Jon Noronha: How Gamma's big bet on AI paid off

Michael Mignano, a partner at Lightspeed, introduces his conversation with Jon Noronha, co-founder of Gamma - an AI-powered platform that helps users create interactive and engaging presentations, websites, and social media assets in minutes.

The conversation aims to explore Gamma's journey, which began in 2020 but took a significant pivot toward AI in 2023, leading to explosive growth with millions of users worldwide. Michael frames the discussion by highlighting how much of the AI industry's focus has been on the model layer, but is now shifting to the application layer where companies like Gamma are finding their place.

Jon acknowledges the focus shift to applications and notes that while Gamma didn't start as an AI company, it has evolved substantially in that direction.

Timestamp: [00:02-00:43]

Jon shares Gamma's origin story, which began in 2020 after his team left Optimizely following its acquisition. The founding team saw a massive opportunity to disrupt the presentation space, which had seen little fundamental innovation since PowerPoint's creation in 1987.

"PowerPoint's coming up on 40 years old, created in 1987. If you think about how the product has changed over that time, there have been some big changes like color was one they added in the early '90s... but the paradigm has been unchanged."

Jon explains that while there were other presentation tools on the market, they merely represented PowerPoint on different platforms - Google Slides was "PowerPoint on the internet" and Keynote was "PowerPoint for Macs." The Gamma team wanted to fundamentally rethink presentations rather than just creating a better editor.

"We saw this opportunity to rethink presentations from the fundamentals. So don't just try to make a better editor for PowerPoints, try to make a new format for how people present ideas."

Their initial pitch was building "the anti-PowerPoint," and they believed the COVID pandemic created the perfect "why now" moment. With companies like Zoom, Loom, and Slack experiencing explosive growth, the team felt the timing was right to revisit standard workplace tools for an era of remote work.

"We were all about rethinking presentations for an era of remote work, where people are not all in the same room, maybe not even all experiencing the content at the same time."

Their vision was to create a hybrid between presentations and documents - something that could be shared before or after meetings, would capture discussion notes, and would combine rich visuals with substantial content density.

Timestamp: [00:44-03:57]

By 2022, despite having launched a beta product and made a bigger launch on Product Hunt, Gamma found itself in what Jon describes as "the dead zone of partial product-market fit" - a challenging position familiar to many startups.

"We'd launched something that had a number of die-hard fans, but our number of active users was in the thousands, maybe if we're being really strict, maybe it was in the hundreds. We had a couple of passionate people, but it just wasn't there - certainly wasn't there for like a venture-sized outcome."

The company faced a critical juncture. With their burn rate, they were likely to run out of money within a year. The team had to make a difficult choice: pivot the company or make one final big bet on their current presentation-focused path.

Fortunately, timing played a crucial role. In mid-2022, generative AI was finally becoming good enough to be useful. The key moment for Gamma wasn't ChatGPT, but Stable Diffusion - the first popular AI image model.

"Stable Diffusion was the first really popular AI image model that came out, and it blew up on Twitter. It was people sharing these crazy pictures they created. To be honest, they still weren't that good yet... but still, there was something magical about people creating illustrations for anything."

This sparked a realization about what makes creating presentations painful - the formatting and decorating, finding clip art and images to fill space. The team began to see AI as a potential solution to this problem, initially viewing it as a way to solve their "cold start" problem by helping users quickly create their first presentation.

"What if we could speedrun you through that whole first hour of using the product and make it a first five minutes instead by using AI to generate that first presentation? That was kind of our initial narrow idea."

The team threw everything into this approach, experimenting with various AI applications from generating presentations and images to editing presentations. The concept of generating a presentation from a prompt turned out to be particularly compelling, and they focused their efforts on launching this feature.

Timestamp: [04:04-07:01]

Jon describes how a second stroke of luck came when ChatGPT was released in November 2022, just as Gamma was about to run out of money. The world's sudden fascination with AI created an "insatiable demand" for AI applications.

"Chat came out and we all know what happened next in terms of the whole world went crazy for AI, created this sort of insatiable demand for 'what can I do with AI,' people talking about AI and social media sharing what they made."

At the same time, AI models were rapidly improving. OpenAI released GPT-3.5, which enabled Gamma to launch their AI generator in the first few months of 2023. This proved to be a complete game-changer for the company.

"It was a total inflection point for us. We went from this sort of like middling zone of product market fit like 'I don't know if this is really a thing' to 'Oh my gosh, people are begging us to pay, they're begging us to monetize,' and it's just been a totally different trajectory ever since."

When Michael notes Jon's repeated mentions of "luck," Jon reflects on the nature of timing in startups, using a sailing metaphor to explain.

"Luck in startups is like sailing. You have to catch the right wind, but you also have to have rigged up your sail and put your boat in the right place to be able to ride the wind."

Jon emphasizes that while timing was crucial, their three years of prior work building a product, team, and technology base positioned them to capitalize on the opportunity.

"We had to be there when lightning struck, but we also had to have been building for basically three years ahead of that... so that when it came, we were in the best position to seize that moment."

He concludes that their years of development before the ChatGPT moment proved "immensely valuable" for Gamma, as it gave them the robust infrastructure needed to scale when the opportunity arose.

Timestamp: [07:04-08:31]

• PowerPoint was created in 1987 and while it's had incremental improvements, the fundamental paradigm has remained unchanged for nearly 40 years
• Gamma started with a vision to completely rethink presentations, not just create a better editor, aiming for a hybrid between presentations and documents
• COVID created what seemed like the perfect "why now" moment for disrupting workplace tools as companies like Zoom and Slack saw explosive growth
• By 2022, Gamma had achieved only partial product-market fit with a few hundred active users, facing potential business failure within a year
• Stable Diffusion (not ChatGPT) was the initial AI breakthrough that sparked Gamma's pivot to AI, addressing the pain point of formatting and decorating presentations
• Gamma's initial AI strategy was narrow: using AI to solve their "cold start" problem by generating users' first presentations
• The launch of ChatGPT in November 2022 and the subsequent AI boom created perfect timing for Gamma's AI-powered presentation generator
• The AI pivot completely transformed Gamma's trajectory, shifting them from struggling for product-market fit to having users "begging to pay"
• Successful startups need both timing ("luck") and preparation - having the infrastructure and foundation ready when opportunities arise

Timestamp: [00:02-08:31]

Companies & Products:

• Gamma - AI-powered platform for creating presentations, websites, and social media assets
• Optimizely - Previous company of Jon and co-founders that was acquired in 2020
• PowerPoint - Microsoft's presentation software created in 1987, mentioned as overdue for disruption
• Google Slides - Described as "PowerPoint on the internet"
• Keynote - Described as "PowerPoint for Macs"
• Zoom, Loom, Slack - Companies that saw explosive growth during COVID, creating the "future of work" VC thesis

Technologies:

• Stable Diffusion - The first popular AI image generation model that influenced Gamma's pivot to AI
• ChatGPT - AI chatbot by OpenAI whose release in November 2022 created global AI enthusiasm
• GPT-3.5 - OpenAI's language model that enabled Gamma's AI presentation generator

People:

• Jon Noronha - Co-founder of Gamma, interview subject
• Michael Mignano - Partner at Lightspeed, interviewer

Concepts:

• Product-market fit - Mentioned as a struggle for Gamma before their AI pivot
• Cold start problem - The challenge of getting new users to see value quickly in a product
• Future of work - VC investment thesis during COVID pandemic
• Startup luck/timing - Compared to sailing - needing to catch the wind but also having your sails ready

Timestamp: [00:02-08:31]

Michael brings up how many creative tools need to expand beyond a single format to become successful businesses, noting that Gamma has moved beyond just presentations. He wonders if this was a strategic decision to drive growth.

Jon confirms this insight, reflecting on their fundraising challenges:

"When we were in those dark times of wondering if we had a path as a company, one of the biggest things standing in our way was that so few investors wanted to invest in a presentations company."

Despite PowerPoint and Google Slides each having hundreds of millions of users, investors viewed presentations as:

• An "episodic" rather than "sticky" use case
• Competing against massive bundled suites like Office and Google Workspace
• A "graveyard of startups" with "a lot of bodies buried"

Jon explains that even before their AI pivot, they were thinking about expanding their offering, inspired by Canva's success:

"Maybe the best inspiration for us has been Canva, which has really shown the ability to build a blockbuster business by diversifying across different assets and diversifying both in terms of making different stuff but also winning over the same customer at different points in their week or their month."

Timestamp: [8:38-10:30]

Jon details how Gamma has expanded to multiple formats beyond presentations while maintaining a coherent product strategy.

"Presentations was sort of our starting point, it's still our biggest hook, but we've now added all these other formats that we support."

These additional formats include:

• Document builder for "shiny PDFs and brochures and ebooks"
• Websites, which Jon notes are often functionally similar to presentations
• Social media assets like LinkedIn carousels

Jon explains how these formats are more interconnected than they might seem:

"A lot of websites are often quite simple and they are kind of just an online brochure. If you think about a small business owner or a coach who's trying to represent themselves online, actually their pitch deck and their website are almost the same thing just in slightly different shapes."

This interconnectedness enables what Jon calls "malleability," which AI enhances:

"We kind of lean to this idea of malleability which AI really enhances - start with a presentation, turn it into a website, turn it into a PDF that you can leave behind by printing out at a client's site."

Their goal is to expand from the current formats to potentially ten different formats, transforming users from monthly to weekly to daily active users.

Timestamp: [10:30-11:35]

Michael asks whether these different formats are attracting the same users or different customer segments. Jon's answer challenges conventional startup wisdom about focusing on vertical niches rather than horizontal platforms.

"We have been amazed by the surprising similarity across different parts of our product offering and the ability to go horizontal."

Jon notes that during fundraising, investors typically pushed for vertical focus:

"So many investors, especially at that early sort of pre-seed stage, don't want to hear about your horizontal product. They want to hear about your one persona that you're winning over."

He acknowledges the logic behind this advice - "if you try to serve everyone, you serve no one" - and admits Gamma almost "crashed and burned" pursuing their horizontal vision.

However, their initial user research revealed surprisingly consistent pain points across diverse users:

"The very first thing we did before building anything was user research. We just reached out to our own networks and we basically talked to a hundred people and just asked them 'tell me about the last presentation you made'... It turns out everybody had made a presentation because it's just common across any jobs, and they actually all had the same gripes."

Teachers, students, consultants, and doctors all struggled with formatting, being judged on visuals, and keeping audience attention - suggesting a horizontal opportunity.

Their expansion into websites reinforced this pattern:

"The people who are making websites - there is a distinction of like the designer/developer versus everybody else... but 98% of people are in the 'everybody else' and they have the same needs: 'I just want to get something professional looking live without stressing about code and needing to pay somebody to maintain my site or change my phone number on it.'"

Timestamp: [11:36-13:32]

Michael asks a fundamental question about what's actually driving Gamma's success: is it the AI capabilities that are drawing people in, or is it the core product experience?

Jon provides a clear formula for their growth strategy:

"The AI is pulling them in, but the core product is making them stay."

He elaborates that AI drives their virality and provides the initial wow factor:

"AI is what drives the virality. It's something we can impress you with in the first five minutes of sign up, and it truly does answer that question of 'how is this 10x better than what came before?' You can literally do something in two hours that took you 20 hours before."

However, Jon pushes back against the oversimplified "GPT wrapper" characterization that some apply to AI applications:

"I think maybe what that perspective misses is this idea that often the wrapper is much thicker than the thing that is being wrapped. It's almost like you have this magical jewel or something that you're putting into Thanos's arm, but you have to have the whole rest of the stuff too."

Jon reveals that Gamma actually spends much more of their development time on core product features than on AI improvements:

"We spend much more of our time on features like having really good PDF export than we do on making the AI smarter, quite frankly."

This approach differs from some AI applications like coding tools where the UI is "pretty thin" and primarily focused on integrating AI into a codebase. Instead, Gamma focuses intensively on user experience:

"Presentations are a highly user experience intensive product, so is website building. Most of our work is actually UX design and front-end engineering - it's how do we actually make it all work together."

This focus is reflected in their team composition:

"Our team is like one-third UX designers... zero ML engineers. Very different than maybe your typical AI company, but probably more similar to what you would see out of a consumer software company."

Timestamp: [13:33-15:44]

• Despite PowerPoint and Google Slides having hundreds of millions of users, investors were hesitant to back presentation companies due to the "episodic" nature of the product and competition with bundled suites
• Gamma's expansion strategy was inspired by Canva's success in diversifying across different creative formats to transform users from monthly to daily active users
• Traditional startup advice pushes founders toward vertical focus on specific personas, but Gamma found surprising similarity in needs across diverse users
• Initial user research showed consistent pain points across presentations (formatting struggles, visual judgment, keeping attention) regardless of profession or background
• For website creation, Gamma found that 98% of people share the same basic need: getting something professional without worrying about code or maintenance
• In Gamma's product strategy, "AI is pulling them in, but the core product is making them stay" - AI provides the initial wow factor, but retention depends on the overall experience
• Gamma challenges the "GPT wrapper" criticism by investing heavily in UX design and front-end engineering rather than ML improvements
• Unlike many AI companies, Gamma's team is one-third UX designers with zero ML engineers, resembling a consumer software company more than a typical AI company
• Gamma's "malleability" concept allows users to transform content between formats (presentation → website → PDF → social media), which AI enhances

Timestamp: [08:38-15:44]

Companies & Products:

• Gamma - The company expanding from presentations to multiple content formats
• PowerPoint - Microsoft's presentation software with hundreds of millions of users
• Google Slides - Google's presentation software with hundreds of millions of users
• Office - Microsoft's productivity suite that bundles PowerPoint
• Google Workspace - Google's productivity suite that bundles Google Slides
• Canva - Creative platform that inspired Gamma's multi-format strategy
• Cursor - AI coding tool mentioned as an example of thin UI with deep AI integration

Formats & Content Types:

• Presentations - Gamma's initial and still primary format
• Documents - Described as "shiny PDFs and brochures and ebooks"
• Websites - Described as often functioning like "online brochures"
• Social media assets - Including LinkedIn carousels
• PDF export - Mentioned as a core feature Gamma invests heavily in developing

Concepts:

• Malleability - The ability to transform content between different formats
• GPT wrapper - Criticism of AI applications as merely thin interfaces to language models
• Horizontal vs. vertical focus - Contrasting business strategies discussed in fundraising

People:

• Thanos - Marvel character referenced in an analogy about AI integration ("like a magical jewel in Thanos's arm")

Timestamp: [08:38-15:44]

Michael asks Jon whether Gamma is developing or fine-tuning their own AI models, or if they're primarily using off-the-shelf models and focusing on product design.

Jon clarifies their approach:

"It's off-the-shelf but really heavy prompt engineering. We've kind of dabbled in things like fine-tuning, but we've actually found that prompt engineering is where a lot of the real power comes."

He emphasizes that most users don't know how to prompt and don't want to learn, so simply providing a prompt box with an output wouldn't create a successful product. Instead, Gamma creates extensive middleware between what users type and what the AI model receives.

"There's an enormous amount of stuff in between what the user types into our product and what the AI model sees, which is where I think a lot of our value comes in."

Jon explains how this approach has influenced their organizational structure:

"Our UX designers are also prompt engineers are also front-end developers. We have these very messy overlapping roles where people have multiple skill sets that they combine to provide the right experience at the end of the day."

He provides a concrete example of this integration with AI image generation:

"We've generated almost a billion AI images... When you make a presentation that has 15 images in it, we do a lot of work to make sure they all have the same visual style and feel like they all came from the same place versus just a bunch of random prompts being sent off to DALL-E or whatever it is."

Timestamp: [15:50-17:18]

Michael raises a common criticism that if a company's defensibility is based primarily on prompting, they're vulnerable as AI models improve. As an investor in the application layer, Michael disagrees with this view but asks Jon how he thinks about Gamma's defensibility.

Jon acknowledges that simple text generation applications might face challenges:

"If all your product did was, for example, a blog post generator where you type in a line of text and you get out a 10 paragraph blog post, then I would tend to agree that you're toast. All you are doing is standing in between a really good text generator and the output."

However, he argues that the situation is different for complex applications:

"It's different when you talk about a real application that somebody lives in to do sophisticated work."

For Gamma, generating presentation text is actually not the hard part. Their real value comes from:

1. Domain-specific layout and visualization: "Laying out content in a presentation... structuring ideas into multiple slides, figuring out where content goes on each slide, coming up with different ways of visualizing stuff."

2. Creating proper UX abstractions: "Abstracting away all of that so that the user can give very simple prompts."

3. Ensuring human editability: "Crucial for us, it all still needs to be human editable at the end of the day."

Jon contrasts their approach with AI web app builders:

"There's currently this craze of all these web app builders like your Bolts and your Lovables. I think they're really cool products, but they generally help somebody go from idea to prototype... To make it a fully editable app, you almost need an interface like Webflow."

Gamma's investment has been in creating a "no-code interface all around it" that allows users to manipulate content easily, present it, export to PowerPoint, and publish to custom domains.

"That stuff is actually where all of the pain goes into. I see all of the coding work that our team is doing every day, all the pull requests, and 99% of it is that kind of thing, not to mention making the whole application scale."

Timestamp: [17:19-20:12]

Michael summarizes what he's hearing from Jon - that Gamma isn't just a "wrapper" that takes model outputs and delivers them to users with minimal changes. Similar to pre-AI software companies, they're providing value by improving workflows, creating a complete product experience.

Jon strongly agrees:

"Absolutely, and that product is great because of the way that it fits into their life and their workflow. It's because you are establishing a trusted relationship with that user that they can come back again and again and get a predictable result by working with you."

This predictability is challenging because AI is inherently unpredictable:

"So much of the magic is smoothing over the bumps of what AI can do and exposing it in a way that has nice, clean handles for people to grab onto."

Jon emphasizes that users aren't choosing Gamma because of the AI technology itself:

"They're not using it because it's AI. In fact, I would argue customers probably don't even care about that. Nobody uses a product for the technology."

Michael agrees, highlighting how the technology is just an enabler for solving real user problems.

Timestamp: [20:13-21:19]

Michael points out that every company now wants to leverage AI, including large incumbents who are injecting AI into their existing creative suites. He asks how Gamma plans to compete with these giants who are moving in the same direction.

Jon acknowledges the inevitability of this competition:

"You either die an obscure startup or live long enough to fight the big suites, and so our hope is to live long enough to fight the big suites."

He explains their competitive advantage is being lean and nimble in a rapidly changing landscape:

"Imagine if AI development were to totally pause... In that world, what you would see is that the PowerPoints and Google Slides of the world would then have the predictability to be able to invest and plan something. But that's not quite the world we live in."

Instead, capabilities are changing "drastically and quite frequently," forcing companies to rethink their products roughly every two months. Large organizations struggle with this pace:

"I got my start of my career working at Microsoft, and I just know that large organizations don't have the kind of tight decision-making loop where they can change their core products that quickly."

Jon explains that established products face constraints that startups don't:

"Most PowerPoint users don't want PowerPoint to change that fast. Enterprise sellers of PowerPoint don't want the product they're selling to change under their feet... What their buyers are buying is the predictability of 'I just want the same thing that I learned when I was in high school in 2005.'"

This creates an opportunity for nimble startups:

"It creates a really nice window when there is technology change for new companies to come in who are not bound by the past and who can make very quick decisions about pivoting their product."

To maintain this advantage, Gamma has:

1. Kept their team intentionally small: "We are about 35 people right now, where many companies I think with our scale of users and revenue would already be in the hundreds, maybe 200 or 300 people."

2. Focused on product-led growth rather than enterprise sales: "Many companies in our position really feel the pressure of 'sell to really big businesses'... Once you go enterprise and build those very long-term contracts and relationships, you start optimizing for predictability."

Jon concludes that while they see the revenue opportunity in enterprise, their current priority is innovation:

"Right now we think the most important existential thing for us is staying ahead and innovating, and that means preserving our chance to be very, very nimble."

Timestamp: [21:20-24:59]

• Gamma uses off-the-shelf AI models with heavy prompt engineering rather than developing or fine-tuning their own models
• Most users don't want to learn prompt engineering, so Gamma creates extensive middleware between user inputs and AI models
• Gamma's team structure blurs traditional boundaries, with UX designers also serving as prompt engineers and front-end developers
• For complex applications like Gamma, defensibility comes from the entire product experience, not just AI text generation
• The real value comes from domain-specific features like layout, visualization, abstraction, and human editability
• 99% of Gamma's engineering work focuses on the surrounding product experience, not the AI capabilities themselves
• Users choose products for their utility and reliability, not for the underlying technology like AI
• Large incumbents like Microsoft and Google struggle to rapidly adapt their core products to fast-changing AI capabilities
• Enterprise customers often value predictability and stability, creating an opportunity for nimble startups during periods of technological change
• Gamma has intentionally maintained a small team (35 people) to preserve agility and has focused on product-led growth rather than enterprise sales to maintain flexibility

Timestamp: [15:50-24:59]

Companies & Products:

• Gamma - AI-powered presentation and content platform focused on user experience
• Microsoft - Mentioned as Jon's former employer and PowerPoint creator
• PowerPoint - Presentation software known for stable, predictable user experience
• Google Slides - Google's presentation software, mentioned as a major incumbent
• Webflow - Referenced as an example of a fully editable no-code website builder
• Bolt/Lovable - AI web app builders mentioned as helping with prototypes but not full products
• DALL-E - AI image generation model referenced as one Gamma might use

Technologies & Concepts:

• Prompt engineering - Described as where "a lot of the real power comes" in AI applications
• Fine-tuning - AI model customization approach that Gamma has "dabbled in"
• AI image generation - Area where Gamma has created "almost a billion AI images"
• AI wrappers - Criticized as just standing between AI models and users with minimal value-add
• Product-led growth - Growth strategy Gamma has chosen over enterprise sales
• No-code interface - Type of user interface Gamma has invested heavily in building
• PowerPoint export - Feature mentioned as requiring significant development effort
• Custom domains - Publishing capability mentioned as part of Gamma's complete solution

Team Structure:

• UX designers - Compose one-third of Gamma's team and often serve multiple roles
• Prompt engineers - Role that overlaps with UX designers at Gamma
• Front-end developers - Another overlapping skillset in Gamma's "messy" organizational structure
• ML engineers - Noted as absent from Gamma's team composition

Timestamp: [15:50-24:59]

Michael asks Jon about future directions for Gamma and how they'll adapt as technology advances. Jon identifies a significant shift they're already working on:

"One of the big ones that we're working on right now actually is this shift from maybe tool-based interfaces to agentic interfaces."

Jon acknowledges that "agentic" has become a somewhat meaningless buzzword, but explains the core product design trade-off they face:

"How much should Gamma be conversational in the sense that you talk to it though you would talk to ChatGPT versus how much should it feel more like a traditional software tool where you just sort of push some button and clip inputs and then generate something?"

He reveals that Gamma has "flip-flopped on this a few times" and expects to continue oscillating because each approach has distinct advantages:

• Tool-based interfaces are familiar to users who aren't interested in AI and don't want to become prompt engineers
• Conversational interfaces match how models are optimized and allow users to clarify intent and make corrections

Despite some predictions that chat interfaces are "dead," Jon doesn't believe this is true, pointing to ChatGPT's dominance in consumer AI:

"ChatGPT [is] by far the dominant consumer AI product by like probably a factor of 10 compared to anything else."

He highlights two key advantages of conversational interfaces:

1. AI models are optimized for conversation
2. Conversation allows for clarification and correction: "When they say 'make me a presentation about XYZ,' there's a chance for AI to say 'when you say you want a presentation about this, are you picturing something more professional or more whatever? Are you thinking like 30 slides or 15?'"

Timestamp: [25:07-27:28]

Michael acknowledges the longstanding debate between conversational and graphical interfaces, noting that there's been apprehension about returning to chat interfaces after decades of GUI development:

"We've always had terminals, this is why we evolved to having the GUI and tactile buttons... Why are we going backwards? Why are we just chatting with everything?"

He points out that creating dynamic interfaces that evolve based on user input is technically challenging and potentially disorienting for users since "the best product designs are ones that have a lot of clarity and the user has a deep understanding of where things are."

However, he suggests that chat interfaces feel natural because we already spend our days messaging with others:

"We understand it not just because we use ChatGPT but we message with each other all day every day on iMessage and Slack and Discord."

Jon frames this as an open question about the end state of the application layer, presenting two contrasting visions:

1. Rich graphical applications: "These super heavily designed software applications like imagine a Slack or a Notion," which dominated previous decades.

2. AI as junior employees: "Is the application layer going in the direction of a junior employee who works for you, where the interface is you talk to them?" Jon cites products like Devin where "the whole interface to you is you're just messaging them on Slack."

He observes that this distinction has profound business implications:

"GUI software apps are priced per seat, but employees are priced per hour. Depending on which one of those paths we go, we might end up with a different business model as well."

Timestamp: [27:29-29:42]

Michael asks Jon about the role of voice interaction in future AI interfaces, specifically whether there's a world where users could talk to Gamma and find it more efficient.

Jon believes voice interaction is likely in Gamma's future:

"I think there is. I think that's a pretty likely thing that we will do on our roadmap."

He explains why voice makes sense for Gamma's vision:

"Conversation is such a natural medium. Speaking is often easier than writing, and we kind of have this mindset of 'we are trying to build your AI design partner.'"

Jon draws a parallel to how high-level executives work with human presentation designers:

"CEOs, high-level executives often don't make their own slides. Instead, they have people who do that for them... They just give these very vague intentions: 'I want something that shows the three products we're launching, and it needs to feel like a Steve Jobs keynote.'"

This executive-designer workflow involves iteration and feedback:

"The poor human has to come up with 'well, here's three ideas of what I think you meant,' and they say 'no, no, not like that at all, and by the way, the middle one, the launch is delayed, so rework the whole thing.'"

Jon predicts AI-powered presentation creation will follow a similar pattern but with key advantages:

"Instead of it being a human who takes a day and a half to get the results back and is stressed out the whole time, it's an AI who gives you those three options in a matter of minutes... When it's not three options, it's like 10 options, because variety is so free when it comes to AI."

He notes that the multimodal capabilities of modern AI models make this even more powerful:

"The fact these models are multimodal just makes it even more analogous to interacting with a human on the other end... 'Oh, I'm answering your questions, but also let me show you this thing... Here are the actual images. Oh, and I also built the actual product. Here's the software, go try it out.'"

Timestamp: [29:43-31:37]

Michael asks Jon about Gamma's concrete plans for the future - what they're actively working on rather than just adapting to. Jon shares several key focus areas:

1.Enhancing the Editing Experience

"We focused a lot on the initial generation of a presentation. So you give us your notes, we turn it into a rough first draft of slides. That's all well and good, but the rough first draft is only a starting point. Where do you go from there?"

Jon explains that their current interface has limitations:

"We kind of have this disappointing interface right now where we get you that first draft, but it's still up to you to get it the rest of the way yourself."

Their plan is to invest heavily in the editing experience through:

1. AI-powered feedback loops: "Being able to give feedback on what was generated and get another version that adjusts what you wanted."

2. High-level transformations: "I want to make this 20-slide deck be 30 slides instead" or "I want to put less emphasis on topic A and more emphasis on topic B."

3. Structural changes: "Imagine if it's on a website, you have this website with five or six pages, and you say 'I think we should split out this section to be its own page.'"

2.Enhanced Visuals and Styling

Jon emphasizes that presentations are fundamentally visual, so they're investing in:

• Better visualization capabilities to "make ideas come to life visually"
• More styling options to "make things look very distinct and match your brand"

3.New Output Format: Video

Beyond their current four formats (websites, documents, social, presentations), Jon reveals they're intrigued by video as a potential fifth format:

"One that we're really intrigued by is video as a new output format. We all spend a lot of time on YouTube, and a lot of the stuff that you find on YouTube is actually people presenting things."

He describes several potential AI applications for video:

• Writing speaker notes for video presentations
• Creating AI avatars to present content "without all the ums and a's"
• Eliminating the need to re-record when mistakes happen: "If you ever tried to record like a 30-minute presentation, you always mess up in the middle of slide four and have to start again and then stitch it together... Being able to just make that all really seamless would be a huge deal."

Timestamp: [31:38-34:30]

• Gamma is actively exploring the shift from tool-based interfaces to more conversational, "agentic" interfaces, but recognizes advantages in both approaches
• Despite predictions of chat interfaces dying out, ChatGPT remains "the dominant consumer AI product by like probably a factor of 10" over alternatives
• Conversational interfaces allow for clarification and refinement of user intent in ways that button-based interfaces cannot
• The application layer faces a fundamental choice between evolving into richly designed GUI applications or AI "junior employees" that work via conversation
• This interface choice has major business model implications - GUI software is priced per seat while employee-like services are typically priced per hour
• Voice interfaces could mirror how executives work with human presentation designers - providing vague direction, receiving options, and giving feedback
• AI's ability to generate multiple options quickly gives it an advantage over human designers who might take "a day and a half" for revisions
• Gamma's current limitation is that it creates a good first draft but leaves users to handle the refinement process themselves
• Their roadmap focuses on enhancing the editing experience with AI-assisted feedback loops and high-level transformations
• Video creation is a potential new format for Gamma, addressing pain points like recording errors and editing complexity

Timestamp: [25:07-34:30]

Companies & Products:

• Gamma - AI-powered presentation platform considering more conversational interfaces
• ChatGPT - Referenced as the dominant consumer AI product, exceeding others "by a factor of 10"
• Slack - Mentioned as both a messaging platform and an example of a richly designed GUI application
• Notion - Cited as an example of well-designed GUI software that dominated the 2010s application layer
• Devin - AI programming assistant referenced as an example of the "junior employee" interface model
• YouTube - Platform where many presentation-style videos appear, potential output format for Gamma

Interface Paradigms:

• Tool-based interfaces - Traditional button/input-focused software interfaces
• Agentic interfaces - More conversational, AI-powered interfaces similar to human assistants
• GUI (Graphical User Interface) - Evolved from terminals, using visual elements and buttons
• Terminal interfaces - Text-based computing predating GUIs
• Voice interfaces - Speaking to AI systems, potentially more natural for certain workflows

Content Formats:

• Presentations - Gamma's original format, with roadmap to enhance editing experience
• Websites - Current Gamma format with planned improvements for structural editing
• Documents - Current Gamma format
• Social media assets - Current Gamma format
• Video - Potential new format Gamma is exploring for future expansion

Concepts:

• Multimodal AI - Models that can work with multiple types of data (text, images, etc.)
• AI avatars - Mentioned as potential presenters for video content
• First draft generation - Gamma's current strength, creating initial presentation versions
• High-level transformations - Planned feature for adjusting presentation length, emphasis, etc.
• Per-seat vs. per-hour pricing - Different business models tied to interface paradigms

Timestamp: [25:07-34:30]

Michael asks Jon about managing Gamma's explosive growth since their AI pivot, and how it has affected the team, culture, and processes.

Jon reflects on the contrast between their pre- and post-product-market fit experiences:

"It's still been hard, but it's a totally different kind of hard. It's helping me to really understand what product-market fit really is."

He shares a humorous perspective on the difference:

"I actually saw someone on Twitter joke that not having product-market fit is really fun because there's the existential dread, but you also just get to do whatever you want. You just build the things you believe in at kind of whatever pace you feel like."

This changes dramatically once you achieve product-market fit:

"Once you get on the roller coaster of having a huge number of customers demanding your product, you no longer control the pace and the direction. In many ways, you are being pulled by the market in various ways."

Jon reveals the scale of Gamma's growth:

"We announced recently we crossed 50 million users... That's like multiple orders of magnitude beyond what we had ever thought this thing would do."

This rapid growth created several challenges:

1. Systems breaking under scale: "So many systems break. We spent a lot of our times just keeping up with demand."

2. Unexpected user demographics: "We discovered that more than half those 50 million users didn't even speak English, weren't in the US. So you have to shift your priorities around and suddenly build internationalization."

3. Roadmap disruption: These urgent needs "hurt other things on your roadmap."

Despite these challenges, Jon says the team has embraced the experience:

"I think our team has been very resilient and adaptable. It's a lot of change, but it's also really exciting. I think we all know it's a privilege to work on such a fun product that is going through such intense growth. It's the kind of thing we all only get to experience maybe a couple times in our career, if ever."

He shares an analogy from one of their engineers who described this phase as "Gamma's puberty":

"We've gone from our basic childhood, but we're not the big grown-up company either. We are figuring out what it means to grow responsibly and how to build in a lot of structure we might not have had before, but really keep those roots of this nimble company that's still paranoid it's going to die."

Timestamp: [34:37-36:51]

Michael asks Jon what model innovations from companies like Anthropic or OpenAI he's most looking forward to for Gamma.

Jon immediately identifies one recent breakthrough:

"One of them actually just happened, which is our holy grail for the longest time: images with text in them."

He admits he's surprised this capability wasn't developed sooner:

"I actually thought with the pace of AI this would have already happened, that we would get really good at making images with text in them, but it's always felt just out of reach."

Jon notes that OpenAI's new GPT image model is the first that "really seems like it's there," though it still has limitations:

"It's still not perfect and it's really slow. So I think what we are looking forward to is a world where it's almost effortless and very quick, in a matter of seconds, to make sophisticated images with text in them."

This represents a potential paradigm shift for Gamma:

"In some ways, all a slide is is images with text in it, or vice versa, text with images in it. And so depending on how the technology evolves, it might retool everything about how we approach our product roadmap."

Looking further ahead, Jon is excited about adding motion to these capabilities:

"Take those images with text in them and then turn them into videos, and all of a sudden things like slide animations could be done in a totally different way. They could be totally procedurally generated."

Beyond these "shiny cool" innovations, Jon emphasizes their dependence on fundamental model capabilities:

"We are students of model quality because we are so dependent on it, and we pass that quality on to our users."

He notes that despite talk of AGI, large language models still struggle with basic tasks that are crucial for Gamma:

"Can it follow multiple instructions at the same time? Can we feed it 20 pages of content and make sure that the output it has doesn't make up any facts that weren't in the original content?"

Jon reveals that Gamma rigorously evaluates model performance:

"Coming from our kind of roots in Optimizely, we have our own evals, but we also A/B test them quite heavily. So we're constantly testing: does this new Anthropic model beat this new Gemini model or this new OpenAI model?"

Their testing reveals a more measured pace of progress than what model companies often suggest:

"We kind of have this interesting quantitative sense of model capabilities changing based on user preferences, and they are moving forward. But maybe it's not as drastic as many of the foundation model companies want us to think it is."

Timestamp: [36:52-39:09]

Michael suggests that both AI models and Gamma itself are likely to evolve dramatically over the next year:

"I'm sure these models are going to evolve dramatically over the next 12 months, so maybe if we do this again sometime, we'll be surprised by how different they are, or maybe how different Gamma is by then."

Jon agrees: "Hopefully both are much improved."

Michael compliments Gamma's progress and thanks Jon for the conversation:

"Gamma is doing a great job obviously, so I don't think you have much to worry about there. But Jon, thank you so much, this has been incredible. Really appreciate the time and congrats on everything with Gamma."

Jon expresses his gratitude: "Thanks, I really appreciate it, Mike, and it was so awesome to be here."

The episode concludes with Michael asking listeners to rate and review the show on Spotify and Apple Podcasts, and to follow Lightseed at LightseedVP on various platforms. He mentions that Generative Now is produced by Lightseed in partnership with Pod People, and promises to return the following week with another conversation.

Timestamp: [39:10-39:52]

Podcast Information:

• Name: Generative Now
• Host: Michael Mignano, Partner at Lightseed
• Production: Produced by Lightseed in partnership with Pod People

Call to Action:

• "If you like this episode, please do us a favor and rate and review the show on Spotify and Apple Podcasts."
• "Follow Lightseed at LightseedVP on YouTube, X, LinkedIn and everywhere else."

Next Episode:

• "We will be back next week with another conversation."

Timestamp: [39:33-39:52]

• Achieving product-market fit transforms a company's experience from "existential dread but freedom to build" to being "pulled by the market" in various directions
• Gamma has experienced extraordinary growth, crossing 50 million users, far beyond what the founders had initially anticipated
• Rapid growth created unexpected challenges including systems breaking under scale and discovering that over half their users were non-English speakers
• One engineer described Gamma's current phase as "puberty" - no longer a child company but not yet fully grown up, requiring new structures while maintaining agility
• Images with accurate text has been Gamma's "holy grail," with OpenAI's latest model being the first to come close to their needs, though still imperfect and slow
• The ability to generate quality images with text could fundamentally change Gamma's product approach since "all a slide is is images with text in it"
• Video generation represents the next frontier, potentially revolutionizing slide animations through procedural generation
• Despite the hype around AGI, large language models still struggle with basic capabilities critical to Gamma, like following multiple instructions or avoiding hallucinations
• Gamma rigorously evaluates AI models through both their own evaluations and A/B testing, finding that progress is real but perhaps less dramatic than model companies suggest
• Maintaining a startup's nimbleness and "paranoia" while scaling is a delicate balance that requires intentional culture-building

Timestamp: [34:37-39:52]

Companies & Products:

• Gamma - AI-powered presentation platform that has reached 50 million users
• Optimizely - Company where Jon previously worked, mentioned as influencing Gamma's A/B testing approach
• OpenAI - Company whose new image model represents a breakthrough for text in images
• Anthropic - AI company whose models Gamma evaluates against competitors
• Gemini - Google's AI model mentioned as part of Gamma's comparative testing
• Lightseed - Venture capital firm where Michael is a partner, producer of the Generative Now podcast
• Pod People - Production partner for the Generative Now podcast

Concepts & Terms:

• Product-market fit - Discussed as transforming a company's experience from building freely to being "pulled by the market"
• Internationalization - Urgent development priority after discovering over half of users were non-English speakers
• A/B testing - Method Gamma uses to rigorously compare different AI models' performance
• Images with text - Described as Gamma's "holy grail" capability from AI models
• Procedurally generated animations - Potential future approach enabled by AI video generation
• Model evals - Gamma's internal evaluation system for measuring AI model quality

Growth Metrics:

• 50 million users - Recently announced milestone for Gamma
• Non-English users - More than half of Gamma's user base

Metaphors:

• Roller coaster - How Jon describes the experience of rapid growth after product-market fit
• Gamma's puberty - Engineer's description of the company's current developmental stage

Timestamp: [34:37-39:52]