Tony Thai, Founder and CEO of HyperDraft, Revolutionizing Legal Tech with Engineering Precision

0:00 Jon McLachlan

Okay, welcome, everyone, to the Security Podcast of the Silicon Valley. I'm here today with a very special guest, Tony Thai, the founder and CEO of HyperDraft. Welcome to the show, Tony.

0:13 Tony Thai

Hey, Jon, thanks for having me, man.

0:14 Jon McLachlan

So, looking at your background, you've worked in many different jobs in many different industries. How did you end up where you are today as the founder of HyperDraft?

0:22 Tony Thai

Yeah. You and I met when I was in that last phase as an attorney at a big law firm, but it all started actually as a software engineer developing enterprise B2B software in my early twenties. And during that time, I was working with a lot of lawyers and, frankly, a little bit frustrated by the feeling that they didn't understand what the business objectives were. Like any self-respecting engineer, I thought I could do it better and just went to law school as a result. I think when I talk to engineers, like such as yourself and like other folks, like when I say, oh, wasn't happy with the legal service that I got, so I thought I could do it better, like a lot of engineers, like, that's the most engineering, like, engineer thing you could have said. There's just not a lot of people that are wired to go and do something themselves. And yeah, I mean, I went to law school, didn't think I'd practice in private practice, like, representing clients like yourself.

1:16 Jon McLachlan

But thank you very much for the representation.

1:18 Tony Thai

Yeah, I was most [appreciative]. No, people like you [were] one of the many reasons why I stuck around, because I was like, oh, well, if I'm facing this problem, then there's going to be a handful of folks like Jon who needed my help. Might as well help the community out instead of you having to go to law school.

1:35 Jon McLachlan

So, for all of our listeners, just for the context, back in the day when I was a founder at a company called Peacemakr, Tony worked at, where was it? It was Goodwin. And you got paired up with us, and you were our primary contact for all of our legal needs. And it was just a spectacular back and forth. And we had to fix a whole bunch of things, but we messed up with our founding of the company, and then we did some patents, and Tony was always on the ball. And then, you know, man, like, all in all, like,

2:04 Tony Thai

You guys were pretty okay case, if we're getting surgical with it, or medical with it. Like, come in with a lot of different cases. I think your guys's was pretty routinized. I want to say you came through [John Runyon] and [McCusker], but I can't remember. I don't remember what partner it was. Yeah. Do you know [John Runyon]?

2:23 Jon McLachlan

No. Okay. One of the co-founders [at] Peacemakr, he had worked with someone there before, had lots of good experiences, and that's how we ended up with...

2:34 Tony Thai

Gotcha. Very cool.

2:36 Jon McLachlan

But I remember one day you showed up and you're like, oh, I have to share with you some news. I'm moving on. I'm going to do something new. And you made me guess, made me guess what you were doing next. And I just remember thinking to myself, wow, he's so excited. Like, there's a little twinkle in his eye and everything, and like just all of this positive energy. I'm like, you're way too excited to be doing anything except starting your own company. Why didn't you warn me, man? You were already on the other side. You should have been like, what are you doing? Well, I thought, I thought you were getting into it. It's the best thing and the worst thing, both at the same time. There's never any, like, right time to do that. And you just got to pull the trigger. You just jump in. And now look at you. This is amazing.

3:22 Tony Thai

Yeah. I saw how much fun you guys were having, and I'm like, I [wanted in on] that action. Yeah, we started HyperDraft actually a few years prior. I started with a moonlight bootstrap project. It was, and I don't know if you remember my schedule, but like, working at a big law firm was not very conducive to those types of arrangements, where you're grinding from 7 a.m. to 11 p.m. and coding from 11 p.m. to 2 a.m. It was a passion project, to say the least. But yeah, we started a few years earlier. One of the things that you learn, a lot of [parallels] in security with this, but like, one of the things you learn or are trained to do at the law firm is to be risk averse, right? They constantly tell you, because, like, our jobs are to help clients navigate risk and reduce risk as much as possible. If I'm being honest with you, one of the reasons why I didn't leave was, like, I was so trained at that time to avoid risk. And there's really nothing riskier, especially in a professional, like, services vertical, than to start your own business. And it took a handful of my clients to convince me over the course of 12, 13 months that I should be doing this full time, not half time.

4:29 Jon McLachlan

[It's amazing] to have that type of support in your life and surround yourself with people that believed in you and saw the potential in you and in the company and in solving the problem that you're going after. And that's awesome, too, actually, that you came up through the ranks through a software engineering background and then got over into law a little bit.

4:50 Tony Thai

Yeah, a lot of people look at me funny, because they're like, there's just no way you could be decent at both. And it's funny because, for me, it's the same thing. It's just, like, slightly different context, right? Depends on which hat I have to wear. I'm still a lawyer. I still practice on occasion, and I had to put on a different hat, but you still try to use at least the same logical patterns when you think through stuff. So honestly, the coding background was really important for my development as an attorney, because having that ability to, or the need to craft logic through code, translated perfectly over to doing contracts and deal work and anything requiring legal documentation. So that was helpful. And it, I would say, it seeded concept around the business, right? I'm sure that folks listening to the podcast that still don't know what we do. So at HyperDraft, we build dev tooling for lawyers and business professionals. So what does that mean? Like, we focus on document generation and digital workflows, because we found that the legal field was lacking tooling in that space. And as an engineer, we built so many tools for ourselves. For me, the [natural] progression was, as a lawyer, like, I need to build [tooling] for myself and the other folks in this vertical.

6:10 Jon McLachlan

I remember you demoed it to me way back in the day, when it was an early version of your product. And I remember thinking to myself, wow, now lawyers have an IDE, an integrated development environment, and they could just sit down, and they could share snippets and auto-completion. And it was really, it looked really nice.

6:29 Tony Thai

Yeah.

6:29 Jon McLachlan

A big time saver.

6:31 Tony Thai

It was. When I built this software, when we started designing all that stuff, we did ask a lot of lawyers, like, their input. But the reality of it was, these are tools that most people in the outside world assume that we have, which is funny, right? Because you hear this common complaint that lawyers bill by the hour, and they take too long on stuff, and that they're being fraudulent with their... No, the reality is we're just that slow, right? If a contract took a long time, it actually took a long time. It probably took longer than they billed you, frankly. So I wanted to reduce that, because if we look at it, like, these lawyers have such amazing raw compute, right? And not to objectify them, but they have such amazing raw compute to spend their time on so many, like, small, tedious things. Not that it's not important, it's important, but to do it in a more streamlined manner, like, that's going to be key to apply that raw compute to stuff that matters, rather than just the mundane.

7:24 Jon McLachlan

Yeah, yeah, absolutely. And so, as you've built out this product, I'm sure you've gotten a much wider view of the legal community in general. And I'm really curious to hear your insight of what sort of relationship do you see that cybersecurity has with legal?

7:42 Tony Thai

Yeah, just before we started, like I was talking, I mentioned one of my good friends and former colleagues and an advisor to HyperDraft is a guy by the name of Steve Singer. He's currently heading up LegalZoom's InfoSec. We met when I was in-house at this private equity portfolio company, and we hit it off, like, immediately. And there's a ton of overlap, and I'll tell you the story with a few stories that he would tell me. As a corporate attorney in-house, one of the things that we're taught is, and we like to remind people, is that if we're doing our jobs, you're not going to hear about us, right? A corporate attorney should be not non-impactful, very impactful, but in the sense that if they're doing their job, then risk is being avoided. So you're not hearing about them and then the work that they're doing constantly. Same goes for InfoSec, right? The same goes for cybersecurity. If the job's being done correctly, then you're not really going to hear about it. And that causes a lot of issues, because that's a surefire way for business units and certain stakeholders to take for granted the need to have those teams, the need to have the tooling, the need to think about the objectives for those teams, because they're controlling for risk. So you don't really need them until you absolutely need them, right? Whether you've got a security breach or you've breached some sort of agreement, right? You don't need them until you realize, oh crap, I'm far out of my depth in terms of the potential downside for the risk that this may create. So in that sense, just from a perspective of attack surface, like, we share a lot of the same kind of risk management and neglect from businesses, frankly.

9:21 Jon McLachlan

Yeah, I get that. And it's almost like one of those thankless roles. If you're doing your job really well, and you're staying out of the news for all the wrong reasons, and no breaches in companies that you're servicing, or legal conundrums, then the legal space, very quiet. You're not a high visibility.

9:42 Tony Thai

Yeah. Like we said, until it goes wrong, right?

9:44 Jon McLachlan

Until it goes wrong. And then when it goes wrong, like, it goes south, it goes very south. So when it goes south, what sort of cars are you seeing drive down that road as things go south? Anything specific in the, right there, in HyperDraft space that you've noticed?

10:00 Tony Thai

I actually, a lot of woulda, shoulda, couldas, you know. And I'm sure, like, you're rolling your eyes, and plenty of people listening are rolling their eyes, because it's, in many respects, one of our jobs to be the paranoid people to say that, hey guys, like, we should really worry about this. But one of the things you pick up over time and with experience is that to avoid the cry wolf syndrome and to pick and choose when to say, hey, guys, now's the time, right? Like, that's a particular issue. For us in the HyperDraft world, I've learned a lot through the process as well. You mentioned us as an IDE, and we really do truly think of ourselves as the IDE and digital workflow systems for these large enterprises. But that's from the provider perspective, right? From the demand perspective, it's actually something that I've learned over the last few years running this company. It's like, sometimes clients come to you and they don't know what they want, and they don't know what they need. So they think that they can purchase a solution that is going to be the silver bullet for solving their either compliance issue or security issue. But in that respect, it's really just a point solution, in the sense that it's a point solution in time for a particular demand, like, pain point at that particular point. So with the HyperDraft tooling on the legal side, like, we'll have folks come to us and say, hey, this is what we think our problem is. And then, as our team digs in deeper, we realize, actually, that's not the problem, right? Your problem is you have no process. So lack of software, lack of productivity and streamlining, this is not necessarily the problem. The problem is you lack the discipline to set in place all these plans and procedures for how to send a contract through, how to mitigate risk, how to evaluate risk through that pipeline. And we can help with that. But software is not going to do that. But there are plenty of platforms that do claim to be the McAfee of contract generation and workflows. And what I've learned over time is just to tell them, like, give it a go. If somebody is going to argue with you over, like, using an out-of-the-box solution versus actually spending the time to evaluate the problem set and figure out different operating procedures, [they already have] their mindset. And so you wait for them to make that first mistake, and then hopefully it's not fatal.

12:23 Jon McLachlan

Right. Hopefully it's just a lesson learned, and you can move on with it.

12:26 Tony Thai

Exactly.

12:28 Jon McLachlan

So, in terms of folks that are using HyperDraft, what industries do you see are most common? [Is there] different types of law [that] attract a certain type of use case, maybe?

12:39 Tony Thai

Yeah, we serve, like, we're pretty industry agnostic for the most part. However, we have found many companies in finance, real estate, and healthcare have come to us because they have particular needs. And actually, this is where the great crossover for our industries, which is, they need something hyper-customized that's very complex, but also needs to be secure and in a controlled environment. And for us, that's our specialty, right? Just coming from an enterprise development background, like, I started this with enterprise in mind. And so those industries have those particular security and kind of complex deal needs. And so we found ourselves servicing a lot of those folks. And frankly, we turn away probably more work than we take, just because some clients come to us and, like I mentioned, they don't really know their problem set, and there's no use in the overkill of the HyperDraft platform for some folks. Many folks can just get away with mail merge, right? Like, just spend a few hours learning how to do that, and we don't have to get involved. And yeah, we get hit up by everyone. But yeah, I'd say that just three [focus verticals]: finance, real estate, and healthcare.

13:52 Jon McLachlan

Nice, nice. Your go-to-market strategy is, it sounds very bottom up, with people trying it out, maybe coming to you? Or is it a, what's your go-to-market strategy?

14:01 Tony Thai

Go-to-market strategy is actually more top down than anything, because, yeah, it's more top down. It's at that enterprise level. So you need purchasing power. Not only that, but you need purchasing decision-making power. So usually it's a department that says, hey, we have X number of agreements or deal types. We're too slow. Everything's being done manually. Great use case. And this is probably further on the, like, the edge case. We had a client that came in. They did term sheets and these long-form agreements. Term sheets would take them eight hours, and then the long forms would take them anywhere from, like, 20 to 30 hours, everything done manually. And for them, time is money, right? For every one of those contracts that's not going through, they're not making money. So the head of that department came to us and said, I'll pay you guys to make this faster. It doesn't even need to be perfect. It just needs to be faster than what we currently do. And so we cut down the time from eight hours down to 15 minutes for that term sheet, which is insane, like an insane improvement, right?

15:00 Jon McLachlan

Just from a [unclear] payback for every single [term sheet].

15:06 Tony Thai

Yeah, well, get a kick out of this, man. Like, from the 20 to 30 hours it took for the long-form agreements, like, we cut that down to three hours. And that three hours was mostly just manual review of the stuff that needed to be done, could be done in an automated fashion. And this is where the lawyers that bring all of those superpowers to the table can focus on their superpowers, instead of, like, [futz] around or putz around with their mundane bits that are important, but that software can help with, and that you can reuse, maybe. Jon, I'm actually curious about your experience here, because let me describe to you the lawyers' scenario, and I think the InfoSec, cybersecurity, is similar. I'm putting on my lawyer hat now. When you're a lawyer, your clients don't really appreciate the amount, like, technical work that you do. So, like, when you and I were talking, I could be wrong, but you probably appreciated me more, because, like, we would jump on a phone call and I'd talk you through it. And I'm enough of a business person that I know, hey, this is what Jon wants to hear. And frankly, like, if I'm Jon, I don't want to hear this and this, like, I'll take care of it. Just controlling for that messaging. But I'd go back and stay up all night getting your deal docs done. You didn't know that. You weren't sitting there with me. And what we're trying to get lawyers to do more of is, it's not that you're just losing all that time, that 20 to 30 hours that you're losing worth of quote-unquote billable work. [We want you to] spend that time now working with the client, understanding their needs, and talking through with them, like, how do we see around corners, right? Okay, so we've done these deal types. Like, how can we make this more streamlined? How can we anticipate risk? Where have we failed before? Those are statements that not a lot of lawyers are willing to have with their clients, right? Like, I'm very happy, like, not very happy, but willing to sit down with the client and be like, all right, Jon, let's do the postmortem on this last deal that we did. What did you think that we did well? And then what did you think that we did that you'd like us to change? And I'm not [telling] that we might change it, but I want to hear it. And that's not something that, like, a lot of lawyers want to do, but the reality is the value proposition and their skill set, the clients greatly benefit from those types of interaction. Yeah, I'm curious if that's the same on the cybersecurity side.

17:16 Jon McLachlan

No, it's definitely on the cybersecurity side. And it resonates a lot with the industry, in terms of the evolution I've seen and just how to write software, how to build software. At first, in the beginning, it was very just, like, waterfall design model. You get into the grind, you just turn out code, and you, it's very disconnected. But now, like, with agile development, everyone is talking to each other. We're doing postmortems. There's incidents that happen in the cloud or in your data center. You do, you want to go back and you want to revisit those things. And there's much less time, I would say, banging on the keyboard and much more time spent engaging with each other. And so there was a natural evolution, I would say, in engineering, that sounds like it's nudged in that same direction in the legal space. And I think that will result in better uses of folks' time, better products. It'll help folks see the bigger picture. And I'm a firm believer that when people [see] the bigger picture, that you can trust and you can delegate, and people can help you solve really hard problems, instead of having to get in there, having to micromanage everything, having to, quote unquote, like, not trust folks around you. You do want to trust folks. You want to surround yourself with incredible talent, but you want to help them understand the big picture, so that they can contribute to that mission without you having to be part of all of those little nitty-gritty. That's why I really enjoyed us working together, because I felt like you really got that. You saw that. And you didn't need the tools to help with that. But one of the issues is, if, like, writing documents, writing code, in an engineering perspective, takes so long, there's not a very good motive to move into that space where you talk to each other more. It's communication. Yeah. And speaking of the inner, the overlaps between the two worlds, maybe share with us a little bit about what makes HyperDraft approach to cybersecurity and the resulting legal issues unique.

19:16 Tony Thai

Yeah, worked as a lawyer for so many different clients and seen the security breaches, or, sorry, security incidences.

19:23 Jon McLachlan

Security, security incidents. Yeah, [alleged] things. Security incidents, yeah. [Possibly]. Yeah, yeah.

19:32 Tony Thai

Allegedly, some actor took our [information]. I understand the need for InfoSec's role in so many different deal points. And for us, when we're talking about proprietary information, if you're working with a business, what's more proprietary than, like, how they make money, right? Like, the contracts that are down to how they make money, and the strategy behind it, and what they need to do, and the negotiation. So there needs to be a lot of trust in systems that they're not sharing that data, leveraging that data, leaking that data. So I would say that, first and foremost, that's a main concern for a lot of our clients. And one of the reasons why we've been successful is because I had people like yourself to look up to, and Steve Singer, who I would [lean on for] a lot of these things, because their perspective is, InfoSec, like legal, like HyperDraft, is a first line of defense, right? It's only a first line of defense. Having a good, strong first line defense, that's a good thing. But the rest of it is around communication, training, and trusting your colleagues and business units to do the right thing. I would say a lot of our successes, in terms of winning over clients and having clients pick us over competitors, is our approach to solving the problem. It sounds so cheesy for me to say, that's really, truly in the team and the architecture. There are other solutions out there, plenty of other solutions out there, especially in the doc gen space, that claim to solve, especially in the contract management [space], which, I'll be honest with you, nobody's been able to explain to me, like, what makes their contract management special. That's like you telling me, hey, Tony, like, I have 50 different replacements for GitHub repos. And you're like, that's cool. I don't see the need for it, but sure. And so, like, we take it much more like GitHub and GitLab type approach to it, which is, let's look at the use case. Where are there problems? How do we solve it for them, so that they have an immediate value proposition to them? At the same time, look after them and figure out how they can control for security risks and legal risks. Yeah, I don't know if that's a good answer or response to your question.

21:41 Jon McLachlan

No, that's, no, that's perfect. You bring to heart what matters to your customers. And what matters to your customers, it sounds like a big chunk of it is cyber, cybersecurity, and the approach to actually solving problems and delivering value right away. Tony, you've noticed, no doubt, that AI is exploding right now. Bubbles, or new technologies, and things that, probably, people have thought they'd never see in their lifetimes are now possible. And it's accelerating at a rapid pace. What type of AI features do you guys offer to your customers?

22:15 Tony Thai

Yeah. We have seen the kind of boom of AI features, as we're one of the OGs, right? Like, we did this before the hype. You can testify to that part.

22:25 Jon McLachlan

Absolutely.

22:25 Tony Thai

Our AI features are similar to what's useful for the engineering space. Think of it like completion, Copilot.

22:34 Jon McLachlan

Auto-completion, yeah. Copilot, [text generation].

22:37 Tony Thai

I was actually hoping you'd transition to this, because the same security risk that is evident in AI-driven code is also a big issue in legal. So we are hesitant to push AI-heavy products. Something my CMO and I talk about all the time is, it's often touted as a product. It's just a feature, right? [They assist in] the ability to work in a more streamlined fashion, is a feature. It is not a standalone product. And again, I would say that our customers come to us because we understand the data, and we understand the business practices and the legal models for how it works, not because we claim to be able to replace their headcount with AI.

23:27 Jon McLachlan

I like that. That's, so that's a lead into my next question, is, what's the impact the AI advancements are having on the legal industry as you see it? Does not replace the headcount?

23:38 Tony Thai

I think it's still pretty early to tell. Let me tell you it from my perspective on the coding side, right? Like, I still code. I use Copilot. It's incredibly useful much of the time. It's 60, 70% of the time. But as I use it, and I see our junior engineers use it, I review their code, and I'm constantly reminded of the difference between a senior engineer and a junior engineer. The thing that it can't [replace] is the time and experience that it takes to learn what good architecture, what good patterns look like, thinking ahead. And, fortunately or unfortunately for the legal profession, like, time still needs to be spent learning different mental models required for their practice. There's just no surrogate from that, until Neuralink or one of these other startups can inject knowledge into our brains and basically train the neural pathways without us having to actually experience any of this stuff. It's just not going to happen, in the sense of, oh, we can completely replace it. It can augment. It can help streamline people like myself who have over a decade of experience. And so we already know where we're going with it. We already know how we're architecting this contract, this deal, whatever legal document that we're dealing with. But it's not going to make, my view of it is, it's not going to make that big of an impact. It'll expose, probably, way more of the weaknesses for folks that have been able to hide behind just doing the [rote] work, but I don't think it's going to really upend [that services arm].

25:05 Jon McLachlan

Yeah, I totally see that. I think that all of these LLMs and the code completion models that are out there, they're another tool in our toolbox. Just like an IDE didn't turn a junior developer into a senior developer. But it is a good tool to help save time, to help save some money, maybe ship some code faster, maybe find some bugs earlier in the process rather than later. And I see LLMs and all of these new models and the advancement as another tool. Not here to replace people, it's here to enhance. You used the word augment. I like that word. [It's a good point]. In the security space, I grew up as a hacker. And we've talked about this before. Like, hacker had a different term back then. Like, hacker is someone who likes to tinker. Like, these days, I think they're more like DIYers or stuff like that. But I do think about it from a, like a, how do I break a system type of perspective. If I was a kid, I would be [wetting] my chops seeing all these folks try to build APIs with LLMs. I think that is one of the dumbest things I've ever heard. And especially from a security perspective, like, I could think about so many different attack [surfaces] when we're introducing prompt engineering into API endpoint calling and payload formulation. This is, like, basic stuff that you learn as a security professional, with command injection and stuff like that. Now they're just opening it up for you and just saying, hey, actually, you can send any prompts. We're going to try to use our LLMs, which is a probabilistic model, to interpret what the API call is supposed to be. Here's some dice. It's, in physics, it's the difference between the probabilistic, super small particle science that's all probabilities, versus, like, the Newtonian stuff, which is [pretty set]. Both of those were like that perspective. Yeah, it's different.

26:56 Tony Thai

Yeah. From a, we see it too, in a sense that maybe you take it more on the development side, but like, at the end of the day, we experiment with a lot of these, you know, newfangled tools, only to realize, nope, we're still going to have to do work. It's a, it reminds me of being a kid and reading CliffsNotes. I don't know if you remember that.

27:12 Jon McLachlan

Oh, I remember that. Yeah. Yeah, I mean, like, ginormous, or read just a couple of pages.

27:18 Tony Thai

Yeah. Yeah. Now that I'm older and understand how this works a little bit better, like, there's only so much information you can get from an abstraction layer, right? There's only so much you can get, and to have utmost trust in coursework based on CliffsNotes. And maybe some people are listening to me, like, I did great with CliffsNotes, screw reading. You realize over time that even if it saves you time, like, it doesn't eliminate the need for someone to have reviewed the code or reviewed the contract thoroughly. And not only just review, not only just review, but understand why it's bad, right? If, something that I try to remind people of, this, everybody thinks AI is going to drive everyone to the right side of a curve. That's not how big data works. You've got the law of big numbers and regression to the mean, and most data is garbage, and most public code is, guess what? Garbage. And what's being pumped out to you, unless it's fine-tuned. And in many respects, like coding, much like legal work, it's subjective in many senses, right? Everyone's got a little bit of a different style, a little bit of a different pattern. It can't be relied on for high-quality work. It just can't. You still got to rely on people to understand that overall architecture, both in software and in legal work.

28:38 Jon McLachlan

Yes, yes. What have you learned from your experience from building HyperDraft?

28:43 Tony Thai

So many things.

28:44 Jon McLachlan

Like, just a top item, the one big one. Oh, you'll never do something again because you learned this, or you'll never do something the same again because you've learned this thing.

28:56 Tony Thai

I learned this recently, like, embarrassingly. It actually had to do with something we were architecting. We're proud on our, like, software architecture, but when I originally started the business, the thought was, like, delegate, scale, right? Boy, was that a mistake, man. Like, architecting a solution with the intent that you can delegate certain portions, because somehow you would be able to scale faster, is such a mistake, because you end up paying for it, or somebody pays for it down the line, and you end up redoing the work twice. And not to say that you can't architect stuff where you can build it as a team, work on it as a team, but you should never build or design with delegation in mind. You have to understand from the ground up how the system works. And that will build a more robust product and more robust architecture than thinking, oh, somebody else is going to handle, like, the little gritty parts. So that's a bit more abstract, but yeah, I don't know if that resonates with you or.

29:51 Jon McLachlan

No, it totally [does]. I've been bit by that once or twice. Thinking, like, oh, I'm just going to, here's a little thing, a black box that [I] can just push over here. And then the thing that comes back is not a, doesn't fit the bill. Maybe it did by the API, but how it did that underneath the hood, is no way it's going to possibly scale past a thousand, or can't scale it horizontally, or needs vertical scale, or all sorts of little [guys]. Or, like, the [instance], and you're like, no, why is there state flying around? You could have been stateless, like, if we'd approached it this other way. I've seen that once or twice.

30:35 Tony Thai

Now that you've brought up physics, which is one of my favorite topics, like, I'll bring in two concepts, right? One, you can't create energy, right? There's [the laws of] thermodynamics, right? Like, the preservation of it. And so if we're taking shortcuts, we're paying the price somewhere, somehow, right? That's it. And then, as I remind people with cryptography all the time, the strength of cryptography is directly proportional to the amount of energy put in, directly. So if it didn't take that much compute, didn't take that much time, then the security is lower. That's just how [hashing] algorithms work. That's just how it works. Yeah. So, if you think about that with any task in life, if you take a shortcut, you're going to be paying for it somehow, some way. And [some] things are okay with that, and some things aren't.

31:24 Jon McLachlan

Yeah. And I also, even with all of that in mind, keeping that top of mind, I also will always try to bring a sense of understanding, maybe even a sense of, like, compassion or empathy towards decisions that were made in the past that were maybe the right decision to make in the past. That then you get to a huge success, and then all of a sudden, those decisions in the past look like horrible decisions. Why did we do that? And actually, if you go back in time and you relive the context that you had back then, it was actually the right decision, even though it doesn't work. And so, when I bump into those things, those are, like, a class of their own. And what I love to say is that the right decisions that we made in the past are not the right decisions to get us to the next 10x or the next 100x. And then everyone feels warm and fuzzy about the success that you've had up from the point when that decision was made up to the point. Because it was, if you had made a different decision, the company could have gone into a different direction, because maybe you spent way too much time solving something that didn't really need to be solved yet. So just having compassion for ourselves, I think, is, like, super important for mental health.

32:36 Tony Thai

I laugh about it because so often, like, the engineers, when we go through the code, they'll see comments. I comment a lot. People ask me how engineering has influenced my legal skill set. Seldom do people ask how my legal background has influenced my engineering skill set. I write a lot of comments. And the funniest ones we run into are ones where I'm like, note from Tony, I know this is stupid, right? Like, guys, I know we're going to fix this. And if you're the one fixing it, I'm sorry, I owe you a coffee. There's plenty of those in the code base. I remember I saw one that I wrote. I wrote a line, like, three years ago. It was, like, a throwaway line. I didn't think we'd look at it. So I'm reading through the code, I think a lot of people can relate to this. I was reading through the code, and I'm like, what piece of crap, dumb person wrote? And you click on it, and the git blame's at Tony Thai, and it's, ah, yes.

33:33 Jon McLachlan

Yeah. It happens all the time, man. It's pretty funny. It's really funny. Yeah. And with Git, everything is accounted for. It's right there. You can go back. You can even look at the commit message that showed up with it. What was happening back in the day when that was, like, the thing.

33:47 Tony Thai

Listen, you said you give people kind of room and empathy for having made a decision at the time. Like, I do the same, and [I've grown], right? Every day that I'm writing code, I'm getting better.

33:59 Jon McLachlan

Why LLMs are not going to replace humans. That's a hot take, brother.

34:03 Tony Thai

I'm in agreement with you that it's a hot take, but apparently, I don't know why, but yeah.

34:08 Jon McLachlan

Well, because you're going against all of the grandiose visions that the bubble had. It's the religion of the moment. Again, I jump on board. You're a heretic. From a [hacker] perspective, there's so much in me that just wants to mess with all these, like, all these different systems, because I want to be the guy that [feeds] misinformation, right? Like, that would bring me so much joy.

34:33 Tony Thai

I saw one recently that really got me going. This is the intersect of the two. There was a car dealership. Have you seen this?

34:42 Jon McLachlan

No.

34:42 Tony Thai

A car dealership that had a chatbot. And with prompt engineering, one of the users was able to get them to offer him a car for, like, a hundred bucks and agree to the deal. And I'm thinking to myself, I'm like, he did it as a joke. Imagine. Just imagine the possibilities of, like, how much access to a system you can get. Could we do that with a house in California? [I'd] be down for that, man.

35:07 Jon McLachlan

Okay, okay. What do you think the impact, what would you like the impact of the technology that you're building in HyperDraft have on the greater, the bigger picture, and the legal community in general?

35:21 Tony Thai

One of the things I mentioned earlier was, I want lawyers to spend more time with their clients. I know it sounds weird, but one of the reasons why we want people to save the time is to spend more time in their community, talking to clients, talking to their network, community, whatever it is. Because I think that's where the real value is. There's only really what I consider two really human practices, right? There's the medical field, doesn't get more human than that. We're maintaining the body. We're focused on the body. And then the law is social governance, figuring out what binds us, what responsibilities we have to go through on the day to day. As the work has gotten more grueling, more technical, all that stuff, I feel, what I've seen is a detachment between the lawyer, what used to be referred to as the counsel, that's why we refer to lawyers as your counsel, and moved away from that to technicians. And we'd like to bring that back, because I think we can solve a lot of problems with that, the raw compute of a lot of these brilliant minds [who are] just stuck in the mire of doing the day to day, and not thinking about, nor should they be thinking about, how to remove those bits, so that they can spend more time with their clients and other people.

36:29 Jon McLachlan

Yeah, I love it, because at the end of the day, it is always people. And that's actually a litmus test that I use every once in a while. I find myself doing something, and I can't very quickly articulate or convince myself that I'm improving the life of a fellow human being. And it doesn't even matter if that's, like, a fellow engineer, a salesperson, a customer, or even my future self. I put myself as a human being, too. They're my future self. If I can't convince myself that someone [unclear].

37:00 Tony Thai

Yeah, yeah. Okay, I'll take that. I'll take that.

37:02 Jon McLachlan

Yeah, I stop. And you have to always make sure that the things that we do are connected back to a human, right? Even if you're in the code and you're engineering, and you have to make a decision about, do you use a signed int or an unsigned int? And our non-technical readers are not going to understand what that is, but it's just a very nitty-gritty technical thing that no one ever thinks is going to have an impact on someone's life, but you actually could, and you actually can. And if you can't see the connection between those two things as an engineer, just go down and see the bigger picture.

37:36 Tony Thai

Do you have a, can I tell you? I'll tell you just a quick anecdote on that.

37:40 Jon McLachlan

Please.

37:41 Tony Thai

Because we can be taught stuff all day long, but until you make mistakes, like, I truly believe that until you make mistakes, those neural pathways aren't really that embedded, because I've been warned about all this stuff.

37:52 Jon McLachlan

That's right.

37:53 Tony Thai

When I was a very junior engineer, I'm saying, like, very early twenties.

37:57 Jon McLachlan

[Working in this] finance system when you were eight years old.

38:00 Tony Thai

Yeah, yeah, yeah. Okay, all right.

38:02 Jon McLachlan

[unclear] more than eight years old.

38:04 Tony Thai

Yeah, yeah, yeah. [I'd be surprised if we] trusted an eight-year-old with the system. But I was working with a big accounting system. We were transitioning over from one system over to something that I was working on, and did all this testing. D-Day came, transferred over the primary interface to what I built. And immediately we started running into issues. Like, records were not going through, numbers weren't adding up. And I was scrambling. This is, like, one of my first jobs. I was so stressed out. And then I was going through the records. I'm like, this isn't, this is a financial system. So the minute I'm going through records, there's 10,000 records going in a minute. It's not like I can just sit and go through all of it. And this is way back in the day, right? Our compute's low. Stay up all night. I finally figure out what it is. Users are typing in O instead of zeros into these fields. So I'm excited that I feel like I've solved the problem. Go to my boss at the time, and he just chews me out. And I'm looking at him like, dude, it's the users that are stupid. Like, I'm not the idiot. And he's just like, your job is to anticipate stupid. Your job is to anticipate humans. Like, developing software is for humans, not for other robots who know the input. And I'm thinking to myself, [who] puts an O instead of a zero? Like, [what moron does that]? And he was right. And so, ever since that point, like, I took to heart, like, it's my job to protect against stupid. And, like, clients are, no, but it's really just ignorance, right? Like, my job is to protect against ignorance, both on the legal side and on the software side. So, yeah. Anyways, I forgot how we got on that point. Oh, you're talking about building for humans.

39:42 Jon McLachlan

Human and human, yeah. Yeah, humans are what matter, always. 100%. No, I like that. I like revisiting, like, old stories. And speaking of revisiting, [if you] could go back and visit your younger self, two questions. Would you? And if you would, you have any advice for your younger self?

40:00 Tony Thai

I would slap the crap out of myself.

40:02 Jon McLachlan

[Some room] for self-compassion?

40:03 Tony Thai

Oh, yeah. Well, not for myself. I'm the dumbass. I've said before, or I've been asked this question before in, like, smaller circles. And one thing I've always said is, listen more. Like, when you're young, like, you're always gung-ho. You always think of the right answer, all that stuff. And I like to think that I picked up on listening more earlier than most. If I'm going to give myself some other advice, it'd be something that I constantly still work on, which is, like, compassion, and having more empathy and compassion around, like, other people's perspectives. Not everybody thinks the way that you do. And I think, for me, growing up as a kid and also as a professional, like, I constantly thought that people wanted to operate the same way I did.

40:52 Jon McLachlan

Where did you grow up? Where did you grow up?

40:54 Tony Thai

I grew up in Garden Grove, Orange County, California, so in SoCal here. And then the family moved out into the Inland Empire, in [Corona]. And then from there, honestly, like, I would say home base for me feels like, and has been, LA, because I went to UCLA for undergrad, which is in Westwood, part of LA. And then I went to USC for law school, which plenty of people give me crap about.

41:17 Jon McLachlan

But why do they give you crap?

41:20 Tony Thai

Are you not familiar with the...

41:22 Jon McLachlan

I don't know. My stereotypes are not very good.

41:29 Tony Thai

[unclear]. Yeah, okay, great. That's one of the stereotypes. You see, [UCLA and] USC are like crosstown rivals, like, in a fun way.

41:36 Jon McLachlan

In a fun way.

41:37 Tony Thai

But yeah, back when both football programs were excellent, yeah, there's a lot of rivalry between the two. So we were called turncoats, right? There's a name for us, like, anyone that converted from UCLA to USC.

41:48 Jon McLachlan

So I see. Kind of what happens between Minnesota and Wisconsin. And as far as Minnesotans are concerned, anyone that lives in Wisconsin or identifies as a Wisconsinite is a [cheesehead]. So we say that with all of the love in the world. So I'm sure [you're] labeled turncoats with all the love in the world.

42:07 Tony Thai

I'm sure. I'm sure. Yeah. Yeah, exactly.

42:09 Jon McLachlan

So we had a lot of entrepreneurs, actually, that listened to this show. And this is a little bit of a leading question, but I'm super curious to hear, and I'm sure some other people might be. If there is one pain point that you just wish someone would step forward and solve, and knock it out of the park, and maybe build a company around, is there any pain points that you have, that you see over and over again, that you would definitely pay money to have just disappear?

42:36 Tony Thai

First off, [the lawyers]. That's why I went to law school, [unclear].

42:41 Jon McLachlan

Yeah, you're very engaged in solving [that one], yeah.

42:44 Tony Thai

I think we are solving for one of them. I would say, just because we just passed tax, the tax deadline, man, I think accounting systems are still broken. It's, oh, I hate [that].

42:56 Jon McLachlan

Like, it is. I tried to use, like, all of the standard solutions, I don't even want to say their names, and nothing integrated with anything else, and I had to do it all manually again. And then I heard the IRS is now [farting] around with some trial to allow taxpayers to submit simple taxes directly in, you know, digitally, directly to the IRS, instead of going through TurboTax or some third-party, like, service or something.

43:23 Tony Thai

Yeah, it's so broken.

43:24 Jon McLachlan

The fact, it's definitely a broken system. We know the rules. We got to play it. The thing that bugs me is that we wait until the end of the, we wait for this, like, climax point where all this stuff has to lead into filing the taxes. Like, I was literally thinking about this the other day. I'm like, this should just be a monthly hygiene thing, right? Like, why can't I just update something every month, and then April's just another month? Like, why does everyone wait for me to collect information? And our stuff is, obviously, a little bit more complicated than a simple filing. But I think a lot of founders feel that way, because, like, they've got their own investments, they have to worry about their taxes for their company. It's a complicated affair. And, like, also have another business to run on the op side. So I think a lot of founders can relate to accounting systems being, like, crap. That's all.

44:16 Tony Thai

If you ask me, like, okay, once you solve the legal one, where'd you go into next? It might be accounting, man. It might be the fact that, Jesus, this should be way more binary in terms of decision making than legal. Like, why is it so difficult?

44:30 Jon McLachlan

Yes. Yes, please. Yes, please. I'll second that one, actually. Tony, this has been an incredible conversation. Thank you so much for your time. Great discussion. I'm your host, Jon McLachlan. And thank you to all of our listeners as well for tuning in to another episode of The Security Podcast of Silicon Valley, a YSecurity production. And please stay tuned for another episode. Tony, been amazing.

44:55 Tony Thai

[unclear], brother. Thanks for having me.

44:56 Jon McLachlan

Thanks for joining. Thanks for joining. And thanks to all of our listeners again.

45:00 Tony Thai

Oh, and just a quick plug, if folks are curious what we do and are interested in learning more, they can check us out at hyperdraft.ai or any of our socials under HyperDraft.

45:12 Jon McLachlan

HyperDraft. You've heard it here first. Thanks, Tony.

45:15 Tony Thai

All right. Thank you.

45:16 Jon McLachlan

Thanks.