In this episode of The Next Next, host Jason Jacobs interviews Rob May, a seasoned entrepreneur and tech investor specializing in AI. They discuss Rob's extensive background, his pivot to AI, and his perspective on its evolution. The conversation explores how AI tools are reshaping the tech landscape, impacting startups, and redefining expertise across different industries. The episode also delves into broader implications, such as the future of M&A, investment strategies, and the changing skillsets required for founding new companies. They highlight the potential and challenges of AI in both personal and professional spheres, offering insights into navigating a rapidly evolving technological landscape.
AI Frontiers: Leveraging Technology for Founders with Rob May
In the inaugural episode of The Next Next, host Jason Jacobs speaks with Rob May, a seasoned entrepreneur and AI investor with four startups and over a hundred AI investments. The discussion covers Rob's journey into AI, the evolution of AI, the implications of AI on various industries, and the challenges and opportunities it presents for startups. Rob also shares his insights on the future of expertise, the changing landscape of business funding, and the shifting paradigms in technology development.
Jason grapples with how these advancements might impact his journey as a founder aiming for ambitious goals while maintaining personal and family balance.
00:00 Introduction to Rob May and the Episode
01:01 Jason Jacobs' New Show and Its Purpose
01:46 Rob May's Background and Career Journey
04:38 Rob's AI Journey and Early Ventures
07:44 Challenges and Insights in AI Development
10:52 The Future of AI and Human Expertise
12:10 Navigating the Rapidly Evolving AI Landscape
22:36 The Role of AI in Creative Fields
23:34 The Future of Creativity in the Age of AI
23:57 The Rise of Virtual Influencers
24:21 Branding in the AI Era
25:02 The Impact of AI on Media and Culture
26:16 Automation and Intimacy in Social Media
29:34 The Evolution of Talent and Earnings
31:12 Challenges for Investors in the AI Landscape
31:48 The Future of Business Moats and Defensibility
37:30 Starting Companies in the AI Age
41:21 The Role of Judgment and Experience in an AI-Driven World
44:52 Closing Thoughts and Future Outlook
Jason Jacobs: [00:00:00] Today's guest on The Next Next is Rob May. I've known Rob a long time through the Boston startup scene. He's a four time founder, a multi stage tech investor, and he's done over a hundred AI investments. He's a partner in HalfCourt Ventures, which is a fund that focuses on AI, and he also writes a popular AI newsletter. So you can imagine why when I wanted to learn more about AI and how I might leverage it to build different, Rob is one of the first people that I reached out to.
We have a great discussion in this episode about Rob's journey to becoming an entrepreneur, what led him to AI in the first place. how his thoughts on AI and how AI have evolved during the years that he's been focused on it. We also talk about his thoughts on the implications, the risks, the opportunities, how it might impact how startups get built and funded, and where all this might be going.
We have a great discussion, and [00:01:00] I hope you enjoy it. But before we get started, I'm Jason Jacobs, and this is the next, next, it's not really a show. It's more of a learning journey to explore how founders can build ambitious companies while being present for family and not compromising flexibility and control, and also how emerging AI tools can assist with that. Each week, we bring on guests who are at the tip of the spear on redefining how ambitious companies get built.
And selfishly, the goal is for this to help me better understand how to do that myself. While bringing all of you along for the ride. Not sure where this is going to go, but it's going to be fun.
Okay. Rob May, welcome to the show.
Rob May: Yeah. Thanks for having me, Jason. I'm excited.
Jason Jacobs: Well, this is the very first recording of my brand new show. It's been, gosh, I think well over a year since I've [00:02:00] recorded anything for anyone. So, I mean, since we're of similar age, I kind of feel like when, when Stella Got Her Groove Back or something.
Rob May: I get the reference.
Jason Jacobs: Psyched that you're willing to be the guinea pig and come on the show.
Rob May: Yeah.
Jason Jacobs: Well, for starters, Rob, I gave you a little context on what this is about, but before we jump into that, just maybe talk a bit about yourself and what you're up to these days.
Rob May: Yeah. So first of all, yeah, I think the interesting thing that I bring to a lot of podcasts these days, so I'm a hardware engineer by training. Not a lot of people know that, but I, when I came out of college in 2000 doing FPGA and ASIC design, it like computer chips didn't matter.
So you had to go work in like military and space industry. And now. AI is like crushing GPU workloads. And so there's all this chip innovation, so it's really interesting time to be working in ai. So I worked in chip design for a while, worked for a couple of startups, and then have done a couple of companies.
You and I met Jason when I was doing Backupify in Boston and are in that similar [00:03:00] cohort of people that started companies in the '09
Jason Jacobs: Fogey founders. I'm, you know, I haven't heard the term before, but I'm determined to make it a thing because we are a demographic and who's talking to us? Nobody.
Rob May: I know exactly. And it was fun dude. That was a super fun time to be building the companies. I feel like, particularly because social media was new and useful, but still nerdy and not overly politicized. And it was fun to be on and all that. But, yeah have done a couple of companies, had one blow up in my face, had one really good exit, two mediocre exits and, have been doing a lot of angel investing.
And so now I run a fund called half court ventures. We're on Fund4, we've been investing in just AI since 2015. And then, I'm always working on some other projects. I'm working on a company, this is a side project called Neurometric with an old, co founder of mine. Where we're looking at a bunch of, new AI hardware, and how to build software that straddles these different architectures that are coming up.
You got all these cool AI chips, people don't know how to program them. Working on some stuff like that and, Yeah, just doing stuff to build AI community. I run a group called the AI Innovators Community in Boston and New York, and we do some events. [00:04:00] And, yeah, so that's me.
I just, a little bit restless, maybe, all the time
Jason Jacobs: well, as you know, there's a couple of things on my mind. For myself, I mean, one is just, being a fogey founder and trying to build another company, but being, kind of at an age and at a life stage. And, you know, with my kids and family, where I can't do it the only way that I know, which is to, hurl myself into the fire, and have it be all consuming.
And then I'm also, there's just so much transformation happening, that's AI driven and its early days, although not as early as 2015. And just, starting to look at how those tools might, enable someone like me or others who are. feeling or thinking similarly to build different.
So, I guess for starters, maybe talk a bit about what led you down the AI path in 2015 and what some of the biggest changes are, and surprises as you sit here in, early 2025. I had to remember the date, but yes, 2025.
Rob May: So my AI journey actually started in 1998. And what happened was, so I was at the University of [00:05:00] Kentucky, and they had a dual degree program, and the dual degree program allowed you to get an MBA, And some version of an engineering degree, whatever discipline you wanted at the same time. It took 15 students every year.
I didn't really think much about it. I just, on the brochure, it was like people with an MBA make 20 percent more money. I was like sounds great. I should do this. And I had to take as part of my MBA courses. We took the classes at the same time. So I took graduate level MBA courses and undergraduate electrical engineering courses at the same time.
And the interesting thing was I started to take this like IT decision Sciences course as part of the NBA and it's super boring for me because I was pretty deeply technical and the guy said, Hey, when everybody do a project about something about the future of, something related to IT. So I'm looking through the book and I should find a topic and I get to the very last chapter and it's talking about like virtual reality and artificial intelligence and all sounds like, Oh, artificial intelligence sounds cool. And so the next semester I like, Okay. So I did my report on AI and I got so into it that like the next semester I like sat in the back of all my classes and read AI books instead of [00:06:00] focusing on school, and all that kind of stuff.
Got really into it, got a job in hardware design, in Florida working for Harris Corporation and started a graduate degree down there in AI. But it's like 2002 when I started and A. I. At the time was like symbolic logic programming and lisp. And so I got about halfway through and I was just like, this stuff is brittle and doesn't work and blah, blah, blah.
So I dropped out of that and I did not work in the AI field, but I continued to follow it. And I continued to look at projects that people were doing interesting stuff, read a lot about it. So after I sold my first company, which was December, 2014, now it had been 10 plus years. And I was like, okay, like what's next?
I'm still young. I want to do something else. So I looked at four areas. I looked at corporate messaging because Slack was a new thing and I thought, oh, there's other problems to solve here. I looked at IOT. I looked at blockchain and I looked at, AI and, IOT. I was just like, all kinds of shit's gonna be connected to the [00:07:00] Internet.
It's gonna be great. I don't know what to do with that. Ruled that out. I read all the blockchain papers, I read the early, Bitcoin papers, the Sidechain paper, all that kind of stuff. And I was, at the time, I was just like, I don't get it. I don't understand what this is useful for. So it came down to AI and corporate messaging.
I actually picked corporate messaging and a friend of I friend and I, and I, went out and built a tool that was supposed to be like, better than Slack and all kind stuff. And it turns out one of the problems with messaging is humans change contact context on threads. It became really hard to track.
I might send something to you for one reason, and you send it back and copy somebody else for a different reason to pull them in. And our goal is if you could keep track of the context, you could intelligently sort an inbox. But the context on a single message thread kept changing, is what we found.
We know how to solve that problem. And so I happened to read, about the same time, Google came out with a paper, on what's called word vectors, and it was a pretty stunning paper for that time, because, for those of you who don't know word vectors, what happens is, you take language, and you embed it into a mathematical vector space, it's a multi [00:08:00] dimensional space, but let's say three dimensional space, because we understand that so assume it's a three dimensional space. Words, are mapped into vectors, where if the words are similar, the vectors are parallel.
So they did this, so that allows you to do math with words. Because it maps the relationship between words as how they're used in language in this vector space, so the math that they did on words in that paper was very famous for a while, and they basically said, King minus man plus woman equals what?
And the machine said queen. Because it took the vector for king, it subtracted the vector for man, it added the vector for woman, and that's the point that it got was the word queen. And people were like, wow, that's really cool. And I saw that, and I was like, this is going to change things.
I started a company called Tala with the idea that, machine vision had its big moment in 2012. With a new neural network called AlexNet, that was the first GPU trained neural network to compete [00:09:00] and win a contest. And this was like 2015, and I was like, in the next two years this is going to happen for language.
Turns out we were off by about three years, but we started going down that path. And the idea behind Tala was to build a customer support chatbot that you didn't have to script anything. It was just like, ingest all the documentation, we should just be able to generate an answer. We ended up selling that company for about a third of the money we raised.
And two reasons we weren't successful. One was, we needed about a thousand times more computing data than we had. And we didn't understand and know that, right? We were trying to solve it with, we thought it was a data annotation problem and it was a compute problem. The second thing was we were trying to sell into large enterprises.
And trying to hook new technology into these old systems and get access to their data so we could annotate it and train on it was like, on premise systems we'd never heard of, and so we ended up doing a lot of custom work that was hard to productize.
But it was fun to be part of AI at that time when people said it was ridiculous and wasn't going anywhere. I also started angel investing, created Half Court Ventures fund one, which was a $3 million [00:10:00] fund. Which has been a fantastic fund. It's probably going to be a five or six X net fund.
We still have a couple of portfolio companies to exit there, but we got in some great early deals and, met some awesome early AI founders. And it's just, it's just been awesome to watch how fast things change. And it's this weird space where I think people forget AI is a collection of technologies really.
And so it's not like one thing. And so as a result, it's very nuanced in how you think about it and where it works and where it doesn't it's like simultaneously overhyped and underhyped. Because you have these people that are like AGI is around the corner. It's going to change everything.
Everybody's going to be out of work. I don't really believe all that. And then on the flip side, you have people that are like, this is just fancy statistics and not a breakthrough and doesn't mean anything. I don't believe that either. There's some pretty impressive stuff happening.
But the impressive stuff is pocketed and not across the board. It's my take on the state of the world today and how I got to here.
Jason Jacobs: Sitting, where we are today in early 2025, I mean, I'm kind of new to the AI world, and I'm not [00:11:00] even, like, this is not an AI journey, this is a journey to figure out how to build different and have outsized ambition while having, you know, personal and family ambition not compromise, you know, like, equal parts, like, family ambition and, and professional ambition.
What's interesting to me is like, wow, with AI, you could do more with less. And maybe you could also, have a lot more flexibility and control as well, in terms of just like. When you work, how you work, how much you work, et cetera. But what I've found so far is like, well, there's these LLMs and gosh, they seem to be iterating super quickly and like, are they just gonna kind of gobble everything?
And then you've got all these little tools that are using a lot of the same words and there's so much overlap between them and even actually same thing with the LLMs like so much overlap between them and it's like well Which one do I use for what and how do I stay on top of all these different things?
I'm like, how do I figure out what my personal stack is? And if my personal stack, you know If I figure it out for today, is it even gonna be relevant in three months given how quickly everything is [00:12:00] changing? And like, what should humans do? And, and what should the machines do? And like, and for which use cases, is it okay for the machines to do more?
Like, it's just daunting. Like it's, it's almost paralyzing. So I guess as a newcomer coming in, you know, what advice do you have for, you know, for someone like me and what's your take on, on kind of where the landscape is today and where it's going.
Rob May: Yeah, I, actually I wrote a post about this a couple years ago. One of the barriers you run up against, everybody's probably seen that famous chart where it looks at penetration, like 70 percent market penetration. It's The radio takes, I don't, I'm making this up, but the radio takes 40 years and then the TV takes 30 years and then color TV took 20 and then cell phones took 15 and then, or the internet took 15 and then cell phones took 10 and then Chad, GBT took two, like whatever.
And it's like the cycles keep shortening, but at some point there's a limit, like as a human, you have to learn how to use a new platform and at some point the technology. Can only [00:13:00] change so fast because you can't, I can't learn a new UI every day, right? Even if it's better and even if it's whatever, like at some level it's I just want the thing to work the way it worked yesterday because I the benefits don't benefit me that much to jump to the new thing.
I think technologists have to understand that there are human limits to how fast we can change and learn new platforms and all that. I also think it's why a lot of things are going to collapse into text because text is not the best platform and it's hard to specify things sometimes and describe what you want compared to using visual tools.
But at the same time, it's an interface that you can put a lot of stuff behind and deal with. So it's part of the reason I'm big on agents, but not for every use case. My advice to people would be, when markets are moving quickly, I think you have to experiment and stay lightweight and expect that things will change.
I would not try to particularly for businesses, but also for people, I would not try to [00:14:00] think of it in terms of like, How do I build the AI stack that I'm going to use? I would look at it as how do I commit to using a tool for 90 days so I learn and then see where it goes or what's new and eventually follow on to the, settle into the tools that are for you.
But some of it is just making sure you understand this stuff so you can make good decisions as new things come out because it's very easy to be misled. And again, there's a lot of nuance in how you use these tools. So let me give you an example. If you're a writer. If you're a bad writer, ChatGPT does a great job of making you average.
But if you're a great writer, you probably use ChatGPT for some idea generation, but not much else. ~I don't know that Stephen King's sitting there asking ChatGPT to help him with word selection or whatever. He's way~
Jason Jacobs: ~Did, did you, ~did you see that BCG, study that Ethan Mollock,
Rob May: Yeah, on the Jagged Frontier? ~Yeah.~
Jason Jacobs: ~exactly what, ~what they found, was that for BCG consultants, it helped the bottom ones get closer to the mean, and while it helped the top ones too, it helped them significantly less.
Rob May: Yeah. But there was a Microsoft study That did the reverse for programmers, right? If you're a writer, what's happening is, you're a great writer, [00:15:00] you're a bad writer, the new floor is here. That's been compressed. If you're a programmer, you're a great programmer versus a bad programmer, it does this, because, if you're a bad programmer or a novice programmer and you're using these tools, sometimes the tools make mistakes, you don't really understand the code that was generated and why, you can't fix it, you can't tell if it's right, it actually slows you down.
If you're a great programmer and it can abstract away and automate a bunch of the basic tasks that you would normally have to Churn out some code and any problems that you see you can fix really quickly and edit the code It just makes you super productive So that's what's hard about these AI tools, right?
As you look at the things LLMs can do in some cases it raises the bottom and in other cases it raises the top
Jason Jacobs: I'm finding that with coding like I was never a coder before it's like hey like everyone's saying you can use natural language and code It's like alright. I can like whip up a prototype But if I try to get anything anywhere near production like you know I need someone with more training than me because, like, there's no machine, at least today, [00:16:00] that's gonna, you know, that's gonna help me get anything done even remotely as productively as someone that's got even basic training can do.
Rob May: Yeah, that's true so it's really interesting to see where it's going like that. And look, I do think these things will get better and better, but take the coding example, say it does get to the point where these tools know how to generate a lot of stuff.
You still need some specific built in knowledge because, you're gonna run up against your own ignorance as a user of what you want, right? Like I might be able to specify what I want in a CRM, But would I use one of those tools to rewrite a payroll system? I don't even know all the rules for payroll, right?
There's a bunch of tax crap and state by state compliance crap. And maybe at some point, the AIs will all know that and it can just be like, Oh, if you need a payroll system, you have to do this. But people forget, in almost every system, even People that don't work in fields like accounting or law, they think of those fields as being very black and white, but they're not right.
And in accounting, there's all kinds of different ways to do revenue recognition. There's [00:17:00] all kinds of different ways, for the same company, there's all kinds of ways to decide, should I put this cost in SG&A or somewhere else, right? Like you, you can classify things differently and they can make your margins at certain areas of the business look better or worse depending on what you want to do.
And the same way with law, right? I deal as an investor, we deal a lot with 1202 of the tax code, right? Qualified small business stock. And you'd be shocked how many times we get in situations where the lawyers are like, huh, that's an interesting situation. We don't know if it's 1202 qualified it would seem to be black and white, but it's all very vague. And so you're constantly making trade offs and assumptions. You're constantly playing probabilities. And AI can't always or won't always be able to tell you what trade off to make. And so you're still, as a human, going to have to decide how much risk you want to take, what you want to optimize for.
It's really interesting. Think about if you built an AI program to maximize your health. What does that really mean? If you followed Brian Johnson, this guy who's doing blueprint spent 2 million a year optimizing his body so he doesn't [00:18:00] age, he's starting to run into the trade off issue, right?
Which is let's say there's some substance you can take some vitamin or pill or whatever, and taking that might lower your risk of heart disease by 8%. But it'll counteract something else that was lowering your risk of cancer. By five percent. So it's okay, do you want to optimize for lower heart disease or lower cancer risk because they're different things.
Do you want to optimize your body to run marathons or be a power lifter? Because you can't do both. It just doesn't work that way. It's impossible. And I think people don't understand. I think we have this idea that AI is going to just solve everything and we don't realize it.
The world is fundamentally probabilistic and there are trade offs at every level and We're going to have to deal with those even as AI does a lot of stuff for us.
Jason Jacobs: Well, gosh, I mean, as you were talking there, there's so many different directions that we could go. My mind is racing, but, let me pick one of them, for kicks, and that is, expertise. If you look at expertise today, I am a guitarist, or I am a songwriter, or I [00:19:00] am, a plumber, or I am a coder, or whatever, right?
Like, Pick an expertise. It seems, and, and again, your, your point about how certain use cases it helps the top half and certain use cases it helps the bottom half today, right? But if you look generally over time, is expertise, going to be less Valuable and less unique over time.
Rob May: Wow. That is a great question. I think it's going to be conditional. And here's how I think you figure out what it is and what it isn't. I think in areas where your expertise is because you know a large set of rules. Legislation, understanding, legislation that's been passed, probably in accounting and things like that, I think expertise on those core things, like I know a lot of the rules are, is going to go down, because the AI is going to know more rules than you can know, and it's going to be able to keep up with the changes.[00:20:00]
If your expertise is Synthesizing information. So like I think being a CEO or an investor or something like that is going to be or any maybe a product manager. Anything that's cross functional is going to be more valuable because we're a
Jason Jacobs: How convenient that you're saying the thing that would be more valuable is the thing that you are, Rob.
Rob May: Yeah, what everybody thinks. But I think what happens is that I think synthesizing information is where these things are still going to be pretty far away. And so I think, even in areas like, the areas that are rule based, the gray areas I think will still be very valuable.
So for example, take consulting like McKinsey, junior consultants, junior investment bankers, stuff like that is going to be, that expertise is going to go away, and be automated away.
Jason Jacobs: Then what's the farm system to become senior?
Rob May: I know that's, what's hard, right? Because you're still going to need people at the top that know how to. Like an investment banking, like craft a deal, right? Or pitch to a customer if you're [00:21:00] McKinsey.
Jason Jacobs: Like, what happens when those people die?
Rob May: It is. It's going to make the farm system. Maybe it's more of an apprenticeship where instead of doing the grunt work to get there, you just shadow somebody for years.
Although my counterpoint to that would be the way they typically teach advanced topics is, They still teach you the basic stuff like, we all learn how to work out multiplication tables, even though we just memorize them, right? But you learn how to do multiplication that three times three means adding three, three times, right?
Same way with calculus, right? They walk you through the whole definition of what a derivative is. And it's the limit as the, as things go to zero and the change here and whatever it is, I don't even remember. And then you work through all that and it's a big pain in the ass and they teach you the power law and you're like, This is so easy.
Why don't I do this? And , because we want you to understand what a derivative is. What it could be is there could be some like either training program that gets you through how something works, but then you don't actually do the work maybe they keep 20 percent of the work that is still done by humans, both number one as [00:22:00] a check and number two, in case the AI breaks or makes a mistake.
And then number three, for reasons that people learn, right? It's like a lot of systems still have backups. A lot of people still make paper copies of things that they never look at and they only refer to if there's a disaster and they lose their digital records or whatever. Yeah, so it could be some situation like that, but yeah, I don't know, man.
It's hard, right? Because this is one of those situations where if you the way the world turns out is going to be heavily dependent on the decisions that we are making now about how we want this to work. And so you want to make sure you decisions that optimize the short term at the expense of the long term.
Jason Jacobs: So let's, I mean, that's for the rule based stuff. But if you look at the creative stuff, I mean, here's something I've been thinking about. So, I've been kicking around with this. Cambridge based company here in our backyard, Suno Music. And, and it's cool because you can use natural language to enter a prompt, and it'll write a song, right?
And so, hey, I just turned into a songwriter, sort of. And so for people like me, it helps me be more creative. But what about the creatives who have been songwriters for decades, professionally, Does that devalue their skills? Does [00:23:00] it increase their skills? And, on a similar line, and then we can tie back into like, what do you think about, about all of this?
I've got this idea for the series Fogey Founders, where it's like Silicon Valley all grown up. It's focused on middle aged founder problems and quirks versus the 20 something founder problems and quirks. I'm really excited about it. I think I could test it out with characters, you know, on an individual basis, do character development on someplace like TikTok or Instagram Reels,
so I started reaching out to, some creators. I slid into the DMs of this, big creator, on Instagram and TikTok, who, does great stuff with character development. And he does all his own stuff. So I said, hey, what do you think about AI and where it's going?
Because I'm not a creator, but I want to do what you do without actually doing what you do. And he was basically like, fuck that, right? I don't want any part of that. Like, I'm going to do it all my And so I guess my question is like, you know, is like, in a way it makes us more creative and in a way it makes us less creative.
And so, What is the future for creatives and is it a good thing or a bad thing overall?
Rob May: Have you seen Lil Miquela?
=Jason Jacobs: I've seen Lil Miquela [00:24:00] but not for like three or four years so I have no idea what Lil Miquela has become. But I remember her when she was just getting off the ground.
Rob May: Yeah. She's virtual influencer.
Jason Jacobs: Maybe it was longer than three or four years ago. ~Could have been like eight years ago or~
Rob May: Yeah, I think she's six or eight years old now, but they've got, she's been out there for a while as a virtual influencer, probably the most popular one. And so that's pretty interesting. I think, man, you're asking some tough questions, Jason. The way that I would play it as an investor, if I had to handicap it is I think brand is going to become more important because so AI is going to commoditize a lot of execution skills and it's going to, what that's going to mean is everything is going to get easier to do for everybody.
Which means you're not going to know who to trust, and who to like, and there's going to be more of everything. And so you're going to need filtering mechanisms. And that's going to come down to branding and what people stand for, what brands stand for. People are becoming brands like, Mr.
Beast is, probably a top 500 brand in the world at this point, probably on par with, a lot of big companies. And I think you're going to continue to see that [00:25:00] trend. And the reason I say that is because if you look what happened at with Web 1. 0 when you look at newspapers used to be a great business, and I think a lot of people in 1995 or so would have said, Oh, the Internet's gonna be so great for newspapers because now if you're the Boston Herald, you can reach people all over the country and blah, blah, blah.
And what they missed was by lowering the barrier to publishing, you created more competition for publishers because now Two guys that wanted to cover a certain thing, could just cover it and they started competing for ad dollars and the way that you did that was, link bait and hacking the system and that's how online media got all screwed up and became trashy and low quality You know you have and so I think Problems like that are gonna happen in a lot of these things and you're gonna have the broccoli problem which is like we all know we should eat more broccoli.
We'd rather eat cake so you have that in online media. We, we'd rather look at celeb news and sports news and anything else other than economic data [00:26:00] about why inflation is or isn't happening. And so we just let other people tell us what to think about those important topics, because it's boring to look into it, given all the other media we could consume.
And so I think what's going to happen with a lot of these stars, you already see, I've looked at companies. That automate, chats on some of these platforms like Instagram or OnlyFans or TikTok so that you can, because people want intimacy, they want to feel you want to feel like you're following these people.
And I think what we like about him is you're like, Oh, this person's like my hero, or I'm really interested in their life. And now I can see They're walking their dog or they're drinking coffee in the morning or whatever you want to see. And then you want to be able to chat with them. But they can't have hundreds or thousands of chats a day.
But with these bots, they can. And but then these bots are like faking intimacy. I don't know where the economic value is going to accrue. I think brands are going to matter more. And people who have big brands, I think it's just you're, I think it's going to accentuate that power curve and they're going to become more valuable.
But below that, what the opportunity for sub [00:27:00] brands are, what the opportunity for technology is, I don't have a strong opinion.
Jason Jacobs: I mean, another related thread is, so let's say, there's Like a showrunner or a producer or, you know, a studio that builds different shows, as these AI tools emerge, there'll be firms that lean hard into them, and there'll be firms that, ~that ~take their heels in and say they want it.
No part of it. You know, same thing with songwriting, but also same thing with accounting and same thing with law and some of these more, structured pursuits. Is there a consistent answer in terms of directionally the firms that dig their heels in the ground, like what they become, like, are they blockbuster?
Do they go away or, you know, are they like the record player or, or the record? Are they going to continue to have a meaningful role over time? And also does it depend on what type of category you're talking about, the way that it does back to the whole, like it brings coders, the top half coders up and it brings, the bottom half creators up.
Rob May: Yeah, there's a [00:28:00] really good book called FilterWorld, and,
Jason Jacobs: called? I just want to write it down.
Rob May: And it's about AI algorithms and how they're homogenizing culture. I bring this up because you bring up Blockbuster,
Jason Jacobs: youngins, Blockbuster used to be a video store where you could rent videos. It was a big thing in Rob and I's youth.
Rob May: Yeah, exactly. And the guy that wrote this, Kyle Chayka, the reason that he wrote it is he was traveling internationally and he was like, why do all the independent coffee shops around the world look the same? Like you'd expect them to be different, culturally local, and they're not. And it's the Instagram effect.
To promote yourself and get people to come, you have to post certain things, and you have to have a living plant wall, and you have to have whatever you have to have, and so he started to notice that, and he noticed, so he started doing a bunch of research and realized that these things do flatten culture, the interesting stat is, I'm gonna make these numbers up, they're gonna be a little bit wrong, but they're gonna be directionally right, okay?
All of Netflix in 2023 streamed like 4, 900 different [00:29:00] movies or something like that. A blockbuster superstore at its peak carried over 7, 000 titles. So we actually had, the whole idea is this is gonna provide us more choice and all this AI stuff and all this tech and algorithms are gonna and he's it actually doesn't.
What it does is it forces everybody. It squeezes. There is no long tail. It squeezes it down because the way you find stuff is what is everybody else looking? What is everybody else watching? And, and so it's, it's had the opposite effect in a lot of cultural areas as what we expected. Nassim Taleb once wrote this idea that I think is relevant to your question about how the record player impacted music, which was, let's say you lived in a small village.
And the only way it's weird for us to think, cause we listen to music all day now, right? But if you heard music, in the 1800s or Before that, it was probably because somebody played it live and that was it, right? And if you were the one [00:30:00] dude in the village who knew how to play the piano or was the best pianist or whatever you could probably have concerts and people would come listen and pay a little or whatever.
Suddenly, when somebody better than you in another town could make a record and I could listen to the record of the better guy, your job as a good but not great pianist or violinist or whatever went away. And the guy who was the best sold way more and made way more money than he or she could have made, in the previous world.
And so that is how these things move from, bell curve distributions of talent and earnings to power law distributions of talent and earnings. And I think that's going to happen in more and more things, right? And back to my point about brand battering. Mr. Beast has a burger chain and a candy bar line and stuff like that.
Attention and distribution matter so much. And so if you can garner attention, you can go into almost any business that you want. And that'll hold true because most businesses are going to be easier to do. With [00:31:00] AI without always having the deep expertise. I think it's just saying it's going to be better to be at the top of your game in some fragmented sub niche than it is to be, not top of your game in something that's a bigger industry.
So think about that as an investor. How does that play out as an investor? We're always concerned about TAM. How big is the TAM? Why do we care about that? Because we know it's really rare to monopolize your TAM and get 80 or 90%. So we're just like, we want to see a really big TAM because if you can get 20 or 30 percent of the market, you have competitors, whatever, that's fine.
I think you're gonna have to change for some markets as an investor and say I'm just looking for TAMs where you can get most of the TAM. And maybe it's a smaller TAM, maybe it's a 180 million TAM, but you can get 150 million of it and that's a good business. I don't know.
Jason Jacobs: But how can that play out? Because one of the things that I'm, worried about as a founder, it's kind of two related things. One is, are moats less durable? And two, is platform risk greater?
Rob May: I would say yes to both because If you assume these AI coding [00:32:00] tools get better, where, we're probably still 8 years away from this, but it should be able to copy a lot of functionality. And platforms that you have where the primary defensibility of a platform has been the fact that it has a lot of interconnections to things, or that it was hard to write and took a long time that's going to go away as a point of defensibility.
Zapier would be a company that is pretty hard to unseat because they've connected to so many things, like it would take you years to catch up and write all that. But at some point when AI becomes good enough, you're like, I could spawn 10, 000 new connectors to other apps in a matter of days.
You could do some of that. And so I think in those businesses. You have to have other relationships, special access via business development relationships. I really like businesses that have a physical component. VCs have been against things that have hardware or sensors or anything.
But if you're dealing with components that are in a physical location and that matters, that's harder to replicate because you can't just [00:33:00] spin those up and go replicate, go replace somebody's device automatically. So I think those businesses are going to become. I don't know if they'll become more valuable.
They'll become more defensible. Yeah, I think pure application software businesses are going to be tough to build moats going forward. And I think it's going to take a while for investors to make that mind shift because we've loved SAS, we've loved it. And I think the pricing and packaging of SAS is changing.
I think the competitive nature of SAS markets is going to change. And so I'm not sure that your vertical SAS is going to be where you want to be in the future for most markets. There's always pockets of money, to be made in some places, but just the overwhelming amount of sass that we've done, I think, is not going to make sense in an agentic world.
Agents are going to be able to be more horizontal and do more things for you. And so it's interesting to see how that market breaks out. But I think investors are going to have to go through a big mindset shift. And I do think moats are going to get harder to build. And even when you're building technical businesses, you're going to have to rely more on traditional moats, like geography, regulation, economies of scale of physical [00:34:00] goods and things like that.
And I think brand can start to become a good moat even stronger than it used to be.
Jason Jacobs: What do you think the implications will be on how these businesses get funded in terms of both, ~the, ~the amount of capital they'll need, the type of capital they'll need? And I guess also, as I said, both, but I'll add a third thing. You know, t t time to exit and, and size of, of exits. If, if exit is even a concept that will carry on.
Rob May: Yeah, it just might be going away, So let me give you two, let me give you a contrarian opinion on M& A and then let me give you an opinion that, Actually, I'll give this one first on LPs. LPs in private markets have created their own problem that they complain about, right?
They all say there's no liquidity, there's no IPOs, so we can't continue to invest and create stuff. And it's okay, why are there no IPOs? Because the same LPs that are complaining about the lack of IPOs give tens of billions of dollars to these late stage growth funds to write [00:35:00] billion dollar checks into companies to keep them private long.
And I'm like, The LPs are funding funds to keep companies private longer and then complaining that people don't go public. And you're like okay, pick which one do you want? So that's a problem that I think they don't realize. And then people complain about the M& A markets, right?
And a lot of people have pointed to the FTC and Lendacon. The Biden administration policy, I don't buy it and I don't buy it for a couple of reasons. Number one, I was on the phone. I actually just tweeted this the other day or I posted on LinkedIn. ~I'm not really doing Twitter anymore.~
~I'm very anti~
Jason Jacobs: What's Twitter? I don't even know, I mean, oh, oh, X, you mean X,
Rob May: yeah, I'm anti X. I deleted all my posts. I don't post anymore. I'm on blue sky and on LinkedIn. But, yeah. But I posted on LinkedIn, so I was talking to the head of a fortune 200 company, corp dev the other day and I was like, what's the M and a market like for you guys now?
Are you just, there's all these venture companies that are struggling and he's man, we thought we would be aggressive. There's a lot of stuff we want to buy. We'd love to do this. But these late stage VCs that got into crazy valuations don't want to sell and they'd rather prop up the companies.
And so he blames VCs more than. [00:36:00] any kind of government policy or, FTC stuff, for the lack of M and a activity. And then, people complain that con was looking at breaking up Facebook, breaking up Amazon. So one of the things people don't understand about startup markets is there's only a handful of really big buyers, right?
Google, Facebook, Apple meta, Nvidia now, whatever. And as a result, there are a lot of startup markets where there's 6 startups and in most of those markets, somebody gets bought by one of those companies if they're playing in similar spaces and the rest either don't have an option, maybe there's a second option for one of those companies to go to one of the other, like if you're doing something social, maybe one of those companies will buy you, as a secondary option, but there's not enough buyers.
And so I think this is very contrarian and nobody else in tech I've heard say this, but if you broke up meta, what would you have? You would have two or three, same if you broke up Amazon, instead of having one giant, trillion dollar [00:37:00] acquirer, you'd break it up into three or four acquirers that still had a couple hundred billion dollars of market cap each.
So I think you would have more acquisitions. If you broke up the top tech companies, they would have to compete across more vectors with each other. They'd have to be more acquisitive. They'd still be big enough to do good size acquisitions. Billion dollar acquisitions. More startups would get acquired.
I think it would be good, but people are very linear and first order in their thinking, and I don't think they're thinking about this the right way.
Jason Jacobs: Rob, I know we're, at least starting to run up on time, but one topic that we didn't cover yet that I think is important is, when it comes to starting companies, you know, I've been a technology founder for a long time. I've never learned to write a lick of code. And traditionally, if you look at like YC for example, they have this whole, you know, hacker and hustler, and so, as I said at age 48, I'm not really a hacker or a hustler.
I'm kind of like a, like a lazy chairman. And so I guess my question is, As, as this all plays out, when it comes [00:38:00] to starting new technology companies, how do these tools change the ideal founding skill sets? And, and also, what's your recommendation for, you know, people that maybe historically would be the classic non technical founder that are looking for the technical counterpart to build anything?
Does that change in this new world or do you still need to partner?
Rob May: Great question. That definitely the hacker and hustler thing was big for the fogey founder group. I think it depends a little bit on what you're doing, right? I think if you're innovating at the application layer, you probably don't need much technical expertise now.
I think more and more that's going to be abstracted away. But I think if you're going to do something that's technically different, like I'll give an example. I think you're seeing more innovation at the chip and hardware layer, the infrastructure layer. And I think that's gonna be a trend.
And I think, I think there aren't enough electrical engineers. So I think that job category is gonna crap. So people working at the very bottom of the stack, and people working at the, are gonna need to be more technical and less anything else. And [00:39:00] people work at the very top of the stack are gonna need to be the opposite.
There's big change coming at the bottom of the stack, right? Just to give you like one example, in 1971, this guy basically was looking at all the mathematical equations of circuit elements, resistors, capacitors, inductors, and was like, you know what? You could rewrite these equations one more way, and it would look like this.
And he called it a memristor, and nobody could figure out how to build one until 2005. But a memristor works a little more like a neuron. Which is interesting. People haven't been able to make them into circuits at production level scale, but we're getting closer. But as memristors get out, that's a bottom level thing that's gonna change and ripple through the whole ecosystem.
And probably if you're building a memristor company or a chip based on memristors or attempting that, you probably just need technical people and not really a hustler yet. But again if you're building application layer stuff, I don't know that you're gonna need a lot of technical help.
I think you're gonna be able to do more and more. On your own. Just with tools that are available. So it's really gonna change. I think founding teams are gonna be [00:40:00] smaller. You're gonna scale with more people. We saw this at my last startup. We were doing work when we were, 12 people that probably would have taken 30 people even five years ago.
Jason Jacobs: So, let me make this personal. I don't know what I'm building yet, but I don't think it's venture backed. I want to keep the team small. I want to stay flexible. I don't want to raise any outside capital if I can avoid it. And, I'm not doing heavily tactical stuff. It's a lot of content, but I think what's interesting is that as the content kicks out with a newsletter and a podcast and that library starts to grow, I think there's inferences that can be made across episodes and posts that might be not obvious and interesting where AI might play a role. I also think that there could be tentacles externally trying to keep up with how quickly everything's changing, what companies are emerging, et cetera, that could kind of make us more efficient at the journey itself.
And then there's just fun ideas, you know, like the show idea or like others that, you know, the AI could be really helpful in bringing to life. You know, you said that for this kind of You know, top of the stack kind of stuff. Would this feel, which this feels like that, you know, that, that potentially you don't need any help, but I have to [00:41:00] tell you, trying to make sense of all this stuff, not only is it hard for this old guy to, to make sense of, but it just isn't really what gives me energy and it isn't really my superpower.
So given all of that, what advice would you have for me specifically?
Rob May: You gotta hire your kids to run this business with you because they probably know this stuff better than you, right? No. Yeah, I think. So this is unfortunate, but these AI tools are going to benefit older guys like us who, have a lot of perspective and can synthesize a lot of stuff, right?
It's, because these tools are at a point in their technology history development where they can do task related stuff, but they don't have experience and expertise. They don't have context that's deeper. Yeah, so I think doing something like that, I definitely think you can figure this out if you want to stay lean.
But also you can
Jason Jacobs: I'm going to try in the short term regardless because I think it's healthy for the, for the mission and to, to feel the pain. Right. But it's more of the longer term. In a way, here's a weird analogy. It almost feels like, [00:42:00] you know, hiring an agency to do a rebrand without like a marketing person internally to take point on it when you're not a marketer, you know, like that's kind of what I feel like using all these different tools myself.
Rob May: yeah, it's, there's a really good book. If you want to read one of the best books on AI, it's, it's called Prediction Machines. And what they talk a lot about, and this gets to your point, is, so there's this idea of economic complements, right? Things that go together economically. And when the price of one goes down, the price of the other one actually may go up because you'll demand more of it, right?
If the price of milk goes down and you can buy more milk, people are going to buy more cookies. And if there's not an increase in supply of cookies, maybe the price of cookies goes up, right? That's a simple version. If AI is going to bring down operations, execution, task orientation, predictions, all these kinds of things.
What goes up in value? What's the economic compliment or the work compliment to those and it's judgment, right? It's experience. It's the ability to, when you have those situations, you've run a company, right? And somebody junior comes [00:43:00] to you and they're like, we did this analysis and X, Y, Z, and you're like, I understand why you think that and why you think the analysis says that, but you miss this thing over here.
That's not obvious. That makes me actually not want to do, what you said. And, yeah. ~And so sometimes that's because the machines are only good as the data that you feed them. And when you're doing these big things, like running a small, a specific data set it's not going to tell you everything.~
And people will misuse it. So to give you an example, there's a pretty famous paper out there, on radiology. And it highlights this term that they call correlated uncertainty.
So what happened was, they took all these images and they had a radiologist diagnose it and the algorithm. And the radiologist could override the algorithm or not and do whatever. And what they found was you had a bunch of chances where they both agreed, great, those are easy.
You had, issues where they disagreed, and then in the issues where they disagreed, they also had a confidence level. What the radiologist did was, When they disagreed with the finding of the algorithm, but the radiologist had a high confidence level, they overrode it. And when they disagreed, [00:44:00] but had a low confidence level, but the algorithm had a low confidence level too, they listened to the algorithm.
And it should have been the opposite. So basically what the way the radiologist used is, if I think I know what I'm doing, I ignore the algorithm, and if I don't, I listen to the algorithm. What they actually should have happened and it turned out not to be right not to be The optimal solution is if you think you're right, but the algorithm disagrees with You should second guess yourself because the algorithm is probably right so basically the way it should work is anytime the algorithm has a high confidence level You should listen to it And when it doesn't then you should use your human judgment because you're probably picking up on other cues outside of just the x ray that a human can pick up on.
And so I think you're going to see that problem play out in a lot of markets where you're going to be able to AI is going to do some stuff and people are going to listen to it when they shouldn't. Yeah, it's going to be a very interesting world, man, to see where all this goes.
Jason Jacobs: Well, gosh, this has been so helpful and informative. I feel like I've got, more questions than answers, although we did cover a lot of ground. Anything I didn't [00:45:00] ask that I should have or any parting words for listeners?
Rob May: I think this is great. I'd love to come back on at some point as your Going through your journey and learning more and talk about some other stuff. I'm always looking for new investments, so send.
Jason Jacobs: Oh, no, that was my last question, which is just for anyone listening, that, you know, that's inspired by your work. Who do you want to hear from or, how can listeners be helpful to you?
Rob May: Yeah, and just Rob May on LinkedIn, or rob at halfcourt. vc. Seed and series A deals in, infrastructure and compute. Love what we call services as software, which is automating human tasks. And
anything with a heavy AI bent. We're a pretty technical firm. So we don't do a lot of consumer and all that, but, we'd love to take a look at all those kinds of deals and, be helpful, any way that we can, we want to be a good part of the ecosystem,
Jason Jacobs: Well, thanks for coming on, Rob. This was awesome. Episode one in the books and we'll keep this dialogue going for sure.
Rob May: all right. Thanks for having me, Jason.
Thank you for tuning into the next next. If you enjoyed it, you can subscribe from your favorite podcast player. [00:46:00] In addition to the podcast, which typically publishes weekly, there's also a weekly newsletter on sub stack at the next, next dot sub stack. com. That's essentially for weekly accountability of the ground.
I'm covering areas I'm tackling next and where I could use some help as well. And it's a great area to foster discussion and dialogue around the topics that we cover on the show. Thanks for tuning in. See you next week.