Building the Ultimate Patent Assistant with Evan Zimmerman

Welcome to the Axial podcast. Axial is an early-stage investment firm based in San Francisco. We partner with great founders and inventors investing in early-stage life science companies, often when they are no more than an idea. Axial is fanatical about helping the rare inventor who's compelled to build their own enduring business. Evan, great to have you on. I think you're one of the most thoughtful people um, just in science. And you're not even a scientist, so you're a lawyer. So I think it'd be really great to talk about your company edge and then also your general outlook on software and science and maybe just, you know, the whole field in general of like scientific research. Um, but maybe to start off things, can you just tell what Edge is and why it's important?

Yeah, first of all, thanks so much for having me. I mean, you're also one of the most thoughtful investors around the life sciences, so I really appreciate your saying that, and it makes a great match for the podcast, right? So in terms of what Edge does, we help people get patents. Patenting is something that has a lot of different aspects to it. We're focused on right now disclosure, so helping you make disclosures quickly and consistently, and then from the disclosure, writing the patent. And again, it's the same mission of doing it with greater speed and greater consistency. The way that we do that is by providing a tool. It's basically a doc editor with an assistant built in. Now, some of what the assistant does is what today you would call AI. It's LLM powered. It will help you edit what you're doing. It'll help you get first drafts of some of the things that are more laborious or maybe a little more technically difficult. But some of what we're doing, and I'm sure we'll talk about this a little more, is actually good design. Patents are very unique documents with lots of different components to them, and there's not really a good tool to just help you write a patent. And so one of the things that we do is we provide in our editor tools where even if you never use our AI component of our assistant, it's just going to be the best damn patent experience you've ever had. From things like managing your references to getting figures done in minutes, not hours.

I have so much to talk about because I read a lot of patents and then I have a lot of portfolio companies paying lawyers a lot of money to do some of this work. And I think it seems like the time is right for new software for patents. But could you talk more about the initial spark? What was the spark that led you to patent management, disclosures, you know, writing patents and starting Edge?

Definitely, and we'll definitely get to the companies because the disclosure piece, that's actually the most underappreciated part of the patent process. So we'll get back to that. But the way that we sort of started this is that patents have actually been a part of my entire career. It's something a lot of people don't know about, about my career, but, you know, sort of go down the road, right? My dad was an inventor. He has almost 30 patents, so that had a big influence on me. Then I went and got involved with patent licensing and a consumer product that was protected by a patent. Then I went to law school, specialized in technology, which at Berkeley, which is the world's top IP law school, technically is really mostly focused on patents. And then I started my own investment firm and our big exit, which was ZKit, which sold to Walmart. It actually had a patent. The board helped it get a software patent, which is really hard to get. So pretty much my entire career, I've seen how patents can actually have a huge material impact on a business outcome if done right. I've seen the other way it's done, in a way that mostly wastes money. So, um, basically I looked at what was going on and I said, you know, patents are this thing that's incredibly powerful. It's kind of this sleeper market. A lot of people don't really think about it, but writing patents alone is an over $30 billion a year market. And so I looked at what was going on, um, you know, in the tools that were available to patent lawyers, not very good. I also saw some of the things that were happening in the technological landscape and I thought, wow, not only is this a good fit, I'm sure we'll talk about legal tech in general. Patents are a particularly good fit for LLMs as for the strengths that they have today and avoiding some of the weaknesses that they have, which do come up actually in other areas of legal tech. So to me it was just something where I thought, you know, you got to try it. Um, we put together a little bit of a demo, we put out a blog post announcing the product, we actually went through Y Combinator, and the response was way stronger than we expected. And we went, wow, like, we aren't even trying to get signups, we're just trying to get people to give us feedback. Um, and today we are, you know, making quite a bit of progress on the company. It's a product that you can reach out and sign up for today. We're starting to turn on charging for some of our users who are extremely happy. Everyone from, you know, a PI at Johns Hopkins to individual— to patent attorneys at some of the biggest firms. So it's been, you know, and even some, you know, some pretty big companies that you may have heard of are trying us out right now. So it's really been a great and rapid rise, kind of what you want to see with a startup.

Yeah, you've definitely, like, in terms of awareness, last, like, few months, killing it. Like, top of Hacker News, like, several times. Uh, really good articles, and it's, like, only the beginning. And I think that kind of shows this is a really— this is a problem that a lot of people are frustrated with, including me. And so, and it resonates across not only, like, biotech, but, like, all of science.— or not just science, just like people. I didn't know, you know, I didn't know about your dad. What did your dad invent? Is it like, uh, did your dad invent anything funny or interesting?

Oh man, so many things. I mean, so my dad, he was a pilot, um, and he has a degree in aerospace engineering. So actually a lot of pilots have engineering degrees. And so, uh, basically he— because he, he learned about patents really when he went to law school, he went to law school himself. Um, it was, you know, it was a thing he was doing while he was a pilot. He was actually trying to be a pioneer in space launch, was just coming around at the time that he went. And, you know, he created a patent for a hide-a-wire, which I remember helping him out with as like a 5-year-old. And then he sold the patent to Monster Cable, and he was like, wait a minute, first of all, I paid for our kitchen, and second of all, he was like, these are a good idea. So basically, his patents focus more on everyday problems that people have. So things like a new curler, a new way of cooking with a pot, and they're all really cool. Like, the kind of trick is that there's one unique physics concept behind each, behind all of the inventions. But, you know, that's, that's really what the focus is. And as I got older, I actually helped him I actually filed for my own patent when I was 13 with my dad's help. So for my science fair, I went to Israel. So this was during the Second Intifada, which was a very scary time. And there were just all of these people who were attacking civilians at malls. And the thing that was kind of crazy is that the way they dealt with this problem was they had security guards. Now, security guards are a good idea, but then you would say, What are the security guards doing? They were literally squinting, like trying to, like, identify people who had clothes that fit weird or were acting fidgety, which is just not really what you would want to do. So I thought, well, what if we had goggles that could identify people? And I saw these heat-seeking goggles. I thought, well, you could do that, but it turns out there was a problem with them. So heat-seeking goggles, there are a lot of things that can block a signal. And so I thought, well, what is something that they have that bombs do not have in common with anything else? And what I, what I found out was that you could actually look at the gradient of the temperature and the density, and that was actually a really good way of finding foreign objects that were unlikely to be, say, a medical device like a pacemaker. And so this was something that actually got a lot of attention. I got a letter from the Israeli Consulate General, and I filed a patent for it. Um, so yeah, patents have been in my life for quite a long time. I had a—

that's awesome that your dad sold a patent. I had a— growing up in Orange County, I had one of the richest families in our neighborhood. Their dad invented— had a patent on the TV stand for the hospitals. You know, you had a little TV. He invented— I guess he patented it, at least. Yeah, yeah. The little swivel or whatever.

I agree.

I think patents have a lot of value for certain industries. I think in tech, patents are kind of ignored or derided because patent trolls are also a thing. And that's another maybe something we could discuss later on. Yeah, there's a lot of research on it, by the way.

What's that? There's a lot of research on patent trolls, so I can talk about that research. We're actually putting out a really big long read next week, our holiday read on, you know, the evidence on the impact of patents on progress and, and their value to society. So I'm happy to talk about that.

Yeah, man, I think later on. I think, I think there's like a— I think patents create a lot of value, but they can be, you know, like any technology, any, anything can be misused. But for the early stages of EDGE, so you identify like a problem to work on, you kind of, kind of have some ingredients of like why now, like there's new technologies, this is a big problem for a lot of people, How do you hone in on the initial market? Doing Y Combinator, you know, maybe some of those, doing things that don't scale, but how did you think about customer discovery, especially which markets? You know, biotech where I exist, patents are extremely valuable, but you know, I can't assume that patents are valuable in every industry. How did you go about like picking specific verticals to focus on?

Yeah, so that's a great question. There are sort of 3 big factors that we think about. So the first one is how different are the patents? So one thing that's kind of a true lie about patents is that there's only one patent. This is the thing that USPTO really goes on about. In practice, this isn't actually really true because you have different categories of patents. And although it is true that you go through the same USPTO, it is true that it's 20 years. And that the regulations are the same with the exception of hat waxmen, there are different norms for each patent. What counts as a figure? What is a good specification? So that was one of the first considerations is, but some areas are more specialized than others. So that was the first consideration. The second consideration that we had was who's filing more patents? And this is somewhat of a gray area because you can't just look at the categorizations. So for example, let's take the auto industry, which is actually one of the top 3 areas for patent filing. GM, of course, they file things that are mechanical, but they also file things that are software, especially because they own Cruise now. So you can't just look at one category and say, OK, this is the category with the most activity. And then the last thing that we thought about is who's actually using the patents. So the software industry actually files a lot more patents than most people think. Than people think, but they're primarily defensive. You mentioned patent trolls. That's actually one of the main ways you deter a patent troll is you have patents that you can use to, you know, for a countersuit or sorry, for other companies with patent trolls, you go to invalidate the patents. So that's a key thing that a lot of patent trolls work through an operating company they have a license with. So you deter the partner with those other patents. But the thing about the software industry is although they have more patents than people think, because they don't use them very much, the value to them is lower. And so whereas if you think about the pharma industry, just as one example, not only do they have patents, they use them a lot, like in-licensing and out-licensing. That's one of the primary things that your in-house counsel is going to do. Fun fact, the pharma industry litigates their patents so much that when you file a federal lawsuit, you have to say what the purpose of the lawsuit is. So one of them is patent litigation. There's a special subsection that says pharmaceutical patent litigation, number 835. And so, you know, there are other examples of places where they use a lot of their patents that they get. So, you know, there are categories where there are fewer patents, but they, but they use them a lot. So for those people, there's actually a lot of value to them in getting their patents. And that's basically how we thought about this verticalization. Now, there's one other thing that a lot of people don't think about, but people in legal tech think about all the time, and that's in-house versus outside counsel. Yeah. So for that, I mean, We basically, so this gets to the disclosure versus patent writing thing, but this is why you have to do the YC thing of talking to your customers. There is way more heterogeneity than I expected. I basically thought that the world was, there's a bunch of things that in-house does, like managing your portfolio, and outside counsel does the writing. And that's pretty much how it is. Turns out that's actually not true and that the things that I thought in-house was doing were actually different than what I thought they were doing. So the way that we think of it today is actually more about vertical. So what industries do you serve? Because patent attorneys do specialize, like I'm a software guy, I'm a mechanical guy. And the other thing we think of is what bundle of features are we selling you?

Because as opposed—

and so whether you're in-house or outside may affect the pricing structure, but But fundamentally, we think about this as what features do you need? Because even we've talked to in-house peers and, you know, same industry, same scale, and they have a completely different organization for how they run their patent strategy. So you have to be able to accommodate those different strategies and be willing to talk to them to understand what their needs are instead of assuming you know what their needs are.

Interesting. So like within that, like, so like each market, each patent class is really really, like, I wanna say, serves different purposes in an organization and it has even a different workflow. You know, it makes sense that software companies patent for defense, whereas like maybe a drug company patents because they actually invented something, they need to protect it and then, you know, monetize it over 20 years. And so, how do you, within that complexity of like talking to all these customers, What, like, I would assume biotech patents have the most value, but have there been any non-intuitive places where it's like, okay, maybe we should focus on these customers better? Is it, I would imagine, like, in biotech, the barriers to entry from the law firms are pretty high. What I'm trying to get to is, like, what considerations have you for Edge like in terms of finding new customers, it can just be based on money, or it's probably not based just on money or the value of the patents. Maybe it's based on, like you're discussing, the underlying structure of licensing technology. Maybe some industries just have easier, lower barriers to entry.

Yeah, it's basically barriers to entry and churn, right? So the way that I sort of think about it is Pick any industry, we'll have you one day. The most important question is at the beginning, you just wanna get people using you. And ideally, you can't really control for this, but ideally you want people who aren't going to churn. And so because it's not just because it makes your numbers look worse in a SaaS business, it's also because you learn something qualitatively. The more people who you have who are working with you on your business over time, And so those long relationships are excellent customer references. You learn the deeper you get into a problem, the more you realize what your more, what some of your bigger features should be and so on. I would say that the industry that most surprised me when I did my research was energy.

Mm-hmm.

Because when I think of energy companies, I think of the most important thing for their business being, you know, the land rights that they basically negotiate. And I think of operations, but actually energy is one of the top 3 areas where people patent things. And I already knew they used a lot of software because I had friends who made software that served the energy industry. But man, I did not realize how much they care about filing for patents. And they use them too. So I think that one of the reasons people didn't know about it is just that a lot of people, when they think of energy companies, they think of ExxonMobil. And yeah, the big guys like Exxon, they file a lot, but there are a lot of companies in the energy space that are subcontractors or providers of capital equipment. And those people need patents to protect the things that they're selling to the ExxonMobils of the world. And the other thing too is, of course, when we think of energy, we think of oil because that's often a euphemism for that. But that also includes a lot of areas like batteries and green energy where there's a lot of innovation going on right now. And that also connects to bio because some of these are synbio. So there's a bit of a gray area doing these categorizations. So like, let's say that I have like a CRISPR-modified bacteria, and you use that to design to do carbon capture. So where do you categorize that in your matrix for when you're deciding who to target in the IP, for an IP product? It's a little bit of a gray area, right? Because I've talked to people who will describe themselves differently who are both doing that category of thing. So I think that's why we focus more on the functionality of what they're doing. When it comes to barriers of entry, I will say sometimes barriers are an advantage. So one of the things that's a big barrier for patents is that you have figures. That's actually a super important part of patents. The exception, of course, is chemistry. Chemistry, actually, often you'll see no figures at all, but you're required almost to provide a figure. And so working with figures is one of the most painful parts of working with a patent. You have to label them. You have to get them made with a patent drafts person who can take weeks to get back to you, especially for complex things like medical device patents. Then you also have to describe them in great detail, and there are rules actually for how you describe them. And so this takes— if you say there's 15 hours being dedicated to the patent, perhaps 5 to even 7 are being dedicated to problems that relate to the figures. And so This is something that you're not gonna get outta ChatGPT, right? So what we basically did was, this is why the design of what we do is so important. We've made a bunch of Figure tools that integrate what you're doing really tightly that make it really easy to work with. And so these things that used to take hours now take minutes. And from there, that also gives us data that makes, that allows us to offer you a more compelling product. We worked with a PI at a big university to help him get his disclosure together. What he said when we worked at the end, we were just getting user feedback from him. He said, this is better than GPT-4. The question, of course, would be like, why? You're seeing the state of the art of the LLMs converging right now with companies like Mistral and product, open source products like Llama with fine tuning. And the reason is that a lot of people focus on the LLM and the LLM tech. And of course there's so much more to do there, but a lot of what this is, is how are you getting your users to give you better inputs? How are you cleverly using the multimodal stuff to make something that your users could not make themselves, or even if they could make it themselves, wouldn't be worth the time investment to make it themselves. So it's, you know, what we're making is not like just an AI product. It's actually like a product that uses AI for some things, but it's a really full-featured, fleshed-out product.

Yeah, I think there's just always these arguments online like, are you a wrapper or a product? And, you know, I don't— I think it's a nonsensical argument. You want users, you just solve problems, and it doesn't matter if you're a wrapper as long as you get growth and then you can then figure out how to retain your users over time. I think Perplexity is definitely— I'm a fan of what they've done. I'm a big time user of Perplexity. I would recommend Edge to take some hits from Perplexity's interface. It's pretty powerful. But on that point of features, what are the main features of Edge? I think the headline features, right, a patent. Yeah, underneath it all are a lot of maybe more simpler functions, like just file a patent, manage it, uh, like things that maybe are not sexy but are pretty important, easier to ignore. Uh, I think just one person from personal experience, a lot of portfolio companies paying lawyers a good amount of money just to file paperwork. It seems like a product like Edge just like solve that and the AI stuff would be extra. That's just extra. If you were to write a pad in, that's extra. Maybe you could describe the product feature set of Edge and where you see the most usage so far, like the most traction. What's the killer feature?

Without a doubt. By the way, I will say one thing about AI, which is You know, the things you mentioned, right? The, the filing, which we're going to be tackling next year, we're going to actually have a product that's going to basically take that work away. It sucks for the lawyers too, because they have paralegals and they don't want to do it either. Some of those things are pure, like just calculating stuff. So like forecasting your fees, telling you when to file, that's totally deterministic. But some of those of the things that we're going to come out with that help with that are actually going to use AI. Filling in an IDS statement, there's parts of it that are deterministic, but there's parts of it where you can use LLMs to make part of it go faster. It just goes to show how for us, as I tell you about our features, we think of AI as like if you're cooking, you have all these different spices and sometimes you use a spice for one dish because it makes it taste better and sometimes it doesn't go with what you're making. That's how we think about it.

It's a tool. AI is a spice, right? Exactly. Okay, cool.

One of the spices. It's a tool in the toolkit, but we don't use it for everything. So what does Edge do? So a few more key features. So I'm gonna break it into 3 pieces. The first one is an assistant. So what the assistant does is it lives within the doc editor and lives in the figure editor, which are features 2 and 3. And it has the full context of what you're doing. And also knows about publicly available patents, litigation, research, and so on. What it— so what it'll do is it'll work with your disclosure. So you have to have the disclosure and give it to the assistant. And then what it'll do is it will do things like draft the first version of the claims, write the detailed description of the figure, write your background section. Maybe you wanted to write boilerplate paragraphs. Maybe you want to rewrite Maybe. So, for example, maybe you got a method section from, from an inventor and you want to turn that in, you know, copy and paste that into the editor and then make it patenty. So like, that's an example. So those are examples of things that you can do with the Edge Assistant, and it's incredibly easy. We have built-in commands and we have freeform where you can just ask them to do whatever you want. The world is your oyster, and it's incredibly easy to use. We think it's the best. One of the things that we aspire to is great— is bestness. And we think that we have the best design of an AI-integrated doc editor anywhere, not just in law, not just in science. The next thing we have, though, is the doc editor. The doc editor is actually a huge part of the value, even if you never use any of our AI features, the doc editor is worth paying for alone. It's specifically made for patents. It makes sure that you're in compliance with USPTO requirements. We automatically create the references for you, which is one of the most painful parts. We next year we're actually going to introduce a tracker so that you can see, you know, if you're out of compliance or forgetting a tracker or a record to track, you can look at your figures while you're writing all that good stuff. And the reason we do this is because patents are kind of unique document because they're very long. But unlike a lot of long documents, these details really matter. A lot of people forget about this if they're not patent attorneys, but you can actually get an office action over really what are basically clerical errors. And those are more likely to proliferate because you have, you know, because these are such big documents that, you know, mistakes creep in. So one example that I love is there's this case involving Apple, the Apple one just a few months ago, where the company that was suing them, they had this huge document and the guy wrote, first we do X, second we do Y, third we do Z. And in the middle of drafting the patent, the guy accidentally deleted second. Oh, and it slipped through not only them, it also slipped through USPTO into the issued patent. And so the case said, sorry, Like, you can't infer that there was a third thing in there. It's not in there. So, um, so you have to really look for those kinds of things. That leads to the third part. This is where some of the most errors come from, is actually the figures. We make it incredibly easy to edit your figures, to label them. We actually have a USPTO-compliant text-to-figure thing, so you don't have to go— or creator— so you don't have to go to a patent drafts person and wait weeks. You can just upload your CAD file to us if you want, you know, the line file that you get from your CADs. We're creating things to make it easy to make flowcharts, you name it, chemical figures. Those things are all in the works. And that gets to something that's also really critical about the details, which is, you know, we focus on the things that are unique to patents. So You're going to get something out that is properly formatted, that is properly designed to output these specific technical things that you need. I mentioned chemicals, that's coming out pretty soon. And so you're going to just get the SMILE code that we automatically render in a chemical formula. How great is that? And obviously, that'll connect to Markoosh groups later. The list kind of goes on for all the different things that we're going to be building. And this also feeds to some of the things that we're going to be doing later. So right now, we have basic management. So, you know, if you are filing one patent, you're probably filing more than one. But of course, we're going to be building down the road a lot more in the way of helping you manage your portfolio. And this gets to the in-house stuff where that's something that every in-house patent attorney is doing. You shouldn't have to pay your attorney anymore after we come out with this feature next year. The other thing is in-house you shouldn't have to do is be writing disclosures. So We actually have an assistant that helps in-house counsel and also outside counsel work with their clients, but in-house counsel work with the inventors so that the inventors can just get their disclosure done super fast, not spend hours writing it, and also minimize the number of meetings they have to have with the in-house counsel who's already probably super overwhelmed. So there's a lot of stuff that's bound together specifically for in-house and for people who are actually high-volume filer companies, even if you don't necessarily have in-house counsel yet. Makes sense.

I think the doc product feature seems like really extensible even beyond patents. And I can imagine like there's various successful products that just help you do citations, help you just edit a scientific paper or just a technical paper. But on that point of like that, that these are the doc feature language in general, what was the logic? —based on values or just based on growth. Like, why not just sell this— sell Edge to law firms? And maybe the feature set would be a little different versus selling software directly to inventors. That seems to be the most radical thing of Edge. Yeah. Technology and all these different features, but just selling the product to different constituencies. But all the successful legal tech SaaS companies like Clio and stuff, you could build it for law firms and they haven't really taken the step to sell directly to inventors. But what was the kind of the outlook for Edge to do that and how do you see that becoming an advantage over time?

Yeah, so the reason why we did that was because when we first started talking, that was a huge number of the people who reached out to us. So to be clear, we do sell to lawyers. I mean, we have, we have law firms, you know, who use Edge today. But what we wanted to do was we wanted to be thoughtful about it. So if you are an inventor, you're not going to be replacing your patent attorney yet. But what we have had is we've had quite a few inventors ask their attorneys to use us. And on top of that, as you mentioned, you have portfolio companies, right? They have patent-related problems that aren't like writing the patent. So that was the other thing is people, some people didn't even realize we were a patent writer. Some people thought that we were actually like a portfolio tracker or something. And so we realized like, wow, there's this ginormous underserved market of people on the inventor side. So not only are we able to serve, and ultimately, by the way, it's for the clients, right? Like that's why it all exists. So not only can we make a tool for the client to improve their satisfaction, It also serves as a way for attorneys to hear about us. And so it actually drives both sides of the business because one of the things that we're starting to see too is attorneys recommending us to their clients. Because ultimately, like, you know, your attorney will do certain things for you if they ask, but most attorneys are overworked. And so there are things they would probably rather not do so they could focus on the things they view as higher value for their clients. So if they can portfolio manage off their plate, they'll do it.

So, and that's a great point then. So like maybe you're not selling directly to law firms upfront, but maybe they still use you and recommend you. There seems to be some tension between building software to handle legal work that could potentially reduce fees for law firms. Okay. And that becomes like a tension that maybe gets bigger over time. Mm-hmm. But I can imagine there's, like you're alluding to, they're overworked and maybe if you can handle some of the busy work for them and help them focus on more high— help lawyers charge fees on higher value work. I think Carta is like an example of this in some ways where they've been able to take some of the low end of the market lawyers used to do. Does it maybe help the law firms in general? But how do you think about alignment between edge lawyers and their clients, because you have to have all law firms that like get pissed at you or think that you're taking away fees.

Um, it's essential, and we're not doing that. I think that the key thing that we— that, that you have to consider is, as you said, they're all overworked. So the question that I sort of say both to the clients and to the outside counsel because again, we do directly sell to outside counsel for the patent writers specifically, is, okay, like, what work do you want to be doing, right? You have 24 hours in a day. You gotta spend some of that time sleeping, some of that time eating, some of that time with your family. There's only so many hours left for work. And do you think that you will be more valuable doing, doing things like their IP strategy and helping them really draft their high-value patents and making them better? Or do you think it will be having a conference once a week with a client over their upcoming fees? That's probably not where you're getting more value. And you've actually seen more and more patent attorneys call themselves IP strategists because that's higher value. And so I think that the tension is totally true, but I think the way that we sort of deal with that is we don't philosophically think that lawyers are going away. We've had people say, hey, will you replace my patent attorney? And we've said no, that's not what we do. But what we can do is make the process better for everyone. That's one of the reasons why patent attorneys are such a great market to be in, because people like their patent attorneys. How many people like their employment attorney? Far fewer, right? And by the way, it has nothing to do with the employment attorneys. Like, every employment attorney I've ever worked with has been great. The problem is the nature of the work. Patents are the only legal work where at the end you get a thing and that thing has value. Almost every type of lawyer is actually about creating or preventing or resolving a conflict, and Oftentimes it's in a way that's illegible to the customer, and so they don't fully understand why they need a lawyer. They just understand that they do need one. Whereas patents, I get a patent at the end and people know what patents are and they see the business value of it. So it's, I think, this market where the tension is less. It's still there, but it's less. And the best way you solve it, of course, providing value to both.

That's a great point. I think that patents, you have to get something, you just value created. And on that point, like, I could imagine Edge helping that process become more efficient and it helps everyone. Do you think there's some untapped market of like unfiled patent, unaccepted patents? I don't have an intuition on that, but like, I don't think that's true in biotech. But do you think there's like some— like, how do I say this? A lot of patents attempted to be filed, rejected, or just left, left, left hanging, and maybe with EDGE could help get over the hump and get approved? Or do you think just like the patent market is generally good at getting patents filed, it's just kind of maybe inefficient right now?

Oh, I mean, there's totally a lot of unfiled patents that should be, and By the way, the biotech industry is not immune from this. And this is actually why we created our disclosure tool. The primary— there are two primary bottlenecks. The first one is lawyer time, which we solve with our patent writer. And again, that can be used by in-house or outside counsel. But the other thing is actually on the inventor side. Lots of inventors do not disclose ideas to be considered for patenting. And the reason they don't do it is because they don't have the time. Because right now, the primary way that you do that is that you actually reach out to either your counsel, who costs money, or to your in-house, who's overworked. And you as the inventor have to spend hours doing this. If you have a scientist whose job is to discover breakthrough medicines, do you really want them spending 10 hours a week filing inventor disclosure forms? You don't. That's not That's not high value for anyone. This problem is especially bad in academia, where a lot of patents come from. Remember, a lot of patents are actually outlicensed from a university, so professors are actually worse at this than scientists who are working at companies. So, um, and that's the reason for that, of course, is that patents have a tension between I need to file a patent but I can't publish the paper yet versus scientists are primarily rewarded by publishing papers. So that's a huge tension that you have. So the disclosure tool helps with that by making it so that the inventor can get a disclosure done in under an hour, sometimes as few as, as little as minutes. And on top of that, they do without having to schedule a meeting and get on someone's calendar, which can take days. And so the result is that you get a lot done in a really short amount of time, and it's also higher quality, so you have fewer iterations. And that not only saves time, it saves a lot of money. So there's a ton of unfiled IP. I think that one of the, the only downside to the tool that we're working on is that we're going to, oh, you know, the USPTO is also kind of overworked, and we're going to help create more patents, and I'm sure that that will be a further problem. And at that time, I'm happy to work with USPTO to develop a product for them to evaluate patents faster.

That makes sense. I think just like— so the idea is like you think more— make high-quality disclosures easier, that will lead to more patents. And, you know, you expand the market in some ways. And law firms in particular would be happy because you get rid of the busy work, but they still retain the high-value work. And you're not destroying the market, actually expanding it. Exactly. One point on that, back to patent trolls. So, okay, make it easier to file a patent and get it done. It's dual use, so potentially patent trolls could— more can emerge because there's like one big group in Seattle and they seem like good people, but they just have a wonky business model. Could you potentially see— what's your view on patent trolls? Yeah.

So first of all, intellectual ventures, there are a lot of conspiracy theories about them. So look, here's the thing. So patent trolls, which are also known as the literature as non-practicing entities, or NPEs, are a real issue actually. But it's worth noting that not all NPEs are bad. The reason why when NPEs become a problem is when they don't provide any value to the ecosystem. So ARM is an example actually of a non-practicing entity. They don't make any chips, but what they do do is they invent a lot of technology. Licensing patents is just their monetization strategy. So those NPEs, no one thinks that they're a problem. Um, the problem is, Guy, is that there's— is that there are these companies, they are colloquially called patent trolls, And people don't like them for two reasons. The first reason they don't like them is because usually— so our tool would not be relevant for patent trolls because patent trolls acquire patents. They don't do any R&D, they don't develop, they don't even write the patents. So like, they literally don't even buy the rights to file patents. They only get patents that exist. Um, so, um, the other re— so that's the, the first reason people don't like them. Uh, the second reason that they— that people don't like them is because there are many patent trolls are perceived as filing frivolous lawsuits, especially in the early 2010s. A lot of patent trolls would shake down companies that didn't seem like they had a lot of money to get them to settle for like $200K because that's better than a million-dollar lawsuit. So that's kind of why people don't like them. I don't think this will really make the patent troll process worse. If anything, I think it'll probably make it a little better because a lot of patent trolls work in conjunction with a licensee to give them standing. And if you have a defensive patent or a thicket of defensive patents, then it's— you're actually in a stronger position to— if they have a licensee, you know, have a defensive countersuit against the licensee, which will make them think, oh, twice. The other thing is we're coming out with more tools over time that will help you on the defense side, like FTO. And so when you have that, you'll just go to PTAB and you'll invalidate the patent. Basically, you can make life harder for them too. So I think that the likely impact will be neutral to a little worse for the patent troll. Like, the final thing that you need to do to affect patent trolls It's actually not really us, it's not really USPTO. It's really that there needs to be like a cost to sitting on a patent and not doing anything with it. Um, and that's really a policy decision from Congress. Um, so, and I think, so I think that if I were to try and address the patent troll issue, I would actually say talk to your local lobbyist, um, and, and tell them that you think there should be a cost for having patents that don't do anything if you're not also doing R&D.

Yeah, I totally agree. Then maybe to shift gears in terms of like going from product features to like the business model, the one fascinating part of AI I'm observing and I'm trying to wrangle my head around is like pricing of software, you know, going away maybe from SaaS to just selling work. But for Edge, How do you think about pricing? And maybe it's still being figured out. Do you think, do you think like people want to pay a subscription to the software, or do people want to pay for on a per-patent basis, or like just work that's accomplished? That's a big topic in AI software, just like you potentially just— people just want to buy work, they don't want to buy software. And that seems— I don't think anybody has a good answer. People are so— we're as a collective, we're all trying to figure it out. That seems to be the most interesting business model part of AI and software is just pricing. So maybe for Edge, how are you thinking about it right now and how do you see it evolving over time?

Oh man, I have so many thoughts on this. So the, the person who I think wrote the best post on this so far is Sarah Tavel at Benchmark. And I would say, first of all, read that post. Pause the podcast. All right. Now come back. So there are a few things I would like to— that I think are important to expand on for what Sarah wrote. So she was basically talking about the idea that SaaS has evolved to be an enabler, and now you can potentially just do the work. And so you might as well pay for that. Now, there's some things that are beneficial to that. Part of it is that the work is actually higher dollar. So, so you can— and the thing is that it's not like the cost of software has gone up all that much. So you would actually make a lot more money. Now, what she doesn't talk about as much is that if you're doing a work-based approach, then you have more customer acquisition costs potentially. So that could be taken away. And this is actually what happened to direct-to-consumer is that you got more revenue by going direct., but then everyone, but then you had more competitors and so that made your margins were actually worse than when you were doing retail. But that's one thing to consider. Another thing to consider, of course, is what if it's a non-repetitive use? So SaaS was this thing that kind of fit a little strangely into certain business models. So Lattice is a really cool example of this. They do HR software and people, they have a SaaS business model, but the problem with with that for what they do is that you only use their product because it's like 360 reviews, OKRs, you only use it a few times a year. So like, so like, does it make sense for that product? I mean, they're a billion dollar company, so something's working and they're profitable, but it's a thing to think about. And so I think that if you're going to go that approach, what you need to do is you need to Be able to do the whole work. No one is going to pay you for the job if they still have to use a vendor. You have to think about, is it something people do a lot of, in which case they will want to do bulk pricing, which is again just SaaS. And also, are there things you can charge for different people at different times? So for a patent lawyer product, they do patents day in, day out. For the, for the company side, they all, all of them do it day in, day out. So like a Merck is filing several patents a day.. But like a small biotech, right? They're probably not. So though you might have different products and different business models. So like another thing that you see is actually kind of a blend where you have a platform fee. So Notarize is an example of a company that does this. Another example is DocuSign. So for a lot of legal tech companies have actually done this where you have a platform fee and then you have, but it's not the highest. And then you have a per-use fee. And so the idea is, and it can be tiered, they can buy it in bulk. I think Snowflake actually does this too. So that's another sort of option. And so the way that we have sort of thought about this is right now we're doing SaaS, but you can see some of that changing. So for example, when we eventually do trademarks, like that might make sense to do per-use. Like instead of paying $500 to an attorney, just spend, you know, $50 and just get it done at the click of a button. But would that make sense to do unlimited trademarks for, say, $100,000 a year? I'm not so sure. I think maybe just the per use might make more sense, but you might still be paying a subscription in order to access our portfolio management tools because that paradigm still makes sense for that part of the product. So I, what I'm trying to get at here is, is again, for us right now we're SaaS, but with a different pricing scheme based on whether you're an in-house counsel or whether you're outside counsel, specifically based on the feature that you're using. But I think that what you're gonna see is more heterogeneity. And this is a thing that you see in a lot of other industries. Like if you look at trucking, like those are really complex pricing schemes.

So you could imagine that in software too. Cool. I think there's a lot more to unpack there. In terms of just AI in general, what else excites you? I think it seems like LLMs in particular could transform how to form hypotheses in scientific research and even beyond science, probably a lot more. I don't think as much about, but what else excites you about AI? Where else do you see it having an impact in science, biotech, and maybe just beyond?

So it's places where you need some type of creativity. So these are very heterogeneous systems that sometimes have errors, much like us. And so the thing that they have is that they're not creative like us. They're not able to think out of distribution, but they know more about the distribution than you do. And so I think that's one area that's really great. The next thing that I'm excited about is I call it translation. So like some legal tech has this characteristic, some doesn't, but how often are you writing something out of whole cloth versus taking something and filling in the blanks and making it look more like something that's in another place where there's a lot of busy work? I think that's a place where I'm really, really excited about it. So you mentioned hypothesis testing. I think that kind of stuff is incredible. And there is this paper that just came out about another science problem that got solved with an LLM. There are a lot of people who are going to be able to solve some real issues. The thing I'm most excited about for biology, I've always called it the combinatorial science. Because biology is the study of the specific. Like, there are general principles, but more often the thing that's more interesting is, oh, there was this protein we didn't even know it exists, and it turns out that it's actually, you know, the protein that modulates the response to chemotherapy or something. You discover things that are new. And when you have these combinatorial, computationally heavy things, the ability to churn through a lot of data and look for patterns that either prove something or say you should look more or do an experiment. That's the field that I think is going to probably benefit the most from this. If you think about something like physics, yeah, you can do hypothesis generation, but for the most part, it's pretty deterministic. The things where it's more complex, we have numerical approximation methods for these things that don't have closed-form solutions. That's more what you would use some of this stuff for. There's a guy named Terence Tao. He's the most famous living mathematician. He's a genius. He's been using actually LLMs specifically for, I guess you wouldn't call it a hypothesis in math, but we'll say that anyways. He'd been using it for hypothesis testing. He's been using it to get ideas, especially in areas where he doesn't know a lot. Because yes, even Terry Tao has things he doesn't know a lot about. And— oh yeah, no, I was just going to say that, like, if even Terry Tao can say, here, this will help me learn about an area that's new to me, and yeah, he still has to do the work of getting up to speed and do his thing of proving the theorems that are relevant to him, but it can help him get there faster, that's huge. And I will just say one last thing, which is there's a bunch of research that's been coming out about who benefits from LLMs. And what it's finding consistently is that although everyone benefits, the people at the bottom of the distribution benefit a lot more. So these are things— so who's at the bottom? It's people who are maybe not as good, but even more importantly, junior people. And so I think seeing this as a training tool, a tool that makes people a little more, you know, get a little bit of a head start to provide more value early on in their careers, That's really exciting to me.

I agree. I think I wanted to add just that Terry Cowan had an article about this recently I read, and that was pretty fascinating. I think from my experience as a biologist, I would say these models create the wrong answer most of the time, but they create interesting answers, like interesting hypotheses that you don't consider, and then you have to maybe fix them, or maybe there's one little detail that's off. But if you just, you know, but the rest is really interesting. Like, say you upload a bunch of genes and pathways and interactions, it'll give you a suggested mechanism of action that you wouldn't even have thought about because it's non-intuitive. Or mainly as a scientist, even for me, I'm so calcified in what I've been taught and my experiences, and, you know, and it's hard to get out of your box. And so I think LLMs, at least for me, have been helpful to just consider new ideas that I would think don't make sense, but actually do if you really think about it. And just some editing involved. But yeah, I think there's, like you said, there's like an educational component.

The calcification makes me curious because there was a paper that came out last week where they looked at GPT and they said, oh, it did a lot worse than some of these doctors. And like when you talk to people who are going to med school, they'll say that a lot of the people who are, you know, older, you know, presiding physicians are doing what they learned when they were in med school, but they haven't adjusted to some of the new teachings. And it does make me wonder, like, how many, you know, we have some of these evaluation methods by experts, or, you know, how many times are we actually just calcified in our thinking and we're seeing something that's new I'm wondering, how are you using this as an investor? Obviously, you mentioned FTOs, but what are some of the other things that you're doing?

I'm a big fan of Perplexity. They have the best interface of any AI tool right now. And the CEO, I've never met him, but he's incredible in terms of his thinking about— he's been able to accept that he's a rapper for an AI. I think a lot of companies are trying not to be a rapper. He's like, I'm a rapper. But I'm a really successful one. But Perplexity has been kind of my go-to spot to just find new leads of ideas, diligence the market, say, hey, what are these companies doing? And it'll give me suggestions of where to follow on, and I can just find new threads of market ideas or technologies. And then I'll use a Claude mainly to repattern. GPT is too constrained. Too many people use it right now. I was paying for it, but it was just the performance was not as good as they needed it to be. And then there's also like, how do I say this? Like uploading like data, public data, not company data, public data to verify something. Is this how this pathway works or this drug works. It's more often wrong than right, probably always wrong. But if I just read through the results, I can kind of figure out, oh, that doesn't make sense, that gene doesn't do that, or even just the results don't make sense, like the genes are flipped, or the molecule, the drug is said to do something that it's known not to do. But once you correct for that, it gives you validation or new leads to follow through on. For example, you just want to know, does this drug have any side effects? You should think about toxicity. When you invest in a biotech company, you're always worried about toxicity. I want to make sure this thing doesn't have any risks that it's going to kill a human. Other parts would be like No, sometimes we need diligence and technology, like a tool or instrument. There's often like a company already selling a similar service or product that you've never heard about. And so getting out of the whole startup—

sometimes I get—

you get out of the startup echo chamber, there's probably a company out in the world doing the same service you're doing in Ohio or London that got started 20 years ago. That's been helpful to like uncover case studies. Where it doesn't help though, it doesn't give you answers. That's the thing. I don't get answers from it. It's not like it gives me logical answers. It gives me leads. But that's like— this is like, yeah, this is one of the few technologies that have been like part of my daily life. I think it's true for everyone. So yeah, I think On that point for Edge, how do you see— oh, it's 10 years from now, forward-looking. You talk about assistants, you talk about docs and other features. How do you see Edge being part of the daily life of an adventurer? You want it something that becomes their go-to spot for all scientific work? What's the TAM of Edge, or what do you see the role of Edge being in an adventurer's life?

We help inventors spend all of their time inventing. And so the way that I sort of look at this is that I want to make it so that all the things that get in the way of you doing your actual work, they're just gone. So, you know, you need— and this is especially in a business context, right? So you need to file a patent. Well, it's super easy. You maybe need trademarks. Done. You know, you've got to manage it. Not a problem. Just look into your dashboard. You know, you have a competitor and you're curious about what they've filed. Go and take a look. The list kind of goes on where the, you know, you could imagine us doing things like helping you identify licensors. Potential licensees, potentially, of your patents. You can imagine using us to actually have a marketplace to, you know, go and sell your wares. You can imagine us maybe even helping you with things like your compliance, like maybe some of your clinicaltrials.gov stuff goes through Edge. I think, you know, you can also imagine doing other legal work. Obviously, you have employment law issues, you have securities regulations because you're about to go public. A lot of those, maybe some of those things will go through Edge too. You can imagine a lot of really cool stuff happening that basically makes it so that you can focus on the thing that you do super well. Because think about what you've just described. You described your own work and it helped you get an answer you wouldn't have otherwise gotten and helped you get there faster. Those are the two main things that it did. And, uh, it also, by the way, implicitly is the third thing, acted like an associate that you don't have. Yes, basically increased your labor pool beyond what actually exists on planet Earth. There are actually not enough people that for you to have an associate. Um, that is what we're going to be doing for whether you're— for everyone, whether you are a lawyer, whether you're a scientist, whether you're an engineer, whether you're an in-house counsel, you name it. If you are inventing something, you're not going to have things that stand in your way from inventing anymore or monetizing your invention when Edge is complete with its mission.

Well, okay, well, Evan, we'll, we'll do a conversation again pretty soon. I'm really excited for Edge as you launch more features and We'll link the website and people should sign up and get in contact with the company and start using the products. But thanks for doing this, and I think you're onto something big. This is a huge market, and it's like one of the most obvious— it's not obvious, but one of the best, the most suited applications for LLMs because patents are so structured. There's a lot of barriers, and you're solving those barriers. And so I think we're all excited that, you know, as you grow and you announce more features next year, we'll, I'll definitely be tracking that.

I'm super excited and obviously, you know, we'll continue to, you know, we'll have another conversation soon, as you said.

I can't wait. Absolutely.

Thanks so much, Evan. Thank you.