How Gamma pulled off their AI pivot | Jon Noronha (Co-founder and CPO of Gamma)

Brett: Maybe a way to frame part of the discussion is for Gamma, if you go back to six months before the company was started all the way through today, how would you define the distinct phases of product-market fit for the company?

Jon: Well, let's start with the none era of product-market fit. So we started the company in 2020. This was obviously peak pandemic, and it was this period when we saw all these companies, like say Zoom and Loom and Slack, really blowing up because people were taking a new approach to work. And there was this incredible why-now moment of people realizing we're not all in the same room, we're not using the same tools, and so that was the era in which Gamma was born. The early hypothesis was not about AI. It was all about remote work driving this transformation. And so we did a lot of prototyping, a lot of beta concepts.

Brett: Did you start working on the company without a product in mind?

Jon: No. Well, we had a problem in mind, I should say, and we had a target. The target was always PowerPoint. We knew we wanted to reinvent presentations, because first of all, it had a massive TAM. That was something I took from my previous experiences. I wanted to work on something that had a huge TAM if you were successful with a product that nobody liked. And so PowerPoint checked those two boxes really well. It also felt well suited to our team's skills. It was really about front-end design, creativity, productivity. I think there was a great product to be built here. There's like a billion people that use PowerPoint and Google Slides every month, and they're not happy with it. So that was always the driving force. We thought we had a "why now." And so we dove in and we said, "All right, let's prototype, let's explore, let's see what we can do here."

We built a product that did okay, I would say. We had some traction, but if we're talking 2020 to 2022-ish, we had basically no product out, just prototypes. '22, we launched on Product Hunt. We get our first thousands of users, maybe we're in the hundreds of monthly actives. We have slightly linear growth, but I would say we're in this era of weak to middling product-market fit, definitely not red-hot. At this point, runway starts dwindling. We're starting to figure out where do we go from here? How do we drive more of this? And this is when the first green shoots of AI are starting to show up. We're seeing Stable Diffusion as a model out there. We're seeing GPT-3, not even ChatGPT yet. And we realized that Gamma's core problem is this blank-page problem. People come in wanting to make a presentation or whatever, but they're faced with an empty white screen and told, "Go start from scratch."

Brett: But why wasn't that a problem for Google Slides or PowerPoint?

Jon: Well, that's the ironic thing we realized is that in fact it is, but there was no alternative. And that actually is what drove product-market fit for us is we always tell ourselves, "We're 10X better than PowerPoint, 10X better than Google Slides." But until we actually solved that blank-page problem for people, we weren't really. That was actually the foundational problem with those products is they start you in a blank page and you cobble something together out of templates. And there just was no better way technologically to overcome this before. I think the closest was to give people good templates. This is Canva's strategy. It's also what Notion's done really well in its own separate domain. But we got to now use this ultra-powerful kryptonite tool to solve the problem, which was generative AI. And so we plugged that in to solve our onboarding challenges.

Brett: How easy was it to identify that that was the bottleneck? Was it blindingly obvious?

Jon: Once we had already launched a product and had it in market, it was blindingly obvious that our activation rate was low. The conversion rate of someone signing up for our product from a landing page and actually getting to a point of seeing value was appallingly low. It was like we're talking 5% or 10% or something. So as soon as we went to say, "Well, what would get this number to be 20%, 30%, 50%?" it was like, well, this blank page was staring us in the face is the obvious problem. And so at that point, yes, it was obvious.

So we designed a whole launch around let's overcome that blank page. It was our Hail Mary moment as a startup. We were down to maybe a year of runway. We were a team of 12, but we knew this AI thing was going to be big. Nobody knew how big, and we wanted to bet on use AI to overcome the blank-page problem. And so we did a big launch around this. We put all of our resources into it. We launched it in March 2023, and that did take off. That really did scratch an itch and got us through our first stage of product-market fit.

And the best definition I've heard of product-market fit is you stop pushing the rock up the hill and you start chasing the rock down the hill. That was absolutely what it felt like. Especially, I would say, that first year of 2023, there was such a gap between the capabilities of our product and our team and a demand of what people wanted out of it. We were just racing to keep up. We're talking servers going down every other week, at one point, for even days at a time, a support volume we couldn't keep up with. We weren't even monetizing and people were emailing us begging to pay for the product. That's as red-hot as it gets with product-market fit. And we spent the whole next couple of years just trying to keep up with that. Now, we're in an era where I think we've finally caught up with that prosumer product-market fit. We're finally able to service the demand and scale it and all of that.

But now, we've entered this new stage of product-market fit challenge, which is in the last year or so, B2B demand appeared. Suddenly, all these companies were already starting to use our product internally, and also, we saw this phenomenon of the AI mandate, where the CEO says top-down, "We need more AI. Let's drive this as a priority." They survey their employees and they say, "What are the top things you wish you could do with AI?" Well, everyone needs just a chat tool for sure. The engineers all need a tool like a Cursor or Claude Code or whatever. And it turns out the third use case that's bubbling up over and over is presentations. People are wasting all this time just formatting their PowerPoints and Google Slides. And so we hit this new source of product-market fit that we were once again woefully unprepared for. We had no sales team, no compliance, no anything. And so for the last year, we've been building up that B2B motion basically from scratch and getting it in place.

Brett: Before you solved the onboarding problem, was the product for people that got through onboarding insanely high NPS and retentive?

Jon: No, it was not.

Brett: So why did fixing onboarding fix the overall customer sat of the end-to-end product?

Jon: We thought we were solving an onboarding problem, which was overcome the blank page, but it turns out we were discovering product-market fit, because the real job to be done for the customer is also solve the blank page. And so when we solved what we thought was our problem, which was solve the blank page for our onboarding, we actually solved their problem, which was solve the blank page for me to avoid all this time making a presentation. And it turned out that was actually the thing that had product-market fit.

Brett: Can you explain more about that?

Jon: Fundamentally, if you just think about the last time you had to make a presentation and just how that immediately felt to you like, "Oh shit, big, high-stakes presentation to the boss, the client," whatever it is, your mind immediately leaps to all the work you have to do. "I have to come up with a visual design. I have to make a template. I have to structure the story and pull together all the key ideas. I have to lay out each individual slide and make sure it doesn't overflow off the page, and now, I need to fill the space so I got to find some stupid clip art and fill it up." You're suddenly looking at 10 hours of work to make anything good minimum, and every time, you're almost restarting, not to mention the fragility of your team starts to edit it and mess with it.

And so what we did was we just eliminated maybe eight or nine of those 10 hours right off the board. And when you can give that concrete time savings, the value is just so immediate, concrete, and people leapt on it.

Brett: When you were working on onboarding, did you think it could solve the entire basically how satisfying this product is?

Jon: Not originally. That's not why we prioritized it, but we have a strong dogfooding culture. We're using our own product every single day. And I remember this moment, maybe a month or so into this three-month sprint, to get this big new AI version of our product where we were all just sitting around at lunch playing with a mobile version of our product and we started just making presentations about anything we could think of, like different breeds of cats, why does Subway smell so good? All these kinds of questions. And we realized that it was fun in the same way that a social media app is fun or a game that you play is fun. It was this new kind of dopamine reward that you got from generating these things. And suddenly, I got this inkling of, whoa, this could actually be a whole product value proposition. There's this aspect of the slot machine of what am I going to get?

Brett: Did you think about building the product for a specific subset of presentation creators and were trying to work backwards from that and bound the jobs to be done? Or did you think very generically about a generalizable tool that anybody to do anything could use?

Jon: We bet early on and pretty consistently on building a horizontal product rather than a vertical one that was focused on a specific persona. And it was actually a source of constant tension in fundraising in a lot of our early debates, because I think there's a playbook out there, which is you need to pick a very narrow vertical, satisfy that customer type and then build outwards from there. We bet against that because we looked at this productivity space and all of the most inspiring companies that we saw were not verticalized in this way. They were actually quite horizontal. So your Notions, your Slacks, your Looms, it was actually difficult to pin down where their early fit had come from.

You could do it in some cases. Maybe Slack was early-stage startups or whatever it is. But these companies were fairly broad in who they targeted. And even if they had a persona, it wasn't a single narrow type; it was a broad one. So we actually settled on a broad one, which was external presenting, not like the internal company all-hands, but somebody in a role like sales or marketing or consulting who is ultimately trying to get business from somebody else.

Brett: Why?

Jon: Because there was natural virality to it. It would spread between organizations. And there was natural willingness to pay, because the presentation was ultimately tied to some kind of revenue-generating activity for that business. So we've always leaned towards that direction while still being good for other use cases, like internal presentations, but we chose not to narrow any further. And I think that served us well, at least up to this point, because we've proven that with that broad product, you can serve a huge horizontal user base.

Brett: Was that a very easy decision, or did you spend real time working through it?

Jon: For me, it actually always felt easy. It was obvious just looking at the leading tools in this space that that was the path that they had taken. A tool like PowerPoint is not a vertical solution, it's horizontal. But it was a source of constant friction, because I would say every investor asked us about it and challenged us on it, which then caused us to question ourselves over and over.

Brett: Did you think about and do you currently think about building the product as if you are building a consumer product company?

Jon: I think overall, we've leaned consumer, especially early on. Maybe just give some concrete ways in which we have looked more consumer-like. The composition of our team early on, our original team we hit product-market fit was 12 people, of which zero were in any kind of GTM role, so we had no sales or marketing. Four out of those 12 were UX designers. So that's actually a pretty insane ratio of one-third. But it's the kind of ratio you do when you're a consumer company who's betting on user experience and product-led growth being your main driver.

I think another example is that we've relied heavily on techniques like A/B testing coming from my Optimizely heritage to iterate our way towards everything from pricing to AI models to user experience, also, generally a much more consumer-skewed technique. I think one thing that informed this orientation was actually the very first exercise we did when we started the company, before anything else, before we built any product, was we did a hundred user interviews in which to ask people in different walks of life, different verticals basically, "Tell me the last time you made a presentation. What was that like?" We had them walk through what they made, how they made it, what went well and what didn't go well.

And the really extraordinary thing from doing that exercise was hearing that of those hundred people, whether they were a consultant or a teacher or a doctor or a tech employee, they all said pretty much the exact same stuff. It wasn't that the doctor had different presentation needs than the consultant by and large. There were some specifics. It was almost all the same few problems, which is actually this blank-page problem, "I don't know where to start," this feeling of judgment, "I feel like people are judging me based on how my slides look, not what I'm trying to say," and then this huge tax of formatting, "I spend all this time just moving boxes around to align them and make them look right." A lot of the quotes we heard, probably the single most common quote we heard from all of these people, no matter what their job was "I spent 90% of my time on formatting and 10% of my time on content." And so unlocking that, flipping that ratio became the skeleton key that felt totally horizontal, not vertical to one specific user type.

Brett: Talk about how you went from that to this prototyping motion.

Jon: Yeah. So we started with those user problems. We boiled it down to three top ones, which was basically formatting, structure, and content. And then we dove into prototyping. All of our early team were either designer or engineer types, so everybody was capable of building things. Keep in mind, this is pre-AI coding, so we still had to do it all the old-fashioned way, hand-rolled by hand. But we built a ton of prototypes. We would build multiple every week. We would take them back to some of those same users. We really relied heavily on trying these things ourselves. Early on, these products were all so bad you could barely even ask a customer to make a presentation in them, because presentations are naturally high stakes. People use them for real work. The product wasn't ready for real work. And so we made presentations ourselves over and over every day.

Brett: Do you think the fact that you had weak product-market fit ended up being a huge lucky break in that if the company had much stronger product-market fit, it would be harder to rebuild it when these new AI primitives were invented?

Jon: Yes. I absolutely feel that way. I think in many ways, we got extraordinarily lucky by having built so much of the non-AI primitives infrastructure that gave us something to build upon, but not being wedded to any of it and knowing we had to make a lot of changes at the moment AI came along. So we could throw out the parts that weren't serving us, but we weren't building from zero. We weren't building a pure GPT wrapper, so to speak, because we actually had years of technology we'd built up that AI could layer on top of.

And it's an interesting challenge now because as we grow, product-market fit is not guaranteed. You don't get to just keep it because you had it before. We're in this incredibly competitive space, but now, we have the challenge that we have been so successful. We have more than a hundred million users who've signed up for our product, but AI has not slowed down. If anything, it's sped up. And so we're having to now think very carefully about how we chuck out parts of our product we have today and embrace new ones while managing that larger legacy base.

Brett: How do you do that? What does that sound like?

Jon: It means constantly questioning some of our core design principles, a lot of bets that we made early on. Gamma's early success came in this period 2023 when LLMs weren't very good yet. And so a lot of what made our product great was guardrails we put around these dumb ... They weren't even agents yet, dumb prompts to protect the AI from itself. Now, what we're finding actually is that all those guardrails we put in place are holding these much more intelligent AIs back. And so we're now in a mode not of even adding functionality, but trying to throw things out, loosen the requirements, and let AI do more overall.

Brett: Knowing what you know now about the path of the company and the way that the product was reinvented, do you think everything unfolded how it needed to? Or could it have been done faster or differently?

Jon: Well, I think you can always look back with hindsight and say, "We should have skipped that route of the idea maze and gone straight to the good idea," but that's not really possible. Yeah, you don't have the time machine. But I guess your question is more like if you were just to execute again with laser-like focus, could you have gotten there more quickly and more directly? I think for sure, yes. We took some wrong turns, and those wrong turns are still embedded in the DNA of our product in many ways. A lot of the work of scaling through this next stage of growth is figuring out which of those wrong turns to undo. Also, which of those were not really wrong turns, they were just too early and seeds of something greater. So there are some aspects of our product which never hit product-market fit in that early stage, but I think still could actually.

To give a simple example, one of the early bets that Gamma took was interactivity. This idea that a presentation didn't have to be just a linear narrative where you click through slide by slide, but instead you could have live interactive elements on the screen. Early on, that turned out not to resonate with our first product-market fit because people really just wanted to interoperate with the world they're used to of make traditional PowerPoints. And so they actually caused all this pain for us of trying to figure out how to bridge this interactivity with this legacy format that was not interactive. And we're still facing that pain because we're in this crossing the chasm period where we're trying to win over people that want familiarity.

But as we grow our distribution and become a platform in our own right where people spend time, now, we can actually reintroduce those elements and then become actually part of the stickiness of why someone wants to stay in a tool like Gamma versus exporting to somewhere else. And so many of these concepts have a proper time, but I think looking back as a founder, a lot of the mistakes come from misjudging timing versus misjudging ideas.

Brett: Say more about that.

Jon: I think as a first-time founder, a lot of my naiveté came from looking at some product and saying, "I can make it better." And that might be true, but I think there's a couple false assumptions that live inside that declaration of better. The first one is that just because I think it's better doesn't mean the user think it's better. But the other one is thinking about the user in isolation versus the system in which they operate. And with productivity, you're always collaborating with peers, you have work guidelines and templates you have to follow. And so a lot of our best ideas were what if you could be free of all of that? But people are not free of all that. And so you have to meet them where they are. You have to bridge that familiarity gap, and only once you've bridged it and started to bring people into your new system can you now start to layer on what you really thought was magical about the vision.

And I think you really see this with companies that have been in this game much longer than we have at Gamma. And thinking again about these companies that inspire us like a Slack or a Notion or a Canva, early on, they all took different spins on the thing they were replacing like docs and email, and some parts of it were unfamiliar and confusing, and they had to layer in those bridges to make sense. But now that these companies have all been around for, let's say, 10 years, they're getting to really dig into those original parts of their vision and bring them to life.

Brett: If you think about the things that you would have done differently knowing everything that you know, there's a huge chunk of it which is at that time, it was just unknowable. And in that case, the way that you work through the idea maze is required or you couldn't unearth it. The other is from time to time, there are things actually that were totally knowable at that time, but because of your own frame of reference or all sorts of other things, you went in the wrong direction. And I'm curious if in that second bucket, if there are any specific things that come to mind that were important, or did it tended to be more it was unknowable, we had to go down the hallway, realize the door was closed, then go to another door?

Jon: The biggest one that comes to mind for me is the obvious and hindsight realization that we had started by trying to pioneer this new creative format that was somewhere in between presentation, document, and webpage. So it was mobile-responsive, content would adjust and reflow. And that was what created a lot of our early propulsion, which was that this format actually worked really nicely for LLMs.

I think the blind spot for us though was drinking our own Kool-Aid that people always want to operate in our new format versus the obvious realization that no, everybody is still exporting this thing to PowerPoint early on because that's where their team lives. And so we did build a PowerPoint export, but probably a year or two late. I think we should have listened to the signals earlier on of people just telling us this is what they wanted. But we were so excited about our own format and wanting to build our own proprietary frame of reference. This also I think came from perhaps trying to build for durability in moat before building for adoption, and I think you actually have to go in the opposite order. You have to first win adoption, which is kind of a grueling fight with someone who is changing their workflow and only then build some of that stickiness around it.

Brett: One of the things you're touching on, and I think it's really interesting in a consumer or prosumer form factor of software, is you have a lot of people that say the way to build products is just to build the thing that you would want, you're the customer. There's another version of it which is no, you have to get really close to customers and understand what they care about or not necessarily have them design the product, but have them articulate their problems. And then there's maybe some place in between, but there's definitely a customer-centric worldview, and then maybe there's the Jobsian worldview, which is like you have to instantiate the thing you've always wanted. And it feels like in the journey of Gamma, there's been this connection between these two because there's this dogfooding culture, there's this the whole company was started around wanting a thing that didn't exist. What else could you reflect on those two maybe worldviews?

Jon: I think that's really well put, and I feel like we've constantly been sailing between these worldviews. I don't think I would ally us with either one purely. I think you're right that we have had this Jobsian mindset of build a thing that didn't already exist, people don't quite know what they're asking for. And I think in the early stages, you see that in the way that we started from problems. Really make sure we're aligned with the customer on the problems that they're stating, but show them a solution that is not necessarily what they wanted.

But as I also alluded to, there were times where we drank our own Kool-Aid too much and we believed we had built the 10X better solution when really the customer's telling us, "Nah, I actually want this other thing instead." And so we've gone through periods of having to then re-anchor that against exactly what customers are telling us. We're really experiencing this now as we take our consumer-y product and go into B2B where in many ways, there's already product-market fit. People are using this thing undercover, they're pulling it into their companies, but at the same time, we're hitting brick walls on certain aspects of the product that need to change because businesses have very different needs than consumers. And so I would say as a leader, it's been on my mind how to cultivate both mindsets, especially when it comes to our design team, which needs to both contain this "research, stay close to the customer" mindset, but also this maverick, creative, "build a thing they haven't asked for yet" mindset.

Brett: So what's your best cut at that? Or if you are trying to indoctrinate that into the people designing products, building products, engineering products, how does it manifest itself?

Jon: The quote that I heard about this that I have always really liked is that "You diagnose with data and solve with design." And so what I take that to mean is, first of all, when it comes to quantitative data, we try to be a very data-driven organization. We really stick to metrics and analytics and A/B testing. But there's also qualitative data. We try to always stay very close to what the customer is telling us and let us all marinate in that. But they're not going to give us solutions. The way to actually solve solutions is to get out ahead of those problems and find the unlock, whether it's a UX designer making a mock-up, but even more often now, it's an engineer actually prototyping something with AI that sometimes crystallizes or unlocks three or four different problems we've heard into one new solution and then bounce back and forth and diagnose, did this actually solve the problem? Let's shove those in people's faces and let them tell us where it holds up and where it doesn't.

Brett: How did you think about the episodic nature of the product?

Jon: I would say it is the biggest structural challenge with this category. You didn't mention another structural challenge, which is that competitors are all bundled into larger suites.

Brett: And I was going to say, yeah, there's generally, I would imagine, low willingness to pay.

Jon: Yeah, yeah. And I think these two factors are why this category of presentations has been a graveyard of startups. PowerPoint came out in 1987, and in the last 39 years, basically nobody else has managed to unseat them. The companies that have come closest, if you probably say Google Slides and maybe Keynote, were not even really serious competitors. They were ports of the same product to different platforms as part of a different larger suite. And so this was always our biggest doubt and fear about entering the category and probably our, again, naiveté that we even did it at all. But I think what we felt was that these dynamics created stagnation in the category, and that stagnation was what made room for a disruptor to come with the right "why now."

We were actually wrong about the "why now." We though it was remote work in COVID. We got lucky that a better "why now" came along just in time, which was AI and really embraced that. And AI is such a sea change that I do think it is just this anti-incumbent technology across the board. I mean, that's why we have a SaaS apocalypse now, is it's questioning every software company's moat.

And so that's what's given us, let's say, the boldness to be a challenger in this category despite those challenges. I won't claim that we've solved all those problems. We are iterating through them. Although I do think probably what's been most valuable to us is to have the mindset of being okay with a broad horizontal, somewhat episodic user base because we believe we can find pockets of deep value inside of that. And it's not an either/or. It's not build the horizontal product or build the verticalized solution with depth for a different customer base. It's a bit more T-shaped. We've started with that wide horizontal base, which is still growing just given this tidal wave of change coming from AI. But we think we can build pillars of depth within that, and we're already seeing signs of that, particularly in our B2B motion, our API. And I think what's different about those is they represent repeated workflows on teams rather than episodic usage by individuals. And so the challenge our team is trying to navigate is how to keep growing that prosumer user base while also building those pillars of depth.

Brett: And so what's your best crack at that now?

Jon: I think our best crack at it where we've already seen such clear value is when you have a larger company, so not these small prosumer use cases, but companies of thousands of people or even Fortune 10s or Fortune 100s with tens of thousands of people where they have some team at their company that has a repeated presentation use case. The most obvious one you could think of is a sales team, where you've got people just pitching the product over and over and over again in a fairly repeated and templated way. You could think of it as almost following a recipe over and over again.

And where we have now rolled out a B2B offering as well as an API with connectors into common tools. So a simple example to think about is that sales team uses, let's say, Salesforce as their CRM system of record, and anytime an opportunity gets created or advances stages or has a meeting, we can automatically create a presentation based on that, building on all the content that's already happened and hand-deliver it to the sales rep to use in the next meeting. And companies love that because frankly, they don't trust their own sales reps to make good content. They had a very fragile workflow before of copy-pasting templates. If we can really streamline that workflow and turn it to a repeated one, we're seeing that there's a huge opportunity in these companies to do it.

Brett: How does that fit with the nature of this product, which is one that just has infinite roadmap? You talk to 25 different customers, certain things bubble up, but you could build infinitely because of the form factor of the product. And so when you think about either the different chapters of the company's life, how did you think about prioritizing?

Jon: Well, one thing that has both helped us and hurt us in prioritization is being an exceptionally lean company. We have generally had a pretty small team. For a while, we were at over two million ARR per employee, which actually felt like too much. It was almost comically outsized in terms of the amount of work we had to do versus the size of our team. But it forced a lot of clarity and prioritization, and I think it's often helped us do something that is hard for product builders to do, which is build less instead of build more. And particularly in this world of rapidly advancing AI models, it's actually a really healthy instinct because the less you build as product surface area, the more that the AI can actually fill in in actually the white space. The blank page was the problem we started out trying to solve. Now, it's the thing we're trying to give to these models to let them rip.

And so we are trying to engineer an organization that doesn't build a thousand laundry list features, instead builds strong reusable building blocks that AI can work on top of to provide a lot of the remaining range.

Brett: What have been some of your reflections on building a product for a use case where the current status quo is people creating very bad presentations?

Jon: Well, first of all, it's what made us bet on this category. I think there are ... I'll contrast it to maybe the world of spreadsheets. So Excel is such a dominant tool for spreadsheets, and I love Excel. I think it's a great product. I would not want to compete with Excel. Even though I do think there are all these interesting ways you could compete with Excel in the sense that code is coming along, the challenge is that it's so beloved by the people who use it that I would not have seen, personally as a founder, the daylight of where I could add that 10X value. PowerPoint felt so ripe to attack because it doesn't have the same feeling. The output undermines people's confidence and the process of making it is so taxing that it felt like there was white space here, even despite the structural challenges of the industry.

Brett: And why do you think that is? At PowerPoint, you have, I think, a bunch of smart people that are trying to do great work. I think Google Slides has talented technologists and product builders. What's the working theory as to why it is the way that it was?

Jon: I think the same structural challenges that make it hard for a startup to succeed here also give an incumbent no reason to put focus into it. If you think about who's buying the Google Suite or the Microsoft Office Suite, nobody's doing it by comparing the feature set of PowerPoint versus Google Slides. Instead, they're comparing Outlook versus Gmail, and maybe after that, Word versus Google Slides. And PowerPoint and Google Slides are afterthoughts in their product roadmap. And it shows up in their prioritization of how they allocate engineering talent, product talent, design talent. By the way, it also increasingly shows up in how they allocate GPUs, which are the new currency. The thing that I've heard secondhand is that when Google is thinking about where to put its AI resources, of course, they're going to throw those at Search, which is its flagship core product rather than something like Google Slides, which it can count on being sold for free whether it does well or not.

So this is the interesting opportunity for us. They've left this huge UX gap for us that we can go after, but there are still these structural business challenges that we have to solve to make it a successful business around that.

Brett: You could argue Google Sheets is an example of this. It's not a flagship product. It has tremendous adoption, but it feels like Google Sheets is doing a much better job of delivering on the promise than Google Slides is, and just like Excel, to your point, is much more beloved than PowerPoint. But a lot of times, the suite is not sold because of Excel. In some ways, I guess Excel has become Kleenex, and that's an interesting part of also the challenge when something just becomes a standard.

Jon: Yeah, yeah. You don't want to diverge from that standard too much. That's absolutely right. I think probably the biggest difference between the Excel workflow and the PowerPoint workflow is creativity and taste. There's less objective definition of what good looks like, and there's much more room for the user and also the brand to express themselves. That's what makes it a really fun problem for us to work on. It's also what makes it a much more challenging AI problem to tackle.

For example, if you imagine how to write evals for a product like Excel versus a product like PowerPoint, I think it's pretty easy to imagine how to write the Excel ones. Are the formulas correct? If we give it a bunch of inputs and outputs, does it give us the right number at the end? It's actually a fairly straightforward problem. If you now imagine how to do that for PowerPoint, there are these mechanical failures. Did the text get clipped at the bottom of the slide or whatever? But once you get past those, it's a way squishier problem to objectively iterate.

And these products are so horizontal. You've got millions of people using them in different ways. At the scale of PowerPoint and Google Slides, you have hundreds of millions of people using them. And so you start to actually need this difficult quality of taste and creativity. And while I think there is great engineering talent working on those products, I don't think they have treated them as a place where they put their best designers and people who have that taste. It's a hard thing to exercise inside of a big tech company like that.

Brett: So how have you solved the evals problem, or how do you think about it at a philosophical level?

Jon: Well, I won't claim that we have solved it because it's a really tricky problem for us. I think we view it as more of a data problem, and I think we are trying to view this less as an objective function of "Is this good or bad?" and more of a problem of extracting and identifying people's taste and preferences, which are not the same for everyone.

And so there's a few pieces of this. One of them is collecting the data. I think this is also a nice potential advantage we have over some of these FAANG competitors is that we have this large prosumer user base that we can actually learn from and use that to actually build a better enterprise experience as well. And it's in the cloud, so we actually have visibility into what people are making, what they're doing, how they're prompting. So we're using that extract taste and see what works. We're running A/B tests on different models to see what clicks. We're also able to combine different models from different providers in a way that some of our competitors structurally aren't. They're tied into their own house models. And so we get to directly test. Particularly in the world of image models, they all have a different feel to them and a different texture, and we get to provide all of those for different use cases and see where they shine.

So for us, it's not about the one right answer of evals, although we do have evals for a lot of common cases. It's about actually using data to identify user preferences and eventually build models that tune to those.

Brett: How do you think about short-term versus long-term opportunity? One way to think about that is that gap creates the space for the entire company. The other is that ideally you don't want to spend a tremendous amount of time solving problems that you think in 90 days or six months will be completely solved with some simple model integration.

Jon: There's this scaffolding approach where we constantly have to build out temporary supports and then take them away again. It's a very strange approach to product building, and we're even having to reorient our own mindset about how we build internally for that. The classic design mindset is make a quick prototype and throw it away. The classic engineering mindset is build durable systems that can last for 10 years. We find ourselves in this in between where we have to build systems that last for six to 12 months at a scale of a hundred million people plus using them, but we also have to know that we're going to say goodbye to them and throw them away and build something new because model progress is so fast.

Brett: So how do you actually go about doing that?

Jon: I'm not going to claim that we've solved the answer to that, but I think a lot of it comes from engineering things with the mindset of this will get thrown away later. And so it needs to scale by usage, but it doesn't need to scale by time, meaning that we don't need to plan for every future eventuality. We often have to flag things and build such that they can be turned off again later. I would say that's a core part of building, so feature flagging as just an engineering technique. But I think it gets deeper culturally, and it's a cultural lesson that we're still trying to learn. It also jives with this whole move towards AI coding where, gosh, the shame we're going to throw this all away in six months, but at the same time, we can build it four times faster. So we should be willing to actually make more temporary things that we throw out later.

Brett: What's the inverse of this? What are the things that you've decided have to be built for durability if a huge portion of what you build is going to be much more transitory?

Jon: Well, we have certain obligations to our customers. Right? We need to keep their data private and secure, especially as we move into enterprise. We also have to keep serving these presentations, which are not always a one-shot episodic thing. They're often these long-term living artifacts. And so we have to serve them basically forever unchanged. We can't just go mess with someone's content after the fact. And so the things that are about how we store, render, and present things need to be extremely durable and sticky, but the mechanics of how we generate them, how we edit them need to get thrown out even faster than we're throwing them out now.

But there's one more challenge with that, which is worth calling out, which is even as we move to AI doing more of the work, we are also in this world where users are trying to learn a complicated new tool to do a high stakes part of their job. And so we can't just add 10 new buttons every month as new features come out. We still have to actually prune and simplify the user experience and maintain people's expectations of how things will work.

Brett: And so is most of that that there's no particular way in which you've solved that other than it's a consideration and people have to use good judgment to navigate it?

Jon: Yes, but I think there's also maybe a design principle here, which goes back to our prioritization, which is err on the side of building fewer knobs and controls because all of those become things that you need to maintain and honor over time. And they may end up becoming irrelevant as AI capabilities advance. So I think we're now leaning towards this model of have fewer simpler building blocks, but then more open-ended customization on top, including even just writing custom HTML, CSS, going with the grain of what models can already do.

Brett: What are some of the things that you've had to cobble together to get the product experience right along this similar path that we're going down that you're pretty sure will be solved in the future and thus you're creating this temporary solve?

Jon: A lot of what we've had to cobble together along the way have been guardrails that get an LLM to produce consistently good visual output. So to give some simple examples of that, if you are a company with an established brand, you always want to generate in consistent fonts, colors, sizes, also even more basic considerations that matter to everyone, and AI should never write so much text that it overflows off the end of the slide and you can't read it anymore. Text probably shouldn't get too small. Early on, the guardrails we took to that were hard structural constraints of our format. For example, we didn't even let you customize the fonts on an individual block of text. We enforced them from the outside and said, "LLM, you just write text and we will style it for you." Same applies for layout where we wouldn't actually have the AI hardcode X, Y positions of elements. We would have it give more semantic ideas like a three-column layout, and we would actually render that for the AI.

Those things have served us really well for both agents and humans. Even for humans, they remove a lot of the fussing that goes into making a presentation, and that's our core value proposition is taking away the fussy formatting so you can focus on content. But frankly, we're finding now that it's actually hamstringing us. By building all these guardrails in about formatting and design, we're actually limiting what AI can do compared to just letting it loose with full creativity. And so a lot of the hard problems we're digging into now is how do we actually give the AI more of a creative ceiling and, in some cases, kick away these guardrails that were previously holding it up while still maintaining that floor and the editing experience that is simple and focused?

Brett: What has been your important learnings from a monetization standpoint?

Jon: Well, I think the biggest meta learning has been that monetizing AI products is a moving target, and there's no pricing model we could have had starting in 2023 up to now in 2026 that would've worked at every period of our growth or into the future. So we've actually had to have a monetization model that is adaptive with changing trends. So just to maybe call out a few of those key adaptations that we had to face early on. The first discovery was actually a great one, which is AI products are easier to monetize than traditional SaaS products. The willingness to pay is just much higher, and there's a natural customer understanding that these things cost money to run, and there's this natural equation people have of their own hours saved into dollars they pay for a software product. And so when we can save someone eight hours in a month and charge them $20, for many people, it's a no-brainer, and they'll happily do it.

Brett: And they're able to compute that rationally you found?

Jon: Yes. At least they're willing to give it a try. I think they're willing to say, "Yeah, 20 bucks." I mean, I think those numbers are usually far enough apart of what do I pay for eight hours of work versus $20, that many, many people are willing to just try it out.

But then there's a churn component on the other end of that, which is this AI tourism phenomenon. People are very willing to try a new AI product, but just because they've swiped their credit card the first time doesn't mean they're actually committed. And so we've had to learn how to almost filter within our revenue base and figure out who are those sticky users and how do we keep nurturing and growing them?

As you said, presentations are a naturally episodic use case, which means we can't expect on our prosumer base to have more than 100% net revenue retention. We know we're going to lose a lot of these people. We have to be okay with losing those people. It would be a mistake though to block them entirely from subscribing or, for example, limit ourselves to only annual plans and lose those monthly users. First of all, because they're still a great revenue source, even if it's not always recurring. Second of all, because within that larger base, there are some who will really stick and see the value. And third, because there's still this viral loop that we kick off from these episodic users who, by the way, do come back even if it's episodic. It still recurs on a slower rate.

And so for all those reasons, we want to keep nurturing this wide base of people who know and love Gamma and refer to their friends and use it sometimes, but we also need to look within that for the stickier user base. And sometimes, there are different personas or ICPs within these. The type of user who will be the sticky recurring user using us every day of the week is not always the same as the person who's coming episodically, but there are more of those episodic people. And so if you just look at raw numbers, you can get confused about where to focus your time.

Brett: And so how have you actually gone about this? How does that map to the way you've chosen to monetize the product?

Jon: Part of it is by monetizing our product through different tiers, and pricing naturally reveals preference. And so we have at the low end, a $10-a-month plan, which is relatively episodic. It's our entry-level floor. It's also more popular internationally in developing countries who don't have the money for very expensive AI products. It's not actually a loss leader because we do make money off of it, but we never expect it to be the sticky retentive product. But then we have just kept discovering new ways to add more tiers of value on top of that. So we have a Pro plan that's $25 a month on top of that, which is surprisingly popular even at that higher price point and naturally stickier. It's for bigger prosumers, people in more developed countries who have more money to spend.

And then we've now added an Ultra plan that's at the $100-a-month price point. So that's people who have much higher volume of usage, much higher expectations. They're willing to use the fanciest AI models to get the best output and so we can monetize them very differently. And then there's also this new axis of individual versus team where we've built up Team and Business plans where we're now spreading across the entire company, and the retention of those looks obviously much better than the individual plans.

Brett: Is the transition from a consumer, prosumer company to one selling businesses feel very incremental, or it feels just the gulf between those two things is very significant?

Jon: I wanted it to be incremental. The way I thought we would tackle this was that we would inch our way upwards from individual to SMB to mid-market to enterprise to really big enterprise. That felt like the classic way to do it and also the incremental sustainable way to learn as you go. The pace of this AI industry has just upturned all of our plans in that respect for two reasons. One of them is just competition. We can't afford to leisurely work our way up the market because things move so quickly. But then the second one is that the demand is just there. I never would've thought that people in Fortune 25 companies want a product like Gamma right now, because we started with these prosumer-y roots. Early on, maybe a year ago, they would come to us and I would just laugh and say, "There's no way we can serve a company like this. We shouldn't even have the conversation." But what amazed us was even when we tried to turn away these customers, they kept coming back to us.

Brett: The other way of arguing it is that the way a consumer uses a product is similar in that a business or a large enterprise is just a collection of individuals who are all trying to create material to talk to customers or clients or this or that. Why did it seem like it's such a gulf?

Jon: Because I'd been through a verticalized journey at my previous company, Optimizely, where we also started as an SMB product with $10, $20-a-month plans and over a period of, in Optimizely's case, several years, transitioned to an enterprise product with customers paying millions of dollars a year. And I was very involved at every stage of that journey. I saw what went well, but I also saw all the ways in which it was really hard. It's really hard to build up a sales team that can sell to enterprises. It's really hard to market to those people and retool your motion. It's really hard to build a product that works for those. It turns out even when the individual user need is the same, teams have such different needs, and there's this whole world of compliance, privacy, IT approvals and everything that you have to solve. And I saw how small our team was. I think we had maybe, I don't know, 15 or 20 engineers at the time we started feeling this demand.

And so it's not that we didn't want those kinds of customers, it was just knowing the reality of what it would take to serve them was daunting. And I know how easily you can get pulled into these deals where you start making commitments to them. Our philosophy was we didn't want to give up our prosumer product and only bet on enterprise. We wanted to layer enterprise on top of this prosumer business that was already healthy, which meant that you have to divide your focus to some degree. We have since grown the team a lot. Obviously, we've raised money. We've probably more than tripled the size of our engineering team. We've built up go-to-market. We are now serving these companies. But to answer your original question, it was not incremental. It's been a rapid sprint to actually be able to serve companies like that.

Brett: Switching gears just a little bit, I think there's so many people that talk about we're in a time that distribution matters more than product. Even in the early days when you had an inkling of this product, but it wasn't quite right, you had a lot of distribution, you had a bunch of people trying the product. What have you figured out about distribution?

Jon: I think I'll probably disagree with the premise here of I still think product matters much more than distribution. It is true though that increasingly as AI products feed on their own training data, that distribution is how you make your product better. And so I think dynamics may change where people without distribution can't get their products to be good enough. But I still think it's a mistake to conclude from that that, oh, these are all distribution problems and we just need to have a better brand and more marketing dollars and everything. All of that absolutely matters, but I think ultimately, product is the driver, and great products can still overcome a distribution gap.

Brett: But isn't the early lesson of the story that the product wasn't right, but you also had tremendous distribution, thousands and thousands and thousands and thousands of people coming to the product?

Jon: No, I think that until we had nailed the product, we actually had pretty weak distribution. We had hundreds of people coming to the product every day, and every time, we would do a big launch or a big marketing moment, that would spike to a thousand and fall right back down. The inflection point came when we truly solved virality in the product by introducing AI. And at that point, we started getting distribution for free, because people would make something great, they would share it with someone else, those people would try it out with a very low friction loop and tell their friends. And at that point, it didn't even matter how big that initial seed of distribution was. What mattered was the slope of the curve, and I think that slope is all product.

Brett: You spent most of your career building products before LLMs existed and codegen and everything that we have today, and now, you're building a company that's completely predicated on this enabling technology. How would you describe the difference in what it means to be excellent at building products today versus at every other point in your career?

Jon: I think in many ways, LLMs have heightened what it always took to build great products and widen the gap between the teams that are really good at them and not. So I think actually, the core trait that mattered before and still matters now is agility, and by agility, I specifically mean being extremely close to customers about what they actually want, shipping extremely quickly, faster than you're comfortable with, and then being willing to throw away or change your core product mechanisms based on what you learn.

The insane thing about LLMs is that agility used to mean that you would do all those things over, let's say, a six to 12-month cycle, and now, it's become possible, keyword being possible, to do them over a day to week cycle instead. Almost no company can really stomach though the sort of g-forces involved in moving at that pace, and especially larger companies. There's just too many stakeholders involved, too many legacy commitments and responsibilities that I think it's just physically impossible to move at that kind of pace. Not just impossible, but irresponsible to move at that pace.

Brett: And the bottleneck being humans.

Jon: Not only humans. I would say human decision-making is one, but also just the accumulated cruft of backwards compatibility, contracts, legal responsibilities. I think even an org run by a thousand AI agents would actually hit this friction. It's just that they're all a few months old, and so we haven't seen it yet. And so we're in this world where your ability to tighten that cycle time on all of those pieces is the key trait that will drive success. And companies with different profiles have to adapt in different ways. For us as a roughly a hundred person company, we already feel too slow. We already feel like we have to be faster. We see how a company of two or three people can just fly because they have none of this accumulated cruft. And I can only imagine what it's like to operate 100,000 person company in these conditions.

Brett: How do you figure out how fast you should actually be moving?

Jon: The answer's probably always faster at this point. I think ultimately, this goes back to diagnose with data, treat with design. So the way to know if you're not moving fast enough is when you start seeing signals in the data that you're losing even an ounce of product-market fit, so looking at activity metrics, looking at segmentation of your users and how much they're driving your product, being unflinching in listening to why people churn off your product and where they're going and what they're saying. We're trying to really sharpen all of those systems so that all of us are getting basically slapped in the face by data every day, and we can use that as the ultimate signal that says ... It's not about be fast or slow. It's be fast in this specific area. This is an area that needs drastic change, this is an area that can make gradual incremental change, this is an area that's crushing it and just keep going the way it was.

Brett: You gave the qualitative definition of your experience of product-market fit while rolling up the hill versus down the hill. If you had to turn that into quantitative metrics or numbers that actually express that qualitative feeling, what are the most important numbers that express it?

Jon: The most obvious one to me in terms of when I felt that things qualitatively changed was actually just new user signups per day. Obviously, a vanity metric, it's not the one that most drives the business, but it was the one that drove the qualitative feel of product-market fit. And just to give some concrete numbers, before we launched our sort of AI product and it was more esoteric and hard to use, we were in the hundreds of signups a day. As soon as we launched this AI product, it jumped into the low thousands. We're talking 1,000 or 2,000 a day. But the reason we knew product-market fit was happening was that the numbers started inflecting upwards. So it started inflecting to be, very quickly in a matter of a month or two, 10,000 signups a day, 20,000, 30,000.

Brett: Which is the opposite. Normally, you launch something and you get from a thousand-

Jon: Normally, you launch, it goes up, and it spikes back down.

Brett: Yeah, exactly.

Jon: So the fact that it went up, I mean, it reminded me of reading about Facebook in the early days, like a true viral coefficient. Was it work? And we were still doing no-paid marketing at this time. So the only explanation for this was word-of-mouth growth, basically, different kinds of word of mouth. But that 10,000 kept climbing upwards, 10,000, 20,000, 30,000, to the point where in our first year we crossed 100,000 signups a day. And I just tried to step back and visualize 100,000 people. That's like a stadium. I mean, it's really actually unreal when people were signing up for this product that we hadn't even really marketed and still had major weak spots at the time. This was maybe three years ago. And that number has continued to trend upward, not exponentially forever. We'd have the whole human population at some point if it had kept trending upwards, but it has settled in now well above 100,000 signups every day. And that just keeps on going. It amazes me that there's this many people out there. And of course now, we think a lot more about signup quality and different markets and different job functions and all of that, but that was the clearest sign that we had gone through some kind of phase shift.

Brett: Did you then also think a lot about retention and usages as key PMF metrics, or no, it's really-

Jon: The true definition of PMF is that you're so overwhelmed that you can't think of anything else. And that's what we were for the first year probably. We couldn't even think about are they using or retaining? We're just trying to keep all the people coming through the door.

Brett: Back to what we were talking about a second ago, what are some of the things you've had to unlearn or have been hard for you to unlearn in building products in this era versus the rest of your career?

Jon: One of the hardest things is that we, as a company, value craftsmanship as one of our core values, building things really well and really carefully and really thoughtfully. And I think also part of that is we valued a culture of ... I won't say consensus decision-making, but involving a lot of people, having a lot of opinions, and letting those all shape the final output. And a lot of that care from our team is what drove our early success.

So I wouldn't say we've tried to unlearn it, but the pace at which the market is moving and the need to change the product has made us adapt that. Our idea of what craft looks like. Craft can no longer look like going into your little studio and tinkering for several weeks before coming out with something to show for it, and craft also doesn't mean that all of us can feel the same ownership of the product through consensus. We actually have to take these blind leaps in the dark often, put something out there, see if it works. We also have to sunset things a lot more quickly. And so I think figuring out how to square high pace and urgency of innovation with craft and care has been one of the trickiest cultural adaptations to navigate along the way.

Brett: So even if it's not solved, what is the definition of craft in the context of your company today?

Jon: I think it means that we still want to be our own best user of the product. We still use our own love of the product as the ultimate north star, but we have now tried to pair that with a much higher pace and a much higher level of urgency at getting things out there and much less sentimentality of keeping the things that we have in there. So maybe you could say that it's pairing craft down to its essential, which is care without possessiveness.

Brett: When you zoom all the way out and you think about all the different things that you figured out about building an app-layer AI company today, what are the most important things that you're reasonably sure are globally correct?

Jon: A few that come to mind as really obvious ones. Multi-model orchestration, so you can't tie yourselves to a single model and just wrap that. You have to add value by orchestrating multiple pieces together. Simple example for us at Gamma is that we orchestrate both text and image models, but even in those categories, we have probably three-ish major providers of each one and use different models for different tasks, in particular, optimizing for cost, latency and reliability. Building both evals on top and also a data pipeline to actually gather real user preference signals so that you can iterate on the quality I would say is really essential. On pricing, segmenting users into different types and finding ways to serve those different categories. I don't know how obvious this one is, but going international has been huge for us. The AI app-layer opportunity is so much broader than just a narrow segment of people like, say, in the United States and the people right next to you. I think especially for a horizontal prosumer product like ours, it's just a no-brainer, and our company is 10 times bigger than it would've been if we hadn't done that.

I don't know if going horizontal versus vertical in general is a clear-cut answer. I hope what we've shown is that horizontal can work, but certainly, vertical can work, too. I do think it's important to pick, and it's very hard to waffle between those things, and you make pretty different choices if you do. For us as a horizontal company, betting on UX as a differentiator has served us really well, and nurturing a team culture and talent around that has been really good.

Brett: What about high-order bit insights? And you hit on many of these around specifically building a modern prosumer business. Are there hard-earned insights that for other people that are just about to begin a prosumer business, you can say, "Be careful of this, keep this in mind?"

Jon: Rethink pricing. Don't take legacy SaaS pricing as a given in this new world. I think you'll need to test your way to pricing yourself. And this is a live one. I don't think our lessons will even carry forward in six months because the landscape is changing so quickly. Differentiation remains really important. I think you need a really crisp story of why someone should bet on you as a brand and you as a product because of an extremely concrete way in which you're different. I think for us, if we were to boil it down into one word, it was speed. You should choose Gamma over PowerPoint because we are so much faster at getting you to the polished outcomes. Obviously, we care about other things we've tried to layer things on, but we tried to really orient ourselves around that. And then bring that into the onboarding. That's maybe the crucial one for a horizontal prosumer thing is you have to take your one value prop and show it in the first two minutes of when someone signs up. If you can't do that, you're toast.

And probably, the single biggest lesson that you could take from Gamma was focusing on onboarding actually unlocked the real product's value prop. So maybe to go back to your question of what would we do differently? I would've just focused on onboarding. For my next company, I would say just nail onboarding over and over. If you do nothing else, nail onboarding, because if you can nail the first five minutes, you've nailed the value prop of the rest of the product.

Brett: Maybe just to wrap up, something you didn't talk that much about is what was the emotional and psychological journey of the company? So maybe if you go back across your path to PMF and the different chapters, was it despair and anxiety? Was it, it was always pretty light and people were just creative and trying things? What did it feel like as a founding team through the beginning of COVID all the way up until now you're having a hundred thousand people try the product on a daily basis?

Jon: Ironically, I would say the happiest and easiest time was when we had zero product-market fit because zero PMF equals no responsibilities, complete creative freedom.

Brett: But it didn't feel anxiety-producing that we have nothing?

Jon: Eventually, it got very anxiety-producing once that runway counter started ticking down. But early on, I would say our first year or two, we'd raised enough money that we weren't immediately worried about will this hit? We knew we had some time, and so we were in an exploration period, and that was so fun. I got the best sleep of my life during, I would say, those periods of the company. Then probably, the worst period of all was the time when that clock started ticking down and I knew we didn't have PMF, and the music was going to stop. That was maximum dread, despair, anxiety that I felt.

But once we hit product-market fit, it wasn't all sunshine and roses. Now, there was a whole new set of challenges. Oh my gosh, every day, there's a new problem to be solved. There's so much to keep up with. I love this metaphor of pushing the rock up the hill and chasing the rock down the hill because I think it illustrates that neither activity is particularly fun - pushing the rock and chasing the rock. And the frustrating thing about startups is you don't get even a week at the very top to just sit and enjoy the view. You immediately go into chasing mode, and I feel like we've been in chasing mode ever since. Not to say it isn't fun. There's many wonderful parts of the experience. I wouldn't trade it. It's my dream. But don't expect to ever get a break, and if you're a founder, don't ever kid yourself of like, "Oh, there's just this one stretch, but after that, it's going to be so much easier."

Brett: Or this one exec or this.

Jon: Yeah. Yeah. If I hire this one exec, if I close this one candidate, if we get this one deal, it's all going to get easier. It never gets easy.

Brett: So then why is this of interest to you? You go from one version of pain to another version of pain, and as you said, by definition, the entirety of the company will be this.

Jon: I'm a recent new parent and I'm observing all the ways in which having a child is like having a company. And my feeling about both in these last few years is that there is so much pain and just physical exhaustion, but also so much joy and pride that you feel along the way. This feeling that you have built something that didn't exist before and the joy of getting to create something magical with your colleagues every day is so high that it is worth the pain and challenges that you endure along the way.

Brett: In reflection, did you find the lack of intensity and pressure and angst that existed in the early chapter of the company's life, was that a feature, or because you had runway, it created too much space for too much tinkering and slowness? Do you reflect on that at all?

Jon: I do reflect on it, but I don't think it's a clear-cut answer. On the one hand, I think the nature of a creative product that is so design-oriented is that you can't build it in a pressure cooker. This goes back to this larger debate going on around 996 and whether everybody needs to be working 80-hour weeks, and I think there are certain types of company that are well-served by a 996-type culture because it's just a grind where there's very concrete, legible work to do. There's way more work than people and whoever gets through it fastest wins.

But creative products that are oriented around creating that special experience that are often about restraint and doing less I don't think are very well served by this mindset. They're well served by creating space for exploring multiple different paths in parallel, dogfooding them yourself, coming to your own conviction with time. And it's been funny observing the Linear founder and the Corgi founder arguing about this on Twitter because they represent these two extremes of company really well. We, I think, are more of those design-y tool where we needed that space. At the same time, I'm also cognizant that Gamma really only succeeded because of luck. Our original idea didn't work. A world-changing trend came along just when we were at our low point. And I mean, kudos to us for jumping on that trend, but still, if the wind hadn't come at just the right time, our boat would've sank. And I do wonder if we'd had more urgency and speed, if we would've gotten more maybe shots on goal.

Brett: What is the takeaway from that, the point that you just made? Is it that there is more luck in outlier companies?

Jon: Oh, yeah. I mean, there's a huge amount of luck in outlier companies. I have no doubt about that based on my own experience and from talking to other founders as well. But it's not only luck. I mean, I think that the boat metaphor is always valuable of if the wind doesn't come, you're going to sink. But at the same time, dealing with that wind and getting your sails in the right place and zigging and zagging at the right time takes a huge amount of skill and execution. But in terms of what does it mean for company building, my synthesis of these ideas and the one I'm trying to apply at Gamma is you need to both give yourself room to be creative and try a lot of ideas, but you need to do it faster and you need to be extremely decisive and bold on, when you have one, throwing the entire company's energy at it.

Brett: Good place to end. Thank you so much for the time and the conversation.

Jon: Thank you for having me. It's great to be here.

Brett: I really appreciate it. Yeah. Thanks.