The Pattern Breaker After the Model—What Becomes Scarce When Cognition Is Abundant?

Julia Nimchinski:

Thanks, and now we have a very special guest. Welcome, Mike Maples, founding partner at Floodgate and author of Pattern Breakers. Mike is an 8-time Forbes Midas List investor, pioneer of seed stage investing, and he made his early bets on Twitter. Lyft, Twitch, Okta, Rappy, and Implied Intuition. Mike. So excited to have you with us. How have you been?

Mike Maples, Jr.:

I’ve been great, thanks for having me, it’s a pleasure to be here.

Julia Nimchinski:

Amazing. We’ll be sending your book to all of the speakers and partners and our community, so, super excited to dig into some of the concepts today and, just merge it with some new form factors of AI.

Mike Maples, Jr.:

Yeah, let’s do it!

Julia Nimchinski:

Awesome. Let’s start with a simple question. Could we just set the stage with your why? Why did you write this book?

Mike Maples, Jr.:

Yeah, so, I’d been doing this for about 10 years, and I noticed that about 85% of my profits had come from pivots. And so, Twitter had started as a podcasting company called Odeo, Lyft had started as ZimRide, Chegg had started as a classified site at campuses. Twitch started as JustinTV, and so I was like, okay. what’s going on here? You know, and some of the folks in your audience this might resonate with, some companies seem to just have greatness thrust upon them. You can’t really figure out why they succeeded, they didn’t follow any best practices.

You know, Twitter had to fail well all the time, you know, a lot of things seemed kind of screwed up. And so, I started to study this topic a lot more, because conversely, I would find that teams that did all the right things, you know, did the business model canvas, were disciplined about analyzing customer needs, all that stuff, didn’t succeed. And so, Pattern Breakers was my attempt to kind of answer that question, and what I… what I really discovered is that startups win by avoiding comparison games. And so, you know, if a customer can compare you to something else, why would they pick a startup that’s 80% likely to go out of business?

You want to show up with something radically different, not something better. And so, pattern breaking became kind of this metaphor for what does a breakthrough startup really need to do to show up in this world in the right way.

Julia Nimchinski:

Love it. Mike, I’d like to, just right off the bat, transition to B2B. Because when we think about, obviously, pattern breaking, and your methodology. it just… B2C comes just to mind very naturally. When we think about B2B, I’m thinking about all of the MarTech map, sales tech map, now AI, rev ops, name it, maps. You invested in Okta. Could you just share your pattern-breaking lens on the B2B spectrum?

Mike Maples, Jr.:

Yeah, so, you know, to me, pattern breaking starts with an inflection. So, something new has to happen in this world. that allows the entrepreneur to play an unfair game against the present. You know, so business is never a fair fight. The only question is who gets to fight unfair. And so the inflection for Okta was that SaaS adoption was accelerating, and we believed that enterprises were going to use SaaS and have lots of apps. This wasn’t obvious. 2010, people were arguing about whether SaaS could be trusted for enterprise, but we were starting to see that.

So Todd and Freddie had this idea of identity management across lots of cloud apps. And part of what I liked about their approach was that they were very vendor-independent. So they were saying that, you know, that the customer wasn’t going to want to trust Microsoft for identity management, because Microsoft has an agenda to sell you Microsoft products, and the same is true of Salesforce, the same is true of any big company. And so they argued that you needed a credibly neutral broker, to manage identity management, and that sounded right to me. And so, fortunately, that one worked out, right?

But that was… really what they were doing to break the pattern was to, offer some radically new set of capabilities to early adopters of the cloud. And, you know, that’s a good B2B example, actually, Julia. You know, like, the earlier ones I mentioned were consumer… people resonate with that more easily, but B2B applies just as much.

Julia Nimchinski:

100%. And then, just connecting with the, you know, the latest, and I’m actually curious, how do you define it? Inflection inside, non-consensus inside, or a movement serving the eye. How do you see it, and, and… Curious your thoughts on, you know, like, I see it basically as a network effect, just spurred with Palantir. Curious your thoughts.

Mike Maples, Jr.:

Yeah, so the thing that I’ve really been spending a lot of time on, and if people are interested, I’m working on this paper, so, like, you know, maybe as a follow-up, if somebody sends me an email that just says durability paper, I can send them the copy so that they can criticize it. But basically, before AI, if you harnessed it, let’s take Okta, for example. You know, before AI, you find an inflection, SaaS is taking off, you create an identity management solution. You enjoyed an overwhelming advantage if you were first a product-market fit, because customers are shouting your praises from the rooftops, you’re accelerating your penetration of the market, you know, first to product-market fit almost always achieved category dominance.

But in the AI world, you know, you see, like, windsurf. And, you know, Anthropic decides to pull the model from Windsurf, and now they have to sell on fire sale. And so. I started to see over and over again, you know, Anthropic introduced those legal plug-ins, and people were saying, well, what should Harvey do about that? So, more and more, I saw these companies get product-market fit, and what I realized was that, in the AI era, product-market fit is not enough anymore. It’s still necessary, but it’s not sufficient. You also have to build durability. And so, you have to kind of ask, alright. as soon as I achieve product-market fit, I’m not only a leader, I’m also painted a target on myself.

And people can immediately try to execute and copy what I did, and gain knowledge very rapidly about what it takes to come up with a substitute. So that’s, I’d say that’s the main change, is, the AI has a lot of inflections and a lot of empowering new capabilities, but it’s harder to have a fundamental insight that nobody else has, and to bring something to the market that’s completely unique. And then once that’s discovered, to maintain that uniqueness. And so that’s kind of what I’ve been really focusing on these days.

Julia Nimchinski:

Definitely, and what’s your view on… on the general dominant fear now, you know, of your alpha, your IP, leaking to the Frontier Labs? Is it a movement? Is it a real inflection? How do you see it?

Mike Maples, Jr.:

Well, I guess the way I look at it is, AI changes the economics of knowledge. So, it used to be that, you could have knowledge about customers, or you could, you know, you hired employees with certain skills, or, you know, like, if you, if you had an advantage in the market, you had a bunch of proprietary knowledge and insights. that was hard for the rest of the market to replicate. Well, now these models, in many cases, make all pre-existing knowledge effectively free. And so, what happens, I think, is, the relative value of new knowledge is more scarce and more valuable, but it’s also less persistently durable.

And so, like, now when I look at these companies, I ask, first question is related to product-market fit. Are we getting bigger more efficiently? You know, are we selling more customers faster? But then I ask, are we getting better with each new customer? You know, does each new win allow us to better serve the next customer explicitly? And if so, how? But even that isn’t making us stronger. So, making us stronger means I’m doing something that blocks competitors with each successive win, from being able to replicate my advantage. So I’m creating a… not just a benefit, but a barrier.

And then I think durability is, how long can my advantage survive change and attempts by competitors to come after me, or new inflections on the horizon? And then durability is also, how does time buy me a window to come up with new advantages, so that I can, you know, be ready when the competitors try to nullify my existing advantages? But, like, I kind of like this framework of, you know, bigger, better, stronger, durable, because it kind of, it focuses the mind, right? If you’re in a big company or you’re in a startup, you can say, okay, every time I deliver value to a customer, am I just getting bigger? am I also getting better?

Am I getting stronger somehow? If so, how? And am I getting more durable somehow? If so, how? And so that’s, to me, that’s kind of the big change in AI, is that, you can’t just take out market risk at the beginning, you also have to take out durability risk earlier.

Julia Nimchinski:

like, they’re declaring that AGI is here. I’m curious, AI as a technology, do you see it capable to pattern break?

Mike Maples, Jr.:

Generally speaking, I think that it can accelerate it, but I still think that the human needs to be involved. So, if you buy the premise that, at least today’s world, that AI and the LLMs fundamentally commoditize all pre-existing knowledge. Well, breakthroughs, by definition, haven’t happened yet. They haven’t been discovered, and so you can’t exclusively rely on pre-existing knowledge for an undiscovered new knowledge. And so, what I really think that the AI can do is Help a creative person, come up with more conjectures, come up with better ways to error correct their ideas, and come up with more, sort of.

Combinations to, consider. I think Finn must have just, finn, my Chief of staff, Zoom bombed us, per se, there.

Julia Nimchinski:

It’s okay. Yeah, I’d like to continue on this, on this, thread of, you know, living in the future. There’s always this fear dominating today that, you know, you’re missing out. You’re missing out as a company, as an individual, as a startup. You’re so much behind, especially every time in your open acts. Curious, what are you… what are you seeing the most innovative companies, startups, AI-native startups are doing differently compared to everyone else today?

Mike Maples, Jr.:

Yeah, that’s a really good question. So, I’d say that the number one thing that I see happening… so, and my colleague, Anne Meera Koh, has been studying what she calls these AI-pilled companies, and I think that, you know, the AI native company doesn’t just bolt AI on today’s organization and today’s methodologies and processes, right? So, like, if your… if your org chart and your handoff and your meetings and bottlenecks all stay the same. but are just reduced or automated somehow. I don’t think that’s the big idea. I think that the big idea is, Empowering more and more people to, harness the power of software, and empowering more and more people in the company. to realize all the pre-existing knowledge that exists in the company, you know, to make it available to as many people as efficiently as possible.

And then it kind of comes back to the bigger, better, stronger, durable. you know, you have the models, and then you have the Agentic harness around the models, and when I look at these Agentic harnesses, I’m always asking myself. to what extent are we making progress on those four fronts, right? Not just, not just doing things faster, not just making the models run better, but, building something around it that creates, scarcity that only we can supply.

Julia Nimchinski:

Mike, you get to see so many crazy ideas. Back, you know, beginning with the Airbnb days and the famous story. I’m just curious today, especially when anyone can create anything. How do you view any incoming pitch, any incoming crazy idea, and are there any different ideas these days?

Mike Maples, Jr.:

Yeah, so the main… the main challenge I see with the pitches today is, somebody will come in with a new product that is incredibly empowering and exciting, and I’m like, I can totally see why I would want this, and why lots of people would want this. But I’m like, okay, that comes from the inflection, right, from the new capability that’s a breakthrough. What is the fundamental insight here? Like, why aren’t there going to be 10 just like this in 2 weeks? That’s a harder question to answer. And so, I’ve been… very cautious about investing in companies that don’t have a good answer to that question.

Now sometimes, sometimes you create differentiation just by going fast and being first, and, you know, there’s a path dependence to that. But, generally speaking, I tend to like ideas that, Have some fundamental insight that the rest of the market doesn’t see and won’t see for a while.

Julia Nimchinski:

Just say that as a startup, you just have to force a choice and not a comparison. I’m curious, but being AI native in B2B, It’s… you… not only compete with alike AI-native startups, but also with, you know, big incumbents, enterprises, and all the millions of sales forces and customer success agents and whatnot. How do you do it? How do you actually force that choice?

Mike Maples, Jr.:

Yeah, well, I think that the first… the first important aspect of that is to recognize that it’s important in the first place, right? So, Why is that? So… If you’re a customer, and a startup comes in with an idea. And you, as the customer, have any credible alternative than to work with a startup, you will, right? Because startup’s 80% likely to go out of business. And so, like. The only way I can work with a startup rationally is if they are the only provider of something that I desperately want. And I want it right now.

And it’s like, I’m willing to adopt even an incomplete solution if the downside is somewhat mitigated, and I can tolerate that downside. And so, that’s what I mean by forcing a choice, right? It’s like, I’m showing up, everybody else in the world is selling apples. And I don’t say I’m a 10 times better apple, I say I’m the world’s first banana. And, you know, if you value the advantage of bananas, I’m the only guy in town that’s got them. And so, I think that more startups would benefit from that type of thinking. Because if they can compare you, they can replace you, or they can decide not to use you at all.

If they can’t compare you, then they just have to decide whether they want to solve the problem or not that you solve. One of my favorite recent examples is the Tesla Cybertruck, and when I first saw it, I thought Elon was maybe joking. I was like, is he going to really ship this thing? And, you know. you may like the Cybertruck, you may dislike it, you may think it’s ridiculous, right? But, like. nobody ever, after seeing it, ever says, how’s that compared to Ford F-150? You know, it’s like, it’s self-evident that it doesn’t compare. And so, what Elon is doing is he’s forcing a choice.

He’s saying, live in my future with my Cybertruck or don’t. But, like, it doesn’t make… if we’re asking to compare it to an F-150, we’re having the wrong conversation. And so that’s, that’s what I think more startups need to do, is sort of, paint a bright line between what they do and what the rest of the world does.

Julia Nimchinski:

Love it. Mike, what’s your view on modes, specifically in B2B? Because at the moment you… Invent something, or actually pattern break. It feels like that same moment, you’ll have hundreds of competitors and, you know, fundraising that same week. What do you do?

Mike Maples, Jr.:

Yeah, so I think that it kind of does relate to, Motes happen because… so, like, in systems theory, people talk about flow and stock. And I like to say, product-market fit is a flow. You’re creating value one customer at a time, you’re getting bigger, and you’re growing. And most people, when they think about success at a startup, would say, I need a value hypothesis, I need a growth hypothesis, I need to prove both of those, and then I get product-market fit, and I’m off to the races. I think durability and competitive advantage is like a stock.

So, like, what I like to ask people is, whenever you deliver value to a customer. Like, what is a brick? A brick of strategic, competitive, durable value that potentially gets left behind that you can exploit? So, like, you know, one example might be, you learn how to serve the next customer better. But that’s not competitively durable, that just makes you better relative to yourself. The second type of advantage would be… some type of new knowledge that somehow blocks the competitor, from entering your space. Something that you don’t share broadly with the market, that the market doesn’t yet realize.

It’s an insight that only you have still. And then even better is something that, creates a brick of advantage that, persists over time. And, you know, Julie, you’d mentioned network effects earlier. network effects quite often are an example of that. You know, if I can… if I can create an advantage where every new customer gets an advantage because of every prior customer joined the network, then that tends to be the most durable form of advantage. If anybody’s interested, I think that there’s a good book on this called The Seven Powers by Hamilton Helmer, and he talks about the different competitive powers that can be built.

What I’m finding now is that with each win, with each customer win, we should be asking ourselves. what is left behind that I can convert proactively into one of those seven powers over time.

Julia Nimchinski:

Mike, you talk a lot about earned secrets, And I’m curious your lens on the concept of forward-deployed engineers, and could that be The pledge together are in secret from a company.

Mike Maples, Jr.:

I, I think so. So, The reason I think that it’s important, not just for the company, but for the customer, is that when you think about it. When a forward-deployed engineer is at its best. It’s not just a fancier way of describing professional services. It is a way for the company to co-create new knowledge with their customer. So, like, I first started to see four deployed engineers in my ancient days. I was at Silicon Graphics, and we were… we were trying to help industrial light and magic make dinosaurs look real in Jurassic Park. And so we had four deployed engineers at Industrial Light & Magic.

And we were, you know, nobody knew how to make dinosaurs look real, so we were making it up as we went along. And, you know, ILM understood that. We weren’t just selling them our computers for what they are, we were like, okay, together, we’re just gonna deal with all the corner cases and hard things to make dinosaurs look real. So, what was happening there was we weren’t just professional services, we weren’t just helping them make our products work, we were co-creating new knowledge about how to make massively improved graphics effects in the movies happen. You know, ultimately, Industrial Light & Magic became an important provider for CGI for movies, you know, CGI became baseline in movies over time.

But to me, that’s a good example of a forward-deployed engineer. They’re co-creating knowledge with a customer. In many ways, they’re co-creating the future with the customer. And if you can do that, you start to have something that gets left behind that only you and that customer have, and that the rest of the world doesn’t have yet.

Julia Nimchinski:

Mike, what’s your opinion on AI wrappers? You mentioned that they can definitely become defensible if they understand their process unusually well. But what’s unusually well, and how you change your mind?

Mike Maples, Jr.:

Yeah, I, you know, I always get nervous about, sort of the statements like, okay, it’s just… it’s just a wrapper on a model kind of stuff, right? To me, fundamentally, you know, we need to answer two questions in an AI world. The first is, what can we uniquely do that customers desperately want? And then the second question is. how do we make our early wins into something durable? And so, you know, if you’re building a wraparound model, and it gives you, first mover advantage and a first product market fit advantage. it buys you the time to start to spread out and to start to create durable barriers around that.

I’d say, though, the corollary to that is don’t start to, you know, what is it, when Han Solo in Star Wars says, great shot, kid, don’t get cocky? You know, it’s sort of like when you start to have early success. you don’t want to fall into the trap that Windsurf did. And, you know, you’re still dependent, ultimately, on delivering value based on somebody else’s value, you know, in this case, Anthropic’s value. So, you know, you don’t want to be breathing your own fumes when you’ve succeeded, if that’s how you go about it. You know, I think that people thought Cursor was just a rapper.

But I think Cursor did a lot more proactive things to embed their context and, you know, coding and kind of prior artifacts of what the customer had done into their product proposition. And so I think that it was harder for people to switch off of Cursor than it was for Windsurf, ultimately.

Julia Nimchinski:

Like, your books, warrants quite a bit. just against being trapped in the present. But, in its essence, what’s kind of happening in the enterprise now? When, you know, there is a movement to encode all of your knowledge, all of your decision choices, all the information possible, calls, video, conferences, you name it. Into the model. Wouldn’t it trap us all in that type of present past?

Mike Maples, Jr.:

Hmm. Yeah, I think it’s a good point, Maybe I can give an example just how we do stuff at Floodgate. So. Ann and I have spent years thinking about what makes a breakthrough startup, right? We geek out on that in a way that very few people do. Well, now, you know, we create, different mechanisms to look at a startup, or to think about startups. And we share it with the whole team. And, you know, we record everything possible that we can record, and so… and, you know, we upload books, we have, you know, kind of a floodgate brain, we have our own individual brains.

But, you know, not long ago, I was mentioning in meetings, I like Anne’s ideas around, momentum and durability. I’m thinking about really going deep on that. And Anne was starting to write one of her own essays about the software factory, and she’ll literally start to query our, you know, floodgate brain and say, what is Maples really getting at when he talks about durability? And who’s he talked to about that topic in the last few weeks? And, like, who pushed back on him and who didn’t? And, like. what was some of the magic that came out of those conversations and stuff like that?

How does that agree with or disagree with what I’m writing here? And so what you’re trying to do is, you know, in that case. Not just any venture firm could do that, because not just any venture firm has spent time thinking about some of this stuff, you know? And so what you’re… what you’re trying to do is connect dots that are uniquely yours. With the ability to amplify those connections across your organization, and to make everybody in the organization exposed to your specific knowledge, and then combine it in other clever ways with other people’s knowledge.

Julia Nimchinski:

Love it. Mike, are you worried at all that encoding all of that information into your model? Like, I’d like to actually touch a little bit on the open versus closed model debate. Are you worried that your data is going to flood in to all of the frontier labs, and what models do you use? How does your architecture look like at Floodgate?

Mike Maples, Jr.:

Yeah, I’m a little bit less worried about that, but that’s probably just because structurally who we are, right? We only have, like, 15 people, and we’re, you know, we’re much more interested in moving fast and being attackers and not defenders and stuff like that. And I also think that there are ways that you could construct a harness that increasingly commoditizes the value of the model. By commodity, I don’t mean that it’s not useful, or that it’s not improving, or not more powerful. I think of commodity as something that you have, but something that everybody else has, too.

And so, like, when I think about the harness, then, I’m like, okay, when we create our Agentic Harness. how is it making us better, specifically? How is it making us stronger, specifically? How is it making us more durable, specifically? And how is it empowering each of us at the front lines To harness all of the aggregated specific knowledge that we have inside a floodgate. But, like, you know, our Markdown files aren’t sitting inside of Claude or any… anybody’s LLM. You know, they’re sort of in their own lattice, separate from all that.

Julia Nimchinski:

Mike, we always try with this community to kind of live at least a little bit in the future, and the last summit we just wrapped up was focused on agent-to-agent GTM in B2B. Here is your lens on this? Do you believe in this idea that you can encode trust into algorithms and eliminate some of the functions and make them evolve into… Some… some other form factors.

Mike Maples, Jr.:

You know, I guess I’ve seen enough, interesting, output from Loop Engineering that I would say, you know, sometimes an answer can be too predetermined, but, like. The way I look at it is, certain types of knowledge can be encoded. And, you know, it could be, you know, knowledge about how to write code, which is, I think, part of why coding has worked so well with the agents. But I think that there are probably lots of, when you step back and think about it, there are lots of aspects about go-to-market that could be encoded, and I think that those things, you know, lend themselves really well to the kind of these agent swarms and loops and things like that, and having a harness around, specific knowledge.

So I guess I’m pretty optimistic. I think that the… The stuff that’s gonna be harder is gonna be, what should our strategy be? Or, you know, like, let’s say at Floodgate, I’m like, okay, how do I find the very best entrepreneurs in AI and get them to want to take my money and not the other guy’s money? I don’t think AI is as good at helping me figure that out. It may help me research the topic, but I think I’m gonna have to have the breakthrough kernel of the insight about how to do that. I don’t think the AI is quite there yet.

Julia Nimchinski:

And speaking of Floodgate, you just built, you know, the whole company around this thesis of 500K is the new $5 million. I’m curious, how do you see the evolution of this in the age of AI?

Mike Maples, Jr.:

Yeah, it’s it’s… you know, it’s interesting, right, because some people are now talking about 500K as the new $5 million through the lens of, I don’t need a big team, I can just, with a very small team, create massive output and a big company. And I think that that is one thread of it, but I think that the other thread is that, It’s not so much about… it’s not so much anymore about, doing what’s already been done with less, I think it’s now, empowering people to have access and abilities with code that they never had before, and empowering those people to, force multiply.

So I think that that’s, that’s what I’m most excited about today. And so $500,000 as the new $5 million was a good metaphor for venture capital, you know, in the early days of the lean startup. Now, what I’m more interested in is this idea that, everybody is now part of making the company a software factory. And, you know, the software that you ship is not just what you ship to customers, but it’s the enabling software that propagates throughout your organization. to allow high agency people to force multiply, and to, like, just massively expand your company’s capabilities.

So I’m more interested in companies who aren’t doing the same thing more efficiently, it’s more companies who are, Unlocking the unthinkable now, in ways that would have never been possible or would have been cost prohibitive.

Julia Nimchinski:

Thank you so much for the phenomenal discussion. And as we mentioned, we’re going to be sending your book to all of the speakers and partners. Where should our community go? You have an amazing sub stack. Or should they follow you?

Mike Maples, Jr.:

Yeah, so, yeah, patternbreakers.substack.com, you know, my Twitter handle is, at M2JR, although somebody’s trying to hack it right now, so I’ve been having some difficulties with that. And then, and if anybody’s interested in the paper, mmjr at floodgate.com, and I’ll… I’ll send a… it’s, it’s a draft, right? So go, you know, don’t expect too much, but it’s, it’s a paper on, you know, how to think about durability in the AI age, and like, what, you know, what can companies do to, engineer advantage into their products before they attract competitors, when they achieve product-market fit?

Julia Nimchinski:

We’ll definitely amplify it. Thank you so much again.

Mike Maples, Jr.:

Cool. Thanks, Julia!

Julia Nimchinski:

Thanks. Thanks.