A Product Market Fit Show | Startup Podcast for Founders
00:17:05 · Alex Halliday
[00:17:05] S0: Yeah, so one thing we would do is we would figure out the intents that people are expressing that the Product Market Fit show probably isn't a good place to answer. So we would have a couple of agents that would spend their time looking at startup conversations, Hacker News, Reddit. Really understanding what the evergreen questions are that people are asking and the discussions people are having. And then what new questions are emerging, because Claude Code is on the rise or the QBs might be removed from New York State, right? So you really want to think about evergreen questions and then frontier questions. So that's the intent cloud that you have a right to speak to, and then it's the content creation to speak to that. And so, you would put all of the transcripts that you've ever had on the show with a definite bias to the most recent ones into AirOps or into a system like Google Drive or wherever they live and you connect them to us. And then, we would have a process which would on a regular cadence, identify the highest priority questions to answer. Would ping you to then go and approve those along with a ton of research from your shows. Probably some questions for you to answer as well to really make sure that that article had your perspective, your overarching narrative and then we would publish that content rich with references to your shows. Like links to where people could go get them on Apple Podcasts or YouTube, and treat that as an ongoing content production drumbeat. So that you can start to get all of the great value out into the world in an organized way, and then on the pages of content themselves. We don't have to go into all, but there's best practices around structuring that content to make sure you're basically spoon feeding those answers to the models and preparing them for citation. Because ultimately you want to be quoted a nice big paragraph in a ChatGPT response saying on the Product Market Fit show, Alex Halliday said, XYZ, and that's a really useful chunk that matches the user intent that you've identified as one. So that would be one loop that would be really valuable to you. You'd be totally within your rights to create that and it's a durable strategy for you. It's not something like gaming the system and consumers would love it. You're adding to the conversation.
00:19:14 · Pablo Srugo, Alex Halliday
[00:19:14] S1: That's the classic blogging approach, just using AI to make sure that you're always touching the latest frontier questions and leveraging as much of the work you've already done in the past. [00:19:25] S0: Yeah, and I think if you were a brand in that same situation. You would really think about what are the offsite, like not on your own domain. What are all the pages that influence the ChatGPT for those priority questions as well. So we do a full audit of that. You've got, a heat map of influence. So maybe Reddit or a Substack, or a YouTube channel driving most of the sentiment and opinion for the things you care about. And so, the play for those would be different. Obviously, it wouldn't look like publishing. It would look more like outreach or conversation participation. [00:19:56] S1: Let me ask a little bit about competition. Competition has really risen to the top, at least for me as a VC. It used to be a bit more of a secondary thing. I think any investment, obviously, you ask about competition but it was not the most important thing. Lately, it just tends to be number one. Because there's so many ways to deliver insane value. Really, the big question is, why isn't everybody else going to do it? With the example you gave, my question would be something like, why couldn't I, or what's the difference between what you're offering and let's say putting all my transcripts into Claude Code. And then, having a scheduled task that runs every week and checks the latest questions, and writes a block for me.
00:20:31 · Alex Halliday
[00:20:31] S0: Yeah, it's a great question. I mean, you certainly could do that. It actually might work quite well, to be honest, given that use case. But if you think about content in the current market, competition exists on the startup side but it also really exists on the content side as well. So each piece of content is in composition with a sea of other pieces of content and so to get the most from an AI model. It's much less so the capability of the model, it's much more so the context that you can have available for the model, right? When doing any sort of content task inside of AirOps, we have a huge data engineering team whose sole job is to make sure that we can give that model an unfair advantage when it's tackling your problem. So if it's, for example, a page you want to have ranking in Google search or it's a particular question for ChatGPT. Before you even start writing we understand every single page that is in competition with it, we understand all of your pages that could be in competition with that, you don't want to cannibalize your own content. We understand which questions you should even be going after in the first place. We collect almost one point two billion answers every sixty days from these answer engines and we analyze them. And then once you create your content, we also have a full picture of your tone and voice, your do's and don'ts in your language, your sitemap for internal linking. So you can think of this as like if Claude is the model at the beginning in the center of the picture, can you give it these concentric circles of knowledge that are actually quite hard to compile, but give it that unfair advantage so the content you create is more likely to win, and then the nice thing also about doing things through us is we also track the impact. So you know, hey, in January we published twenty eight pieces of content. We know that ten of them are absolutely crushing it and we should go do more of those. And that knowledge feeds back in to our agent on a go forward basis. So I think in the case of like working with your transcripts, you would have a really good result just doing that with Claude Code and that works well. I think when you're dealing with tens of hundreds of thousands of pages or millions of pages.
00:22:33 · Alex Halliday, Pablo Srugo
Prioritization is the most important thing, followed by the context you can provide on competition and those are two things we nail for our customers to make sure they continue winning. [00:22:44] S1: I'm really worried because listen, you've been listening for what? Ten, twenty, thirty minutes now? Clearly you like it and the thing is, the next episode is way better and you're going to miss it. You're going to miss it, because you're not following the show. So take your phone out and hit that follow button. And do you see any reasons for what works broadly? Because again, we think about this strategy, it's the same strategy just done with AI. But is there any reason that this changes? This is something you hear a lot. It's like, well, there's going to be so much AI written stuff, slop. Whatever you want to call it, whether it's good or not. It almost, when you just think about that at a high level, you're like, there's no way this is still the future. There's no way five years from now, it's still all about blogging. There's going to be so many blogs, but is that real? Or is this just going to be, this is still the strategy and there's a reason for it?
00:23:27 · Alex Halliday
[00:23:27] S0: I think the human web will continue to exist. I think AI content gets conflated into one blob. I think some of the best content created uses AI and much of the worst content uses AI. That's not the right distinction. So I think understanding that there's really great applications of AI to improve content quality and also improve content health in the long term, is something that I think folks need to understand on that front. But the human web is designed for humans to click around and right now, we're duct taping agents over the top of that. So ChatGPT is going to a search index and is retrieving pages, and including them in its response. And there's a lot of protocols out there that look to try and unseat that, including one called NLWeb and, obviously, traditional MCP as ways in which ChatGPT can ask us questions about the Product Market Fit show and get back the perfect snippets without ever having to visit your website. Irrespective of what the protocol is, whether it's going through a search index and going to the web or going directly to an endpoint for the Product Market Fit show in this example. The underlying content needs to be really, really good. So I actually look forward to us moving on from creating yet more pages of content to something smarter and we actually interviewed Glenn Coates from OpenAI at our last conference. Who is running the OpenAI apps program and he said they're really actively exploring this for later this year. And the way he positioned it to me was you want to be the sharpest tool for the job. Meaning if ChatGPT wants to know about congestion in Barcelona or the best restaurant to go to in Copenhagen. They are going to start to form an opinion of which tool is the best for the job to answer that. Which can give the model the best quality response and allow the model to do even more for that user's journey. Yeah, it's going to change. I think as we've seen with many of these things, the behavior of going to websites and reading content is such a muscle memory for billions of people. The web isn't going away in its current form, but it will have a sibling of kind of agent first web, and I'm excited to see that develop.
00:25:33 · Pablo Srugo, Alex Halliday
[00:25:33] S1: When in the story did you land on the current idea and maybe share the story of how you went from what you were doing originally for marketers to what you're doing today? [00:25:42] S0: Yeah, so I think my co-founder and I got us to about $1.3 million in ARR selling ourselves. [00:25:50] S1: Just the two of you? [00:25:51] S0: Just the two of us. Yeah, and it was a slog. You're on the road, you're trying to persuade people that you can meet their problems. You don't understand the problem well enough to be honest at that stage. It was very hectic and you're also doing all the deal paperwork. It is full on product roadmap, design, everything. [00:26:08] S1: So before you move on, why did you keep it to just the two of you? I've heard a few people do that, and I actually think it's a good idea. But for you specifically, why did you keep it to just the two of you to get to a million ARR? Why not hire at least, I don't know, four or five more people? [00:26:19] S0: Sorry, I meant two of us actively selling and dealing with customers. Yeah, so we were doing the full deal lifecycle, prospecting, we were doing everything. I think the team was probably about eight or nine at that point, yeah. [00:26:32] S1: That makes more sense, gotcha.