You need to look at the holistic model, and there are other value adds to include in your pricing model that are not standardized amongst a lot of companies. So that is really important. And then on your point about AI, I think the the real value AI has is allowing that creativity to happen that humans have, and let the AI do the repetitive tasks. And like you said, those 90% of the tasks that are monotonous and repetitive and just you have to do. [00:04:46] Evan Zimmerman: I think it's really important that you touched on creativity. I'm kind of a nut about it. And one of the things I think is really powerful about AI especially is that it's nondeterministic. So if you think about, like, a spreadsheet, if you say add up these numbers, it's gonna do that because it's basically a fancy calculator. AIs use probability. And one of the most important aspects of creativity is randomness and, actually, believe it or not, waste. That's actually an important part of creativity too. And so the I think that when we well, I'm sure we'll talk about this a lot, like we talk about with our own customers. You're gonna use different systems in different ways. And the technology today is better at different things than it's gonna be better at even a year from now. And so I think that you should also be thinking about what do I deploy these systems for And for us, of course, creativity is a really great application because it can suggest things that maybe you wouldn't have considered. Is it gonna be the final idea? No. Of course not. But it could spark an idea that takes you down a direction that you've gone before. 00:05:51 Understanding the Relationship Between AI, IP, and Copywriting [00:05:51] Taylor Kenerson: A 100%. And I think it it is that seed, and it can be that catalyst for that creativity that is is needed in today's world. Can you touch on now, like, AI and its relation to IP and copywriting? I am so curious about this. I see so many things that are. Is this a duplicated image? Are you allowed to use this? So speaking as, you know, obviously an expert in this area, but for our fellow founders, what do we need to know about IP and AI and all of that? 00:06:19 Understanding Intellectual Property and Fair Use in the Computer Industry [00:06:19] Evan Zimmerman: Well, I mean, I think the first thing that you should know is that if you want to go and have a patent, you should ask your patent attorney to use our software because you'll get it more efficiently. But the other thing that I'll say is I I think what you're touching on is these lawsuits. Right? And they're really interesting because this is this notion of copyright is something that's come up quite a few times in the history of the computer industry. And I'm gonna dive a little deep into why and why it's a little strange and where these problems are are going. But I will say that for now, it's not something I would worry about too much. After all, you have for a lot of people who are using, like, OpenAI, for example, Microsoft has agreed to indemnify its customers against copyrighted claims. And I think Adobe asked you the same thing for Firefly. So I think at least in the short term, you're okay. Basically, there are 4 types of intellectual property. Athens, which is what we do, is one of the 4 major types. Copyright is another example of a type of intellectual property. And these types of IP have existed for literally centuries. And the core outlines the trade offs you know, there have been changes over the years, but by and large, it's the the same concept that it's been for about, like, you know, 300, 400 plus years. The problem that you have is that the computer era has introduced what I would say are 2 new types of intellectual property that don't neatly fit into any of them. They are data, and they are software. And the challenge is that because they don't neatly fit any into anything, but you have to have it fit into somewhere. Essentially, both of those have been jammed into copyright, and, occasionally, software gets jammed into patents, especially with AI. AI patents are actually you're twice as likely to get them as a normal software patent, believe it or not. But the challenge with this is that copyright is has a few weird characteristics. The first one is that it's extraordinarily easy to get, but the trade off is that it's very weak. There it's like Swiss cheese, and the main limitation was called fair use. It's a phrase that you've probably heard thrown around a lot, but the short description of what fair use is is copyright stops you from copying, literally, people's work and distributing it, especially for profit. So for example, if you had a book, I couldn't just copy the book and then sell it to other people. I would need your permission to do so. And on top of that, there are exceptions. Like, if I want to make something that's similar or if you have characters like, if I wanted to write a sequel to the Harry Potter books, I couldn't do it. But the fair use, the exception is that, well, I can't necessarily stop people from doing things that are related to my work. So here are examples of things that would be legal, and there are actually cases on all of these. If I wanna do a parody of Harry Potter, I would be allowed to do it. If I wanted to make a dictionary or encyclopedia of Harry Potter characters, I would be allowed to do it. Fan fiction, which is not for profit, at least in the United States, perfectly legal. So the question really is, for these models, is there a, is there a fair use case for them? Because we learn from things, and that's totally legal. So, to me, there's really 3 questions. And I'm sure you've seen about these lawsuits like Sarah Silverman, New York Times. You know, what these lawsuits touch on is really 3 distinct issues. And, by the way, it's a 4 factor test for fair use, but there are 3, for me, real issues where some of the 4 factors matter more than others. And for your listeners, we're actually publishing a newsletter on this next week, at blog dot with edge.com. The newsletter is called nonobvious. 00:10:21 Target Customer Segments and Strategies for Scaling at Edge [00:10:21] Evan Zimmerman: You should definitely subscribe to learn more about our analysis of this question, but to preview it. The 3 distinct areas are, was there copying involved in the legal in the actual, training data? So that's the first question. Did they need a license to, like, buy the data? The second question is, is training an acceptable fair use? It's also known as a transformative use. But then there's actually a third question, which is, let's say that you use, like, AI OpenAI just today came out with a new video tool called Soarer. And let's say they use Soarer to create, like, a Super Mario Brothers video, and then you do something that's not okay with that, like, you sell it. Is the is the technology provider responsible, or is the person who used the technology in an inappropriate way the person responsible? Because, you know, you can use Photoshop to create copyright infringing images too, but no one would ever say that Photoshop should be responsible. There were actually a few people who said that when it first came out. So I think that my so our analysis goes into the objective things. My personal take is that these are all fair use. I think that the case is is rather, is really rather weak. And one of the piece of evidence for this is that you see the headlines when someone gets sued, but what you don't see is the headlines a few months later saying that the judge threw out the case. They threw out almost all the claims or narrowed them. The Sarah Silverman case, which I mentioned, was super high profile, and the judge is already told, I think only one claim survives, and they're being required to revise it. And remember, just because they let you continue with the claim, doesn't mean that you win or that it won't get thrown out later. It's just the judge saying that it's kind of ridiculous. So, I I those that's I that's sort of what I would say there is you should definitely be looking out for it. And this is an argument that some people have for open source. But, ultimately, I think that right now, I would say the risks are not super high. And we can delve more into the details if if that's interesting to you, but that's my sort of overview, my summary. [00:12:36] Adil Saleh: Very interesting. And, I encourage everybody to, you know, follow that newsletter. This is writing up, and he's lately speaking a a lot about this. Okay. Just quickly, at at Edge, what kind of, customer segments, that you're targeting for this year, and, what kind of things you're doing for scale? 00:12:56 Customizing AI Solutions for Better Customer Experience [00:12:56] Adil Saleh: Like, any initiatives have you taken starting off? What kind of post sales journey? Just give us a little bit about it, and then we'll build up. [00:13:03] Evan Zimmerman: Sure. I will say on the customer second question, an important learning I actually had relating to this is you need to be open. You may be surprised. So we create software that essentially helps practitioners write, practitioners because, of course, they can also be patent agents, for example. And the thing that we thought was really interesting when we first announced this product, we actually heard a lot from people inside companies. This was shocking to us because we thought it was gonna be all attorneys who reached out and then a few inventors who wanna cut out their attorneys, which you always see in legal tech. A lot of people wish they were paying us some legal fees. What actually happened was we did get those, but we got a lot of people inside companies. And so we realized that the in house counsel is actually a viable customer too. Because we also have a pretty, like, expansive product suite. It's not just like a patent writer, which is what you see from a bunch of other companies. We are a more comprehensive solution, which is one of the thing that makes us unique in the market. And so for us, like, some of these are very large enterprise deals, from very, very large companies. And, of course, there are also law firms, and we serve law firms from sole practitioners to large firms that, you know, have over a 100 attorneys and and agents. And what we've sort of found is that the most important thing is to, you know, not necessarily that we create custom solutions for people. We allow them to actually have some scope of customization to their need. So one that we've done, which is pretty special is style personalization. And I think this is something that you should really consider for lots of AI companies is patent attorneys have their own styles. This is one of the reasons why pre AI solutions were not actually good enough for patent practitioners. And we allow you to tell your style to our assistant so that they'll be able to customize the output to their needs. Now this is actually a thing that I think is really special about AI so they can take instructions and implement them in unique ways. So you can imagine, for example, if you have, say, some kind of agentic sales tool. You know, people have different styles, what they found has worked in their industry, things that represent them, that makes things in their voice. So maybe think about ways that you can allow people to customize the sales output instead of just doing something on your own in a house style. You can imagine the same thing for copywriting. Right? Maybe you do marketing material. Is there a way that you can make it so that, you you know, your if your company has a voice, that everything can be in their company's voice or that it's in compliance with their brand guidelines. Those kinds of customizations make things sticky, then they also make people feel more comfortable. Because one of the things that you see over and over with AI tools is if the output is not good enough, I have to edit it. There's a point where I spent so much time editing that I would have been better off just running it myself the first time, and that drives churn. You really don't want that. You want people to feel we call it the zero shot internally. It's from zero shot learning, but there's zero shot output too. And so we really try to make the zero shot output really incredible, formatted the right way. That's another thing that you can do that's differentiating from ChatGPT is to have it integrated into an interface that's formatted automatically in a certain way. Like, the less prompt engineering your user has to do, the better off you are. So that's another thing that we've learned that really feeds into our product is that you're not just gonna say, do something for me, and then you'll have to spend a lot of time trying to make it work. Instead, it just does it the right way the first time in the USPTO approved format. [00:17:01] Adil Saleh: You just need to make sure you're smart at prompting, you know, as a user. [00:17:06] Evan Zimmerman: Yeah. And maybe tuning as appropriate, Or you're making your own model. I mean, that's getting cheaper all the time. 00:17:11 Building Specialized AI Assistants for Legal Industry with Evan Zimmerman [00:17:11] Adil Saleh: Mhmm. Mhmm. Cool. So you you're you're building something on top of GPT at the back end. Like, as you mentioned, that it it basically it it catches the tone and, you know, the style of, you know, for example, lawyers. You know, like you said, they have, like, different way of, you know, delivering corporate, you know, contracts or all of these these writings and whatever they have. So how does that like, how does on top of chat GPT you're doing in order to make sure you mitigate it and make it as perfect? Because a lot of these, tools that we see, out in the market, they're just, you know, getting it a better UI and getting all the models that they have. They're not building anything intelligent on top. And they're just leaving it to the users, you know, prompted, and they are better off using chat gpt, you know, native version. [00:18:00] Evan Zimmerman: Well, I think that one of the fundamental problems is that a lot of the products I see out there, they're actually too general. So part of what we do, which is different, is think of how I mean, it's a huge market, but think of how specific the use case is. It's intellectual property. And we thought and we know that intellectual property proceeds in stages. And so we can design you know, we don't design an AI. We design our product. Sometimes I talk to customers. I don't even use the term AI. I use the word assistant. And so what we do is because we have a focus, we're narrowly focused on a segment, a big segment, but a segment nonetheless. We are able to some of its prompt engineering and fine tuning to make it specifically produce the output. But the other thing is about the data. You know? Think about what we do. Right? I mean, it's not like it needs to know what Mario is, not to pick on Super Mario Brothers, but that's not really relevant to our assistant. Right? If these are technical matters that deal with certain legal things. And by the way, even if you think about legal tech, because we're focused on a specific segment, it doesn't need to know anything about securities regulation, about m and a. If you think about contract law, it doesn't have to know anything about contract law, which is a huge area of law. So the I would say that that focus is really key to what makes what we do special, and it also provides a base for as we want to grow to do that in a way that is detailed and specific. So I will also say one thing you mentioned about other companies, but I think you also have to think about the fault tolerance. So in our case, right, lawyers have 0 fault tolerance. A mistake by a lawyer can cost sometimes 1,000,000,000 of dollars. Whereas if you're doing, like, you know, let's say, customer support, obviously, you want your customer to be happy every time, but the fault tolerance for let's not say the whole interaction, but let's say, like, an individual message is is not so is not so low. Right? Because you can have, like but, again, you want the inter your interaction, you want to be good. But if you send a message and it doesn't necessarily solve the problem right the first time, well, that's the status quo with the customer service rep no matter what. Right? Is they oftentimes don't solve it in the first attempt. And then over the call, they make 2, 3, 4 attempts. So I think you have to really consider the that fault issue and also what type of faults are acceptable. So the thing about AI is that it makes different there are different types of errors that happen at different rates, with also different levels of severity. So, hallucinations are an example where Hallucinations will happen depending at at different rates, depending on what task you are asking it to do. And Mhmm. Depending on what type of thing you ask it, it may be more likely to hallucinate. So is it so is your use case 1 where there's a lot of hallucination risk, or if you get something wrong, it actually doesn't matter all that much? A lot of times, if you're asking to do, like, a pure computational task, you can make it better with prompt engineering, like asking it to write code to do the computational task and then executing that code. 00:21:27 Discussion on Leveraging AI in Software Development [00:21:27] Evan Zimmerman: But still, like, that's a place where you can get error, but not hallucination. Or even if you ask to summarize something, that's a place where maybe you can make a mistake. There's recall. And so those are things that I think are really important to consider, especially because very often a space is very big. It has lots of potential problems you can address. And so maybe one thing you should ask yourself when designing a product or maybe your next feature is, of all the things that I can choose, is this something that leans into what AI is good at right now and away from things where it's weaker? And, also Is that multiplied by the false tolerance for that type of error in that context? Because we think about that a lot. Even in the legal world, there's actually some level of fault tolerance. So for example, if you file a form wrong, not every form, but certain forms, you can file an amendment. So, the the I'll give you one example where this happens all the time. Although, we do this deterministically is filing an invention information disclosure statement. So when you file a patent, you have to submit a list of references. And when you submit that PDF I mean, people file corrections all the time. If you look at a prosecution history, sometimes you'll see 3, 4, 5 different IDSs that have been submitted at different times. So you can have some error there. Now that also happens to be a case where you can do it deterministically, so there's no reason to use AI. But, still, that's a thing to consider. And I will say one last thing to that point, which is you should use AI for the things where it provides value. We software has been we've been making it good for, like, 60 years now. Software is pretty great. There's still a lot of stuff for software to do. And, think of AI as just another part of software. Like, you wouldn't use a database to solve every problem. You shouldn't use AI to solve every problem either. [00:23:30] Adil Saleh: Yeah. Absolutely. It's it's just there to augment it. And in in many ways, as you mentioned, that it's it's not there yet. You know? 00:23:37 Discussion on Customer Success and Future Plans for 2024 [00:23:37] Adil Saleh: You gotta make sure that you seek for enhancements and advancements, before you put in a solution. Like, just for example, like this, you know, the moment this g b d 4 came in, the next week, HubSpot launched their own, you know, their CRM, you know, AI powered, CRM where they can people can talk to see the next day, Salesforce come up in. Why? Because they have, like, this big database. They've been machine learning that database for years years. Now they have trained the data and made it made that, you know, phone ping and everything very efficient to a level where peep they can deliver value. Whereas Yep. The a lot of companies, they're they're still struggling, you know, to get that that solution to that problem. [00:24:15] Evan Zimmerman: Well, I think another reason too is that those are use cases where you can have some fault tolerance. I will say 2 other points there, which is some of these companies have been anticipating. I mean, we have too. So we follow the state of the art very closely, and we scenario plan for what we think the next state of the art jump will be. And so we have features where they are built, and we just need to plug in the AI so that when it's ready to go, sometimes we churn out some of these new features in, like, a week. And our customers call us and say, how did you do that? That's because we knew where the state of the art was going, and so we were prepared. I will say one other thing, which is about it not being there yet. There's a book that I found incredibly helpful called Crossing the Chasm. It's actually where the term early adopter comes from. It's a really fantastic book. And the thing that's kind of interesting about it is that we've talked about fault tolerance a lot. There's also fault tolerance based on the user. So some people are more on the cutting edge. And so you may determine that you want to do a use case where maybe it's not quite there yet, but lets you offer something that most people are not willing to take the risk to offer. And then what you'd have to do if you're doing that in the crossing the chasm model is then you sell it to people who are willing to take risks that some of the bigger enterprise incumbents won't. Now when you do something like that, you're locking yourself that feature at least out of certain markets, but the advantage you get is that you're a first mover. And so you can learn and develop and improve. And then as the state of the art gets better I mean, you have to assume the state of the art getting better. I think that's a pretty obvious assumption with where AI is right now that you're gonna be ahead of other people. You're gonna have spent, like, a year plus refining your product, and then you can just plug and play as the state of the art model comes out. And now you're gonna be able to go to, like, a Salesforce or a Honeywell and say, look at all these customers that we have. We have a product that's way more refined than these other guys that just pivoted into it last week. And, yeah, like, there were some bugs in the first version, but you're getting version 5. Doesn't that make you feel good? [00:26:25] Adil Saleh: Mhmm. [00:26:26] Evan Zimmerman: You know, when Oracle first came out, it called its software OS 2, even though it was the first version to make enterprise customers feel like they weren't getting a first gen [00:26:34] Adil Saleh: product. So Yeah. I mean, that's, you know, that's what you call the innovation. Like, it it comes with all the garbage, and it comes with all the overhead. You know? And you consistently improve with, with, you know, handing handing it over to more users and install base. Okay. One last thing before, you know, we're pretty much on time. Thank you very much. We are, like, 3 minutes past. You know? So now what is that one thing that you guys are doing in the year 2024 to, you know, make your customers widely successful? Is that, you know, related to pertaining to AI or anything else, like any feature, any new initiative, any segment you're trying to explore? Maybe anything that in terms of GTM or maybe raising funds? What is it? [00:27:19] Evan Zimmerman: I mean, we're a start up, and so we are always exploring segments. I think that GTM is definitely a part a place where we are investing a lot more this year to really, I would say repeatability is the word that we use a lot. You know, how do we get repeatability for everyone from the smallest sole practitioners to the biggest enterprises to not only improve our response rate. Our bounce rate is, like, 0, but, you know, we always wanna improve our response rate, but even more to come to us. You know, some of our customers actually into us through our website. 00:27:53 Discussion on Providing Value to Customers in the Legal Tech Space [00:27:53] Evan Zimmerman: And so how can we do a lot more of that? That's something that we think about a lot. And in terms of providing value to our customers, you know, we have a few ideas. And I think that, you know, the thing about our use case is that we are not, you know, coming in as an ignoramus. Right? We know the different aspects of IP. We know the different things that people do. We know some of the things that make it hard. And so just knocking out those problems 1 by 1 is really where I think the value is gonna come from. And I think the other thing that I'm trying to do is to really get people to understand the value of integrated suite. So, obviously, there's introducing the value of the thing that catches their attention, and getting people to realize and look. We're not a cheap product. Getting people to realize that it's worth paying that because you'll make even more money. There's an ROI. But I will say that showing how the pieces work together is also really cool. You know, I think that one of the reasons why people are afraid of legal tech is that lawyers do not have a lot of experience with procurement. And, partially, it's because I think pre generative AI, the value was just not there for a lot of use cases. I'm gonna be honest with you. I mean, obviously, word processing is an improvement over typewriters, and, obviously, cloud document storage and cloud task management is an improvement over a paralegal and an office. But a lot of the other stuff that was oh, an esignature, that's obviously a game changer. But other than those, you know, small numbers of things, a lot of the value has actually just not been there because the use case is just too complex. Like, almost all of the workflow automation and almost all of the document generation has just been I mean, I'm just gonna be honest, like, hot garbage. Like, I've seen a bunch of it. And so are there companies that do a lot of value for the companies themselves? Clerky is a great example. On the legal automation side where lawyers did it, but maybe lawyers shouldn't have been doing it. Yeah. For sure. Like, there's a reason that Carta has a 250,000,000 ARR business. But a lot of the other ones, you know, they aren't exactly there. So I think showing people what it means to have ROI and to really know how to approach that problem, that is a huge part of what we're trying to do is better customer communication to not only close those deals, but to close them, producing even more win wins. So, these are sort of, I think, typical startup problems. It's not like, you know, we are Salesforce seeing what's our next $1,000,000,000 pillar. 00:30:37 Discussion on Legal AI and Creating a New Product Category [00:30:37] Evan Zimmerman: also not in ServiceNow, right, saying how do we integrate AI and everything. And they're saying AI is the biggest, most successful product add on they've ever had, and it's been less than a year. I think for us, it's really how do we create this new product category of legal AI and really rock it and produce so much value that, essentially, we get a slice of the entire industry. And not only that, we have people nearby start to give us a call and say, you know, our you know, the partners at our firm look pretty happy. What would you think about about talking to, to us over at FDA or something like that? That's really where we're thinking. [00:31:21] Adil Saleh: Absolutely. I I hear you. Because, more than a decade ago, like, people were thinking about customer success as a category, you know, quite the same. Like, I mean, people were thinking had double thoughts about, you know, when Gainsight and Salesforce was emerging, having their CS operations, subscription base, all of that. And then, you know, it became a category. Like, it it it was, with with, you know and YC was one of the Ycomer was one of the, you know, platform that actually, have this, expanded to the to a category. So, I mean, we we got to meet a lot of startups like Stoplight. They get created this, this API category too. Now there's so many products coming up. So I think, I mean, I'll get the team a note about about bringing more of these products just like we have today. Evan, it was really nice meeting you. It was pleasure talking and getting to know, your journey and not just edge your your thought process around it and the way you've been so opinionated, about about about the industry and the trends and, the way AI AI is taking this. [00:32:26] Evan Zimmerman: Thank you so much. I do wanna close with one thought, which is based on something you just said. Learn from the history of technology. We have this weird thing that happens in Silicon Valley where you have a company that's a breakthrough success, and suddenly everyone copies what they're doing even if it doesn't make sense. And then when you have a bunch of companies that copy it and it doesn't make sense, we kinda retcon the learning from that company. And for the success that we're seeing, so much of it comes from, you know, me and my cofounder being real nuts for the history of technology and looking at things and saying, here's examples that worked. Here's examples that didn't work, And so here's how we can apply the learning to our own use case. You know, when we finished YC, we started by going to the Computer History Museum in Mountain View. So Mhmm. Yeah. It's by the way, it's fantastic. You should go. You should donate. Totally good. But all I'm saying is you would be amazed how many Palmer Luckey actually talks about this about, like, the memos from the sixties. A shocking number of good ideas are out there for free, and people just forgot about them, either because the time wasn't right or because it was a victim of this Silicon Valley idea hype cycle that I just talked about. You can learn so so much just by looking at things sometimes from the distance past, sometimes from the recent past. I mean, maybe on the next podcast, we'll talk about how we learn from Facebook and LinkedIn in the creation of our product. Mhmm. So [00:33:59] Adil Saleh: Yeah. I was Yeah. I was also listening to a book, Good to Great, and it says, like, you know, good is an enemy of great. So, you know, that author, he actually explains the journey of, you know, Oracle and Apple, Microsoft, all of these from, you know, early eighties, you know, late late nineties and their growth and the patterns and how they correlate, in in in the coming decades and, you know, how these companies are getting a leap and not a lot of them, actually make it to the great companies. And and he drills down into, you know, these patterns. So, and I'm sure listeners would definitely give it a read. So thank you very much, Evan, again, for for this talk. We'll definitely already have a note for the team to bring more look alike, products, you know, so we can see, you know, where this category is going. [00:34:48] Evan Zimmerman: Thank you very much for having me. [00:34:50] Taylor Kenerson: Thank you, Evan. [00:34:51] Adil Saleh: Likewise. Have a good rest of your day. And you? Bye bye. Bye bye.