Dhaval: This founder built an AI product to help writers beat creative block. He’s a former writer himself, and in this episode we talk about how to approach building AI products in 2023 and beyond. Specifically, Amit shares his journey of how he launched in August of 2022 and generated a profitable company within a short amount of time with long enough runway. To allow his team and himself to chase the product vision without having to raise multiple rounds of capital. Amit Gupta and his team are building. AI writing tool called Sudowrite. He’s a former founder of Photo jojo, a product that he grew to 10 million a year and sold in 2014. After taking a break for a few years, he launched Sudowrite in August of 2020. An app to help create a writers use AI to beat writer’s blog.
Talk to us, Amit. Tell us about your product, your market segment, your ideal users. I would love to hear about it.
Amit:
Sure. Our product is called Sudowrite and we’re building an AI writing tool for creative writers, people who are writing novels, screenplays, any kind of creative story-based narrative content. And today our audience is those individual creators, and I think in the future, We really want to create a story engine, something that helps anyone who wants to tell a story. So that could be any of the people in the world who always thought they had a novel inside of them. It’s too much work. It’s too hard to get that novel out. We want to make it easier.
Dhaval:
Very cool. Tell us a little bit about how you got onto this idea. I know you are a very prolific writer yourself. How did you come up to wanting to solve this problem? And tell us a little bit about your background that you were willing to share with us.
Amit:
Sure. Yeah. I’ve started a couple companies in the past. The last company I started, I sold in 2014. Then I decided to take some time away from Tech. I left San Francisco, I traveled and I really decided I wasn’t going to do the same thing again. I was going to try something different. So I ended up writing and I spent a few years writing fiction, which is where I met my co-founder in a writing group. And as we were writing fiction, we were bemoaning how much harder it was in running startups because, Man, you get no feedback. It’s in a startup, you launch something, right away if it’s working, your users will tell you, you’ll see it in the numbers. And with writing, it can take weeks or even months to get a few pieces of subjective feedback on a longer form piece. So it’s a very different world, much different kind of Learning curve. And we wanted to find a way to bring some of the rapid iterations and the ideation, some of the collaborative elements of working in tech or on a startup to the writing process.
Dhaval:
Very interesting. What was the process like for you to build your MVP? Did you start with wanting to build an AI first creative workflow, or was it writer first and then add AI to it? Tell us a little bit about how you did your MVP.
Amit:
Yeah, we were definitely AI first. My co-founder James started playing with GPT-3 when he came out a couple years back. A few years back. And originally we weren’t planning to start a company. We were just making a tool for ourselves and for our writing group. And then we shared it with some other writers that we knew and respected. And we just got incredible feedback. So we kept tinkering with it. If I think about a year before we decide to really turn it into a company, It was for that year just a hobby project. A lot of writers that are we knew were using it. And we didn’t focus on a lot of the features that most writing tools had. We focused on the things that were unique Sudowrite, the places where we could bring AI to the process and do something new.
Dhaval:
Very interesting. Did you build on top of chat GPT-3 or was that a private beta at that time? How did you get access to that? Tell us all the juicy details of how did you integrate AI in your workflow? And what was the product market fit experience like that like for you?
Amit:
Sure. So this was back in 2020. GPT-3 had just come out. It wasn’t widely available. I think we DM Greg Brockman to see if we could get access and we just wanted it to play with as writers for our writers’ group. So he gave everyone in the writers’ group access, and started playing around with it. Eventually we started building on top of it.
Dhaval:
So where are you in terms of your product? Do Have you launched your MVP do you have revenue? Do you have certain number of users? Yeah. Tell us a little bit about that.
Amit:
Yeah, product-wise, we started building it in. I think like August or July of 2020. And we started letting people in shortly after that. We started charging people, I think last year or late the l year before, not that long ago. So it’s been on the market for a bit now. And I think it’s grown steadily throughout the history. As soon as we started charging for it, people started to sign up. We don’t have a freemium options it’s all paid. And we’ve really kind of seen it hit product market fit in the last couple months, or at least what we think is product market fit. It’s in the past couple months where I think we’ve seen a lot of people signing up, a lot of people kind of entering the community, asking for things, suggesting things that are rate where like it no longer feels like we’re pushing it up a hill. It feels like we’re chasing the ball down the hill trying to catch up. So it’s a, that’s a cool feeling. It’s a very quick reversal.
Dhaval:
Yeah. And do you have any numbers that you can share with us? The number of users you have, your revenue and daily active users or any of that data points? Cause we do have investors also listening to this. So, perhaps they can connect with you if you’re willing to share that. Totally up to you. I understand that.
Amit:
Sure. Yeah. I don’t wanna share too much, but we did share recently that we hit 500K in ARR in December, and I think we’re above 750k right now.
Dhaval:
Wow. Wonderful. Do you have any data points around the number of users you have or market shares or rather market segments that you are going after? Any of that information?
Amit:
Yeah. In terms of market share, I think we obviously have hit like a very, very, very tiny fraction of the market so far. These are really the early adopters. The folks that are willing to experiment with new technology. I think there’s a lot of grounds still to be covered. A lot of things, frankly, we have to improve about the product before it’s really going to be something that someone who is used to a very polished writing experience is going to want to use. But today it’s useful to a lot of people. And I think the market is You know all these people who consider themselves writers, but like I said, in the future, I think that almost anyone who reads has a story in them. So there’s a lot of additional people we think we can bring into the market.
Dhaval:
Yeah, writing is such a fundamental thing, right? Once you figure out writing, you can create many other things on top of it. So I have like two questions that come to my mind. I’ll ask the first question first, which are your writers using your product for instance, notion, like it’s designed for short form documents, type of work, Google Docs notion word or are they using your product for writing? Like find five hundred page, a thousand. Novels or short documents? What type of short stories or long stories do you have a specific workflow in mind for writers? And I’ll let you answer that. I have another question about product management. I’ll get into that in a second question.
Amit:
Yeah. So I think the place where we differentiate is that we’re really heavily focused on long form. And that means that our writers are people who are writing novels or screenplay stuff that can go up to like a hundred thousand words or even longer. That’s something that’s very challenging to do with today’s AI models. It’s something that requires a lot of kind of like work behind the scenes to condense things, summarize things to, to make it work. But I think where a lot of the value lies because it’s very easy to generate short snippets of copy or taglines. It’s much harder to create something Longer and more meaningful and substantive. But that’s where we’re working towards.
Dhaval:
Interesting. And this lends my second question, which is in the product experience, in the customer journey, in the user journey, how did you come to know , this is for AI product creators who are trying to infuse AI in their product. So this is for them, for your product. Amit when you were building it, when you were trying to find that product market fit. At what point was the moment when you decided, okay, at this point in the user journey, we’ll be infusing AI . Tell us a little bit about how did you decide at what point in customer journey was the AI part of.
Amit:
Well, Sudowrite is a AI first product, so there’s really no reason for it to exist without AI today. And I think so even in the onboarding, you’ll start to use AI. You’ll start to experience what it’s like to write with AI. It’s from the very first moment after you sign up.
Dhaval:
Got it. So as if a writer wants to write a book and right away you’ll be like prompting them to get information data points that allows the product to support them in their journey.
Amit:
We’ll show them right away, right after they sign up, what AI can do for them in their process. And I think the eventual goal is that we can help with the whole process from ideation to first draft, to revision, to publication, and just be able to be there with every step in the process. And I think some writers are really good at various pieces and they hate other pieces. So we want something that can help take the drudgery out of writings, wherever that drudgery comes for each writer. because it’s different for everyone
Dhaval:
That’s interesting. Yeah. I’ve talked to many of the AI founders in the last few months, and this is the point where I hear different opinions every time, which is, if you’re a new AI product, if you’re interested in creating a AI product, right? You’re using one of the pre-existing tools, right? Where do you find that personalization? Do you build your own personal models on top of? Pre-trained models like chat GPT-3 or do you go a hundred percent custom? How do you find that differentiation, etc. What are your thoughts on building on top of pre-trained models versus training your own models and building a groundbreaking product from ground up? Where do you like and what are your thoughts there?
Amit:
Yeah, it really depends on the company what stage they’re at, what second on the market and what kind of AI is useful for them. I think with, for instance, stability, there’s a lot you can do as a relatively small company trading that model to do interesting things for you because of the way they’ve sharded it down in size, and I think with large language models, that’s a lot more challenging to run a large language model like GPT-3 requires a lot of infrastructure investments just to have the machines capable of running the thing. And so it’s quite expensive to just run inference before you even get to the concept of like training it, which could cost millions. So I think there’s definitely a lot to be gained from training custom models, but at this point in the evolution of them they’re Iterating and developing some rapidly that a lot more of the gains come from things like fine tuning or developing models that sit on the layer above or below GPT-3 or otherwise finding ways to really get the best quality out of GPT-3.
Dhaval:
So fine tuning on top of pre-trained model is the way to go for a new AI product creator. However, If you’re working for one of those organizations that has the data available and has the resources available, building your own model may be an option.
Amit:
Potentially, you’d have to see how much additional value you really think that custom model will bring, and if it’s worth the expense. I think even if you are able to build it and have the data, it may be the case that still doesn’t make financial sense to build an LLM yourself today. Because as soon as GPT-4 comes out, it’s going to leapfrog whatever you’ve run and then you’re going to have to train a whole new model to just to keep up with the new kind of state of the art.
Dhaval:
Very interesting. Okay, tell us a little bit about your future vision of the product. Where do you see this going? Where do you see Sudowrite going ?
Amit:
So today I think it’s very useful for Sudowrite very useful for people who know how to write and who want to write faster, who want to get unblocked, who want to keep in the flow. And those writers come to us and they tell us that PseudoWrites able to help them double their word count so they can publish twice as much stuff. And they’re very excited about that. We’re in a world where as soon as you finish watching a TV show on Netflix, you want the next season. And the same is true of books. So as soon as a reader finishes reading a book, as soon as a writer has published a book, the next day they’re hearing from readers that they want the next one. And of course, you can’t deliver that, but I think in the future we want to create a system that helps the writer deliver that helps them. Serve their readers, produce more work, produce more artifacts aside from the novel that take place in the world and in the story that they’ve created, and basically find new and interesting ways to monetize their IP. In the same way that bigger pocketed creatives today, like the Marvel Universe, or Star Wars or whatever, are able to like create a lot of storylines in the same world, a lot of different product offerings. We want to bring some of that power to the individual.
Dhaval:
Wow. Turning individual writers into franchise. Wow. That’s a powerful vision. Amit, have you raised capital? Is this a bootstrap startup? Have you raised a seed round and if not, are you planning on doing it soon?
Amit:
Yeah, we raised a seed round late 2021 $3 million. We raised it for primarily from individuals, so people who are in the creative fields, in other related fields, and we have no immediate plans to raise more money. We have quite a bit of runway.
Dhaval:
Would you say that a AI product creator who is interested in getting their feet wet and want to build something bootstrapping to validate the idea and validate the hypothesis may be the best way to go forward? Or is there too much a bar? High bar for them to do that. Trying to bootstrap Were you at all at any point bootstrapping this and tell us a little bit about that experience?
Amit:
Yeah, it was Sudowrite bootstrap for the first year. I think there’s absolutely no limitation on anyone who wants to get started with generative AI today. Building a startup from the ground up with just one person and your spare time, I think the tools are incredibly powerful and there’s a wide open field for so many different solutions that can still be made.
Dhaval:
Thank you for sharing that. That’s very inspiring to hear because you’re a hundred percent right. There are so many remain to be optimized. And it’s up to the creative potential of the individual who’s trying to create that product to decide where they want to go after what they wanna go after. Yeah, that concludes all my questions. Thank you for being so on the point, man. Thank you so much.
Amit:
Yeah, sure. Glad to do it. Happy to meet you.
