Dharmesh Shah, co-founder and CTO of HubSpot, joins Sam and Shaan to discuss how he pioneered inbound marketing through blogging and SEO, why that playbook is changing with the rise of AI search tools, and the emerging discipline of AEO (answer engine optimization). The conversation also covers Dharmesh’s personal framework for building side projects, his philosophy on iteration, and how he thinks about category creation and community as business strategy.
Speakers: Sam Parr (host), Shaan Puri (host), Dharmesh Shah (guest, co-founder and CTO of HubSpot)
Introduction: Who Is Dharmesh Shah? [00:00:00]
Sam: All right, so we had Dharmesh Shah on the podcast today. Dharmesh is interesting. One, he’s probably the most listened-to and viewed guest that we’ve ever had. And two, the reason he is that way is because Dharmesh runs HubSpot, which is a $25 to $30 billion company that makes billions in revenue. He’s a billionaire. He also is sort of like a hacker — he launches little projects all the time. Sometimes they work, sometimes they fail, sometimes they make $5,000 a month. He’s probably the only person on earth who has this perspective of both being a big thinker and also launching new projects.
Shaan: Yeah. On this episode we wanted to talk to him about the early days — like how did you actually get HubSpot off the ground? You had no money, you didn’t have any connections. He figured out a new model of marketing around blogging and SEO, and we wanted to talk about what that looked like then. Because today with AI and ChatGPT it seems like there’s a new opportunity. I have more energy than I started with because he’s got this kind of playful, builder, get-after-it energy that’s pretty contagious.
Sam: All right, give it a listen. Let us know what you think in the YouTube comments.
Catching Up at HubSpot’s Inbound Conference [00:01:00]
Sam: Dharmesh, what’s going on? How are you doing?
Dharmesh: Back from a busy but fun week at Inbound in San Francisco. Did you guys get together in SF?
Shaan: Yeah, we did. Dash is like — you know, this was his Coachella. He’s there. He’s the guy.
So Sam, I go to San Francisco and basically the street is shut down. They take over the Moscone Center and they have this thing called Inbound. Dharmesh was like, “Hey, after your talk, let’s go hang out.” So we go hang out, and my first question was like, “Dude, what is this?” I know it’s a conference or an event — I’m not confused about that — but I’m looking at it and there are these giant block letters that just say INBOUND. It doesn’t even say HubSpot. And there are all these people walking around and I’m just thinking, why are they here? What is this all about?
He actually told an interesting story about the origin of this that I think is a pretty interesting business strategy. Dash, tell the story. You were telling me that Brian was talking to these well-funded companies that had CMOs trying to get website traffic, and here’s Dharmesh just blogging for fun at home and getting way more traffic. And they kind of looked at that like, “They have a chief marketing officer. They have an ad budget. How are you getting ten times more visitors?” There’s something here, right?
Dharmesh: Yeah, that’s exactly the story. I was still in grad school finishing up my thesis. Brian graduated a year ahead of me, but we’d still get together once a week and chat about ideas because we’d considered the possibility of starting a company together. He was working as a venture partner at a VC firm in the Boston area, and his role was to help startups with go-to-market — marketing, sales, how do you get customers.
My thesis was called “Uncharted Startups: The Patterns and Practices of Modern Software Entrepreneurs.” I’m a software guy, and my thesis adviser — one of the tougher advisers — about three-quarters of the way through says, “Dharmesh, you’ve got a lot of words here but not a lot of data. This is not Dharmesh’s opinion. This is supposed to be a thesis. You need research.”
My immediate thought was: what he’s asking for is going to involve interacting with humans. That sounds awfully unpleasant. That’s not what I signed up for.
So I thought blogging was a new thing at the time. Maybe I can write a blog and get comments, and use that as citation. I started a blog. To come up with a name, I just took the first two words of my thesis title — “on startups” — stuck them together and put a .com at the end. That was available, so I registered it for $15 or whatever.
And then it’s like, how do I get people other than my mom and two friends to read it? So I dug in. There’s this thing called Google. They’ve got two types of traffic — you can pay your way in, or you can do search engine optimization and get organic traffic. There are social media sites like Digg and Reddit. Here are the things you do to get people to your website. Those are all the things I obsessively dug into and learned about.
So Ryan is watching me do this when we’re meeting. He’s like, “Dharmesh, how’s that blog going?” I said, “It’s going pretty well. Here’s the traffic.” He says, “How the heck are you doing that?” I say, “Oh, there’s this thing called SEO — here are all the things you do.” He goes, “Well, my portfolio companies can’t get a hundredth of that traffic. They should be doing this instead of whatever it is they’re doing. Stop the madness of spending this budget on things that aren’t driving any engagement.”
That was the genesis of the idea of inbound marketing. Everybody was doing it wrong. Our catchphrase at the time was: try to add value to your prospective customers before you try to extract it. All other marketing was forms of extracting value from you — how do I interrupt you at dinner? We were the opposite of that.
Building the Inbound Movement and Conference [00:08:00]
Dharmesh: So back to the conference. We didn’t have a product yet. The tools I was using were common things — Google Analytics, blogging software. We decided two things. One, the reason so many people weren’t doing inbound marketing is not that it didn’t work — it was that it was just too hard to put all the pieces together. Two, there was a lot of nodding of heads around the idea, so let’s start blogging about it. Let’s get up on stages. And let’s get people who believe in this new way together.
The first conference was at a Marriott across the street from the office. It was 200 people in a ballroom. We made a decision from the start: this is not intended to be a company conference. There’s this thing called inbound marketing which we think is a better way, and all we’re doing is acting as a host to bring the community together to talk about these things.
In fact, we forbade — and I don’t think I’ve ever used the word “forbade” before — HubSpot employees from pitching HubSpot from stage. No one was allowed to pitch their companies. We kept that up. That’s why the conference was never named HubSpot. This is the 17th year we’ve been doing it. It was a convening of the community, not a classic tech conference where CEOs get up and pitch their book.
A few smart things that I think worked and that everyone should try if they’re looking to start a movement or start a category: we did not trademark the term “inbound marketing” even though we coined it. If this thing is actually going to be a thing, we want everyone to use it. We want people to put it in their job titles and their job postings. We want it to be a thing. And the way to make it a thing is to make it accessible to everyone.
Even in the early years we had direct competitors sending half a dozen people — their entire teams — to the Inbound conference even though HubSpot was a competitor. It was an industry event, not a company conference.
Niche Communities and the Pattern of Movements [00:14:00]
Sam: I was part of the early blogging community — first as a reader and then I had my own small blog. From ‘08 to around ‘14 was a really magical period for blogging. Back then we felt like freaks. There were very few of us and we were kind of weirdos — basically all nerds who loved meeting other nerds. Then when I lived in SF with Shaan, there was startups, and then there was crypto, and crypto was the same thing: a very small and very passionate group of people. They were freaks, they were weirdos.
My takeaway from hearing you tell this story is that I’ve seen this happen so many times — niche communities. Early YouTubers. We knew early YouTubers and it really became a thing. The signal is how passionate they are, and it’s a tiny group.
Shaan: Who’s that now?
Sam: I think AI is already past that. So not AI. But I think Peter Levels has done that with the indie hacker thing — that was really cool. Brian Johnson was that a couple years ago. It’s obviously easy for me to say what’s past. I don’t actually know what it is now. What do you think it is now?
Shaan: Well, I was thinking about the HubSpot strategy — the category creation where you make it about a movement and a way of doing things instead of starting with your company and your product. I thought of two examples just while he was talking.
The first was Brian Johnson. When Brian started he didn’t have a product to sell, so he basically created this way of biohacking and longevity — let’s call it the “don’t die” movement. Like, if you believe what I believe, then you sleep at this time, you put this mask on your face, you get this test done every six weeks. You do all these weird things because you believe in this way of living. I thought he did an incredible job of creating that movement. Along with Huberman and others, of course — no one person ever owns these things.
And another example, a little smaller and more niche: what Russell Brunson did with ClickFunnels. First the idea — just the rebranding of the word “funnel.” We already had websites, landing pages, growth — all these other words. But then to own a word like “funnel” the way HubSpot owned “inbound marketing.” They have funnel hacking competitions and conferences not called the ClickFunnels meetup — it’s for people who are great at funnel building, and they make them feel like stars and give them plaques. And then of course, if you’re going to build funnels, you should use ClickFunnels. That’s the secondary thing underneath it.
Making Inefficient Markets Efficient [00:21:00]
Dharmesh: One category of ideas I’ve been obsessed with for 30 years is making inefficient markets efficient. The macro-level definition of an efficient market is that every transaction that should occur does. A good example is the public stock market — someone wants to sell shares at Y price, someone wants to buy at Y price, and so the transaction happens.
What makes markets inefficient? Thing one: the buyer and seller don’t know each other. Thing two: they know each other but can’t come to an objective way to determine price. Thing three: they can come up with a price but there’s not enough trust for the transaction to clear.
Some of the most money ever made has been when someone took an inefficient market, identified the latent economic value being wasted, and removed some of that inefficiency. eBay was that for niche market products like Pez dispensers. Google in a way was creating an efficient market for niche information — “I’m writing a blog post about building fences for llamas, and here’s someone searching for that content.” Google is the market maker for niche content across a very very long tail.
And this idea of a community is really, from a business perspective, a form of creating an efficient market. These are people who would enjoy knowing each other, maybe transacting, maybe hiring each other. Just the act of pulling them together removes some market inefficiency, and as a result you get to participate in some of that economic value you uncovered. That’s how communities make money. And when they do well, they do really, really well on a profit margin basis — relatively cheap to start, almost self-sustaining once you get to critical mass, and you have network effects that make it hard to disrupt once the community is good and going.
Eras of Marketing: SEO, Paid Ads, and Algojacking [00:28:00]
Shaan: One of the things we talked about when we were hanging out was this story about early days blogging — getting crazy amounts of traffic, then realizing there are other people who want to do this, then building a tool. And I started thinking about different eras of marketing.
Early era — blogging and SEO matured and HubSpot became a huge beneficiary. Then when Facebook launched its ad marketplace, around 2010, paid ads across Google and Facebook became the dominant way to market. Then there was one other I thought I was late to even though it was right in front of my face: what I call algojacking.
Two people told me about it specifically to my face — they told me the secret — and I was too dumb to acknowledge it. The first was actually Steve Bartlett. I hired Steve Bartlett when he was probably 18 or 19 years old. He had this one skill that was really interesting — he was building social media pages, random themed accounts, and he could grow them very fast. They were run by a single 19 or 21-year-old person and got ten times more engagement than my corporate accounts.
Sam: What’s an example page?
Shaan: He had one like “Freshman Problems.” He’d post something and all the college kids would laugh at the memes. But once you get the attention and trust of every college kid through your freshman problems page, he would then just shove in a product — Spotify, Nike would come and pay him $100,000 to reach that audience through his accounts.
In the same way, Dharmesh — you were a guy blogging in your bedroom and basically outpacing venture-funded startups with CMOs and budgets who couldn’t get a tenth of your traffic. The same thing was happening with Steve. He was literally sleeping in my office at the time, telling me, “Yeah, you’ve got to think about Jenny in her bedroom. She’s scrolling. What’s going to make Jenny care? She’s not going to care that you launched your app.” He started thinking: work backwards. What does the algorithm want? Give it that, then find a way to slip your product in.
And now if you look at Diary of a CEO — one of the most successful podcasts in the world — the reason it’s one of the most successful is because he figured out how to algojack YouTube and Instagram. He worked backwards from the algorithm. What we do, for example, is we have our product and then we try to share it, and it may or may not work. What he did was change his product completely to be what the algorithm wanted. It started as the diary of a CEO, his personal diary. Not that many people wanted that. Then he started interviewing CEOs about CEO stuff. Some business people wanted that. If you go look at his feed today, it’s Seth Rogen, health, sex, conspiracies — basically mass mainstream, mostly in the format of “they’ve lied to you about carbs, they’re lying to you about this.”
But that’s what the algorithm wants. And when I say the algorithm, it’s really just what people click on when they’re bored. Mr. Beast did the same thing. If you ask Mr. Beast what a great video is, he’ll say: one that gets a lot of views. You’re like, “Well, by that definition, is McDonald’s the best burger?” He’d say, “Yeah, it sells the most burgers, so it’s the best burger.” We have different definitions of “best.” But his thing has always been: I work backwards from what’s going to get a view.
Even though I saw that algojacking trend, I didn’t really understand it. Long story short, Dharmesh — today you see a new trend. What do you see as the next thing?
The Death of SEO and the Rise of AEO [00:39:00]
Dharmesh: Yeah. Let’s talk about this because I think there’s real value in some history first.
The thing I was doing back when HubSpot got started was SEO — search engine optimization. There’s an entire industry that grew up around it: how do you increase the odds that your web page will rank in Google organically — meaning you’re not paying ad dollars, it’s free. How do you rise in the rankings?
And pulling on your thread, Shaan — around algojacking — I prefer to call it optimization. What you’re really doing is just saying, “Knowing what Google is looking for, here’s what I’m going to do to have my website rise in the rankings.” And there are layers to the game.
The first layer is more of the hacking: how can I trick the Google algorithm? There were lots of practices that were later labeled “black hat.” Like, Google really likes it when it sees the keyword someone’s searching for on your web page. So people started putting the keyword on the web page 500 times.
Sam: In white font on white background at the bottom of the page.
Dharmesh: Exactly. And then Google got too smart about that, so it would be in tiny font or white on white. It was just a constant game to try and trick the algorithm.
Shaan: Google would do an update every year or so, and then six months later you’d see an article about a company laying off 500 people. I’m friends with the NerdWallet guys — a company that writes articles about the best credit cards. They were doing poorly for years, then after three years Google made an update and they became number one for “best credit card.” Overnight they went from making basically nothing to making $30 million in revenue in that third year. Google rewarded them because they’d done things the right way and all the other sites using hacks got wiped out.
Dharmesh: Exactly. And I went deep down the rabbit hole on SEO. A couple of key insights that I think are still applicable today and made us money:
Thing number one: there are lots of smart people at Google whose sole job it is to deliver high-quality results to the end user. That means every black hat tactic you’re trying may work right now, but you’ve got an army of really smart people whose sole job is to make what you’re doing not work. So it’s just a matter of time.
Thing number two: are there things you could do that would actually make those people’s jobs easier? Google rewards — sometimes intentionally, sometimes unintentionally — what you’d call good behavior. Make your site clean. Make it load fast. Make the content actually useful.
I got up on stages and wrote about SEO all the time, and my one thing I’d tell small business owners was: whatever term you want to rank for, type it into Google right now. Look at who’s coming up. Then ask yourself honestly: do you have a web page on your website that deserves to rank higher than those results? Not in a self-serving way — if you were the actual human typing that keyword in, would you want to find your website over what Google is showing right now? If the answer is no, job one is to create the content that deserves to rank higher.
Because Google is imperfect, and all those battles are winnable. The top two or three might be pretty good, but if you were really searching, here’s what you’d actually want. Produce that page, and nine times out of ten, ninety-five times out of a hundred, it was just a matter of time.
I have content I wrote for HubSpot 19 years ago that still drives traffic to hubspot.com, still generates leads, still generates revenue 19 years later. That’s one of the big things about SEO: when you’re doing Google Ads, you’re effectively renting someone else’s stage. You pay them to put you on the page, but the day you turn that off, the traffic turns off. You’re renting, and you don’t control the rent. It may cost $5 a click now and $15 a click tomorrow based on supply and demand.
The nice thing about SEO and content is it’s like building your own property on your own land that compounds in value over time. As land prices go up, the value of the thing you built goes up, not down. And it’s under your control.
Intensity as the Strategy [00:49:00]
Shaan: One of the painful lessons I’ve learned is this phrase: intensity is the strategy. I’m the type of guy who always liked to read the next blog, the next tweet, the next book because in my head, life was a game of chess — if you just knew the right moves, you’d win. But chess is the wrong analogy for life. It’s more like taking a chess piece and just battering the other guy as many times as you can in different ways. I’ve learned that for most things in life, I already knew the strategy and the difference between success and failure was the level of intensity.
Dieting is the easiest example. We all know what to do.
I bring this up because I can kind of tell — and correct me if I’m wrong — that your volume was a lot higher than what was normal in those days. It wasn’t even that you were the best writer or the smartest at SEO necessarily. I’m guessing your volume was completely abnormal. Was intensity more the strategy than coming up with the perfect playbook?
Dharmesh: Totally. This has been my personal belief my entire life as far back as I can remember. I’m never the smartest person in any given room. You can outthink me — that’s relatively easy to do. You will not out-grind me, because that’s under my control.
I’ve always believed that the quality of outcomes is almost always directly proportional to number of iterations. If you can squeeze out more iterations compared to the next person over the fullness of time, you win. The universe says, “Ah, you win.” You just have to have the conviction, because the feedback loop is slow in the early years. You blog and four people read it, then six, then it goes to five. What the hell? But you have to have the conviction to say, I know it’s just a matter of putting the reps in. They don’t have to be perfect reps — they just have to be incrementally better each time.
Sam: It’s so funny. In April of 2006, you wrote an article saying “Less Is Sometimes Less” — an article about how people say “less is more” and you disagreed. You’ve been saying the same stuff for 20 years.
Did you guys see this video that’s going somewhat viral of the YouTube founders early on?
Shaan: What’s this?
Sam: It’s 20 years ago, Chad and Steve from YouTube just in their office talking. Basically they’re sitting around and one says, “I was kind of depressed last week. We have 40 videos on the site, and if I came to this site I wouldn’t stick around.” And they’re just like, “Yeah, this sucks.” This is YouTube — the founders of YouTube. And you’re talking about early days blogging: “Yeah, I got six visitors one week, eight the next.”
Being able to persist through that time, and then the volume — you don’t have enough videos. It’s not that YouTube doesn’t work, it’s that YouTube with 40 videos doesn’t work. It’s not that blogging doesn’t work, it’s that blogging once every two weeks doesn’t work.
We didn’t close the loop though, Dharmesh — what’s the new wave?
AEO: Answer Engine Optimization [00:57:00]
Dharmesh: So the new wave — and just like SEO created an opportunity for SEO consultants, agencies, and businesses driving direct organic traffic, I think we have a new thing now.
A couple things are happening. Organic traffic that used to come through SEO very predictably — including to folks like HubSpot, who have played the SEO game for a long time — has gone down by about 20 to 40% depending on the industry. The reason is that historically, the practice of search was: go to Google, type something in, get 10 blue links. That’s the whole game.
As it turns out, there’s this new thing called AI and a new app called ChatGPT, which has 800 million weekly active users. What people are doing now is not going to Google and getting 10 blue links — they’re going to ChatGPT and just asking a question, getting an answer, and that’s it. There’s no click-through. Just tell me the answer instead of making me read through 10 articles and figure it out myself.
You can see it in the actual data. And so what’s happening now is this new emerging discipline — in the same way we had search engine optimization, we have what we’re calling AEO: answer engine optimization. Before, people were going to Google to search for your product, your service, your industry. Your job was to be one of those 10 blue links, higher the better. Now it’s: people are looking for products in my category in ChatGPT. How do I show up as an answer?
A few things that are distinctive about it. One: the click-through rate in terms of web traffic to your site through AEO is going to be down because if they just get the answer, there’s no reason to come read your blog post. Also, it lists far fewer options. If you say, “What’s a good clothing company?” you only get like four options, whereas on Google—
Sam: Yeah.
Dharmesh: And which no one ever clicked next on, by the way. There was a running joke in the SEO world: if you ever had to hide a dead body, hide it on page two of Google search results because no one ever went there. They would just rewrite the query — rephrase the question — instead of clicking next.
Sam: But at least I get 10. What I notice when I ask stuff in ChatGPT is it gives me like two or three.
Dharmesh: Right. So it’s even more important. And it’s what engineers and math people would call a binary outcome. Either you show up or you don’t. If you’re not in the actual citations in the answer, you might as well not have played the game. There’s literally zero difference between being number 5 and being number 500 — you get nothing.
Practical AEO Tactics [01:03:00]
Dharmesh: So then the question is: what should people be doing now to start getting that?
Number one: On the off chance you’re blocking unknown crawler bots — some companies do this because crawling costs them server resources — explicitly go in and turn on the ability for the OAI SearchBot and GPTBot to crawl your website. If you don’t get indexed, you’re not even in the game.
Shaan: Okay, that’s thing one for sure.
Dharmesh: Thing two: they do self-report now. When you look at your Google Analytics or HubSpot web traffic, whatever analytics tool you use, it now self-reports when something comes from one of those AI bots. There’s a UTM source parameter. HubSpot has this in the product now — we’ll tell you what percentage of your traffic is coming from AI referrals broadly defined, just like we used to tell you what you were getting from paid versus organic. So take a look at what traffic you’re getting now. As you do things, you’ll know whether you’re getting better or worse. You have to measure it in order to improve it. See what the baseline is today and start tracking it.
Number three: Reframe your content. Either rewrite it or put new content out there that is closer to a structured question-and-answer form. This is less about the kind of prose and narrative writing and hooks, because you’re not trying to hook a human the way you were when writing blog posts. Now you’re solving for the AI crawler, which is trying to find the best possible answer to the question the user is asking.
It’s easier if the question is “what are the top CRMs for small business and startups” and there’s a web page on the HubSpot website that says “here are the primary factors that influence what you’d pick” and “here’s why HubSpot is the top option, with seven reasons.” Make it easy for the AI to get to what it’s trying to find.
Number four: All the things you used to do in classic SEO still work. You still want page authority because ChatGPT specifically — when you type a question — converts that into search queries. It goes and does a bunch of Bing queries. Bing is what it uses, but most search engines are roughly equivalent technology-wise. So all the things you used to do to rank a web page will still help you, because when Bing gets asked by ChatGPT, it comes back with a list of pages, those pages go back to ChatGPT, ChatGPT puts it through the language model, and that becomes the answer shown to the user. So getting high-authority links, showing up in Bing, doing all the classic white-hat SEO stuff — still applies.
Shaan: Sam, are you getting traffic and sales from AI for your stuff?
Sam: I don’t pay attention to it at that level of detail, but we’ve looked at a few of these tools — the way you’d have HubSpot or Moz for SEO, there are a few startups now funded specifically for tracking this and helping you get into the top answer box. We’re trying those now but I think it’s early to see results.
By the way, I just typed in: “I’m a small business owner and I think I need some kind of place to keep track of customers and prospective deals. What should I be using?” HubSpot was the second answer. It said, “Here’s what a CRM does, here are the top CRMs for small business 2025.” But it looks like in this case the ranking is just pulled straight from a TechRadar article. Like, TechRadar had some article saying they tested 13 of the best CRMs in 2025, and it just pulled that exact list and put a comparison table together and gave me that as the answer. Whatever tech writer said becomes the answer.
For Hampton, we’ve gotten a lot of business this way. It’s crazy how much is happening.
Dharmesh: What are people looking for, Sam, that they’re finding Hampton?
Sam: Sometimes it’s just “what’s an entrepreneur group I can look into?” — that’s a major one. But sometimes people will be asking ChatGPT as if it’s an executive coach, saying “I’m struggling with this problem,” and it will say “have you considered an executive coaching group or peer network?” and then we get shown.
What’s crazy is we don’t try. I was shocked when I started seeing leads and customers come in. HubSpot would flag them as coming from ChatGPT. And I remember the first time we got one, about eight or ten months ago — it was a little light-bulb moment. I honestly think the best way to make it work is just to do the same things you’d do to rank on Google, but with the nuances Dharmesh mentioned. Like the fact that on Google you structure your page a certain way because humans are reading it, but ChatGPT loves question and answer. So formatting your content as Q&A is a simple nuance of the new way of doing things.
Dharmesh: Right. And the other thing that doesn’t exist yet — back in the early blogging days, Google had a keyword tool. You could type in a term, it would tell you ballpark how many people are searching for it and how much competition there is. That doesn’t exist for ChatGPT yet.
The other thing that doesn’t exist yet is: early Google would show you in your traffic analytics which keywords people were searching for. Google has since curtailed that and doesn’t pass that granular intent data. But I think things will evolve.
More Tactical AEO Tips: Structured Data, Reviews, and Reddit [01:15:00]
Dharmesh: A couple more tactical tips that I think are applicable today.
For e-commerce, Shaan — this may already be happening automatically on the Shopify platform — one of the things that AI engines really like is structured product catalogs. It’s like, here’s a standard semantic web way of describing the product: size, color, all those things. And another key thing is inventory. Is it in stock? Can it be shipped? What’s the delivery time? If you put those things in the website, that’s going to influence things. If I’m searching for a specific musical synthesizer, I want to find a site that actually has it. That’s why Amazon gets so much love generally.
For you, Sam — the other thing that AEO or answer engines optimize for is more human-based signals. If you look at review sites like G2 for product-oriented companies, it also highly indexes on things like Reddit — especially if there’s active back-and-forth. If there’s even one comment in there that definitively answers a question, and that answer loosely matches what the user is asking, that thread has a very high likelihood of being a cited source.
So: if there’s a Reddit post asking “what’s the best CRM for small business” and someone in the comments definitively says “it’s HubSpot, and here are the five reasons why,” and that comment gets a bunch of upvotes — ChatGPT has access to that data and that has a very, very high likelihood of being cited.
Sam: This is going to be the biggest company in the world. Imagine if Google had a subscription. Google is already like the third or something largest company in the world — now imagine every customer is also paying $10 a month.
Three episodes ago we were talking about OpenAI. If we have to predict the next trillion-dollar company — this was before any company had reached a trillion — it’s crazy how much ChatGPT with 800 million weekly active users is essentially the new operating system for this current generation.
I had a buddy who was deciding between two very promising companies. One was OpenAI, before ChatGPT launched, when they were already worth $30 billion. He was like, “I don’t know if this equity can grow anymore.” He ended up taking the job anyway and has already made something like $30 million on his stock. He sold a portion so he could be comfortable, and then we were talking about what to do with the rest, and I said, “There’s definitely a world where it could 10 or 20x or more again.” And it’s just crazy to say out loud — what are they worth now?
Dharmesh: $500 billion is the current publicly disclosed round that’s happening.
Rapid Fire: How Dharmesh Gets Ideas [01:24:00]
Sam: Can we do some rapid-fire questions? We have this document where we were preparing different ideas, and Shaan said he wants to know your process from zero to one — how do you get ideas, do you do research, how do you move the ball forward? And you replied with something interesting. You said two things: “This gives me a good chance to talk about algebra,” and “It’s all about iteration.” Can you elaborate on that?
Dharmesh: Yeah. A couple of things. One is I don’t do deep research on ideas. I don’t look at the market like an MBA would. Most of the things I end up pursuing are either personal itches — I need the thing myself, or someone close to me needs it. Either my business needs it, my wife needs it. And these tend to end up being software, because that’s the only thing I know how to produce. If I were a carpenter or a musician, it would be something different.
My running thesis — and this is the algebra reference — is around mathematical induction. Mathematical induction is this simple principle that says: as long as it’s true for n equals one, and it’s also true that if it’s true for n it’s also true for n plus one, then it’s going to be true for everything, all the way to infinity. Distilled down: if you can start (condition one) and you can always take the next step and make the next iteration, it will work. You just have to keep at it.
I’m not saying stubbornly pursue the exact same idea you started with — that doesn’t work. An iteration is truly taking market feedback and doing a true loop that says, “I’m going to try to do it better the next time.” But that’s my approach: find an idea, get started, and then ask, can I write code tonight?
Right now I’ve got three projects in flight — these scratch-my-own-itch AI projects — and I’m almost tempted to tell you what the three are because it’ll force me to launch them before this episode goes live.
Shaan: Yeah, give a one-liner on one or two of them.
The Image Generation Agent: imagegen.ai [01:29:00]
Dharmesh: One is a visual designer as an agent. There are lots of great image generation tools — ChatGPT has one, there’s Imagen from Google. But what I’ve always found lacking is the actual process I go through as a non-designer, as a business person.
If I’m doing a social media post or a blog post, I have a process: I want to come up with visual ideas, I want to see what would represent this, I want to see 10 possible examples, I want to iterate. Then I want to pick five and create my own personal brand style and apply that style to the content. That’s a whole workflow, and it’s painful and I’m too impatient. I have access to designers, but it’s not about saving money — it’s that if I have an idea at 2:00 in the morning and I want to get it out there with a visual that goes along with it.
That’s what I’m building. It goes through my entire flow as an agentic process.
Shaan: Do you have a name for it?
Dharmesh: It’s called imagegen.ai.
Sam: So generation.ai?
Dharmesh: Yes.
Sam: Is this going to be — you know, one of the first or second times you were on MFM, Wordle was popular and you said you made a Wordle alternative. And your side project was making $80,000 a month at the time. Do you have ambitions or guesses as to where imagegen.ai is going to be in six months?
Dharmesh: I don’t know, actually. Part of this — we talked about this a little bit in past episodes — is that as part of my HubSpot role, I’m now working on the agent.ai platform, which is basically a marketplace where you can build agents, find agents, hire agents to do things. So in the back of my mind, I’m building this as an agent that will be on the agent.ai platform.
The hope is that if the utility is high enough, it will either cause a bunch of users to be added to agent.ai — which is now up to 2 million users — or it monetizes on the platform. For me it passes a litmus test: if this thing already existed, would I pay $5 or $10 a month for it? The answer is yes.
I actually posted on LinkedIn yesterday the product roadmap for this thing without disclosing the name. Like, here’s what I’m building, here’s what it’s going to do. Simple things: right now when you do an image generation with text, the image models have gotten better at text but they’re still not perfect. So one thing I wanted to do was pass the generated image back through AI and say, “Tell me the text on this image.” If there’s a typo, just rerun it — don’t waste my time giving me things with typos. By the time it gets to me, it’s already done. It’s like having an intern doing this in the background, and then I show up and it’s got 10 options, in my style, without typos. Sounds like a minor thing, but it’s useful.
Shaan: Sam, do you want to ask him some of these other questions? You had some pretty good ones — like the request for startups, or his ChatGPT history.
Sam: Yeah. I’ll do the ChatGPT history one. I’m just curious — you run a huge company, you’re a scratch-the-itch builder, you’re an investor, you have many hats. If you pulled up your ChatGPT and looked at the last five to seven conversations, what are you talking to ChatGPT about?
How Dharmesh Actually Uses ChatGPT [01:36:00]
Dharmesh: I’m a huge ChatGPT fan. But the way I consume it is not through the app. I consume it through the API. I have my own command-line thing I put in. And there’s a reason I do this.
The next big advancement we’re going to see in AI apps generally is memory — they remember things. And ChatGPT is the best out there in terms of retaining memory, which is awesome. The non-awesome part is that memory is locked into ChatGPT. If you ever go to Claude or Gemini, those things don’t have that memory of you. That’s a very frustrating thing, and it’s great for OpenAI because it makes it very sticky.
Shaan: So your personal version that you’re running in your command line has better memory than ChatGPT. How does that work?
Dharmesh: Because it has infinite memory, and that memory is under my control. For instance, I can go through my email archive — I have two and a half million emails in my “memory” — and I can say, “Here’s the thing I’m trying to get to,” and then combine that with other things. I can take that memory and pass it to Google Gemini, which does video analysis a lot better than ChatGPT, which doesn’t do video analysis yet. It’s the openness of it. I can do more things with it because I can intersect it with other apps.
Sam: But why isn’t this the product? I think everybody would want this.
Dharmesh: Because I have three rules of business. Rule number one: never own a piece of a restaurant. Rule number two: never be a landlord. Rule number three: never compete with Sam Altman from OpenAI. This would violate rule three. I don’t want to create a ChatGPT competitor.
But this will manifest: HubSpot is building memory into our Breeze AI product. And I think over time the industry will evolve where there’ll be some form of federated memory — similar to how we plug into Google Docs and Gmail today. Right now someone needed to do the memory, and OpenAI did the right thing. It wasn’t diabolically planned — it’s just a better product if it has memory, so they added memory.
Sam: Okay, but what do you actually use it for? What are the last few things you’ve used it for?
Dharmesh: I use it a lot for research — if I’m researching ideas. And I use it a lot for domain names, as it turns out. I have a thing that can look at Google Trends, so I can say I’ve come across this idea for a domain name, and I can look it up in the marketplace right now. I can look up the transaction history — has this domain been sold before, what was the last purchase price — like Zillow, but for domains.
One of the agents I’m also working on that will launch this week: I want to be able to say, “Vibe coding is picking up — what is that? I want to know who coined the term, what it is, what domains are available, what’s the price, are any of them free and clear.” Vibe coding, for instance — I was early but late enough that someone had already purchased the domain.
Shaan: By the way, vibe coding would fit your inbound framework. If I’m Replit or one of these companies right now, I’m creating the vibe coding movement — the community, the events, the blog — and then I have the tool underneath it that you could use but could use many others. Putting it behind that the way you did with inbound marketing seems smart right now.
Dharmesh: I’ll tell you one of my ideas — and I’ll sacrifice it, not for free. I paid six figures for a domain that I think captures this.
Shaan: Vibecoding.com? What did you get?
Dharmesh: Vibecoding.com.
Sam: Oh my god.
Dharmesh: And here’s the idea. I’ve logged hundreds of hours doing what some would call vibe coding. There’s another term — “agentic coding” — which I think is more accurate, but it doesn’t have the same vibe as vibe coding.
Sam: That doesn’t exactly roll off the tongue.
Dharmesh: It doesn’t. And just in case, I own that one too. But the difference here, and I would bet a lot of money on this: almost everyone who is not a developer by trade will hit a wall when they try to vibe-code something real.
There’s a class of problems vibe coding is really really good at: if you’re building software for yourself, if it has a short lifespan, if it’s disposable, or if you’re just building it for your team of three people and if it stops working, no big deal. But as you start violating those criteria — if you want to build a production product, a long-lasting software company, a product you’re going to maintain for years — vibe coding will invariably paint you into a corner. You’ll hit some bug, something will break, and it’s just a matter of time based on how ambitious the thing is you’re doing.
When that happens — not if, when — you might actually have an economically viable idea but an unviable codebase because you have no idea what’s happening inside that black box.
So: what if there were a community and a marketplace called vibecoding.com where a bunch of engineers hang out, and what they do is say, “Hey, I’ll answer minor questions for you, but also for $500 an hour I’ll jump into your GitHub repo, figure out your problem, fix it, write unit tests for you, and help you get out of this corner you’ve painted yourself into.” I think that’s going to be a massive market.
Sam: You probably own between 500 and 1,000 domains.
Dharmesh: It’s between 500 and 1,000. Yeah. You just have an idea and you buy it. But what’s crazy is a lot of these domains actually have projects on them. I’m looking at things like pitchcraft.com, opengraph.com, prompt.com, dadjoke.ai, speedround.com — that’s an investment firm, a check or a no in 24 hours. Dad joke.ai — a dad joke generator. It works but not that well.
You said you have three other projects you want to launch but you need this podcast to go out first so I’m setting this up. Shaan came up with the idea of calling you “generative.” This is the most generative I’ve ever seen someone. How do you have enough time to do all this?
The Domain Portfolio and “Living in the Matrix” [01:50:00]
Dharmesh: Just to be clear, the list of domains I put in — I don’t have sites launched on most of them right now. What I was trying to share, and may just put publicly, is: here are my top 20 domains with ideas, but I don’t have the calories to actually pursue them. Make me an offer, but you have to have some street cred — some evidence that you’d actually be able to pursue this idea — and I’ll contribute the domain and maybe some capital to kick it off.
To answer your question directly, Sam: my hours have not changed in the 30-plus years I’ve been doing this as an entrepreneur. I still think of myself as an entrepreneur despite being part of a $25 to $30 billion company.
I mean this in the most positive way I can put it: I live in the matrix. My window into the world is these little talking-character people that sometimes show up on a screen in my office, and I’m just trying to play the game. In the same way people play video games obsessively — I did play video games a lot back when I was younger — this is my version of a video game. And I enjoy it.
If I were a writer, my favorite writers are the ones who spend a lot of time on world-building and character-building — especially the sci-fi and fantasy writers who create an entire universe with its own physical laws and species and powers. They create this entire world that could exist, fabricated out of thin air, in words. I’m trying, in my own small way, to create a version of the universe that solves all of the problems that I have or that the people I love have. I’m kind of poking around at the edges and making a version of the universe that’s a V1.1 of the universe that exists.
Sam: Does your house work like that in your physical life? Is your home life set up in your very particular way?
Dharmesh: No. And I’m probably drawing the wrong visual picture. My life, if you just looked at how I live, is more like the absent-minded professor. My desk is completely messy. It’s not a clean-desk, white-lab-coat, everything-pristine situation. It’s the exact opposite. I’m analytical in some ways, but I actually play things very loose. I’ll have a random idea, drop what I’m working on, go play with something else for a day or two. Right now dropping imagegen.ai for the next few days — and then I’ll finish it and get it to some semblance of working.
I’ve said this before but it bears repeating: if a musician practices 12 to 16 hours a day, or if an athlete is off practicing eight hours a day between games, we never question that. We’re like, of course, that’s what it takes to be top of that game. For some reason there’s a stigma attached to work. Especially if you happen to make money at the thing you’re doing.
Shaan: Yeah.
Dharmesh: I’m not doing this for money at all. I don’t care about money at all. I do enjoy the craft and the practice and the actual process of doing it.
Sam: You know the famous clause that Michael Jordan put in his contract with the Bulls? They had all these rules like you can’t ride motorcycles, you can’t go play random pickup basketball games on concrete. And he had this clause called the “love of the game” clause — Jordan was allowed to play basketball anytime he wanted, for the love of the game.
And I think that’s a clause we all just want in our own mental contract of how we do things. You had this great tweet: “If money doesn’t let you do the things you love doing, why else would you want it?” That’s so good, because I know I have to talk to my mom or my sister and I’ll tell them about my buddy Furqan — he was my co-founder, and then the company he built before that, AppLovin, is now like a $100 billion company. And they’re like, “What’s Furqan up to?” I say, “Oh yeah, he built this lab called Founders Inc. at Fort Mason and he’s in there till 11 p.m., grinding, new startup.” And they say, “Oh my god, if I had the money, I would never do that.”
Part of it is: he’s not doing it for the money. He’s doing it because he wants to. That’s why he got the money. The money came because he is this way — he didn’t become this way because of the money.
Dharmesh: And also whether it’s athletes or musicians, we find it noble when they do it. But if you’re a business person, you’re greedy or you’re a workaholic.
Sam: When my mom says that, you’re like, “Have a seat, Mom, pull up a chair.” You see LeBron James — are you familiar with Jordan’s contract, subsection 3C, the love of the game clause? She’s like, “Oh, I just wanted to know if he has kids yet or not.” She’s operating at a completely different level.
Closing: Dance Like No One’s Watching, Build Like Everyone’s Waiting [02:01:00]
Sam: Dharmesh, we love hanging out with you. I always leave better than I came in. I wish we could do this for many more hours — there are so many different things about life and business we can learn from you. We had Howard Marks on recently. At the end it was the same thing: you came to talk about investing but you’re really just talking about how to live a good life. You do the same thing.
Shaan: It’s kind of funny, right? The legendary investors bring this philosophical energy — patience, observe the world, stay calm. And then Dharmesh and the builders have a slightly different energy. Still philosophical, but it’s a “let’s make it happen” energy. Let’s build it. Why not? We could do this. Let’s try. It’s okay if it doesn’t work. It’s a totally different, playful builder energy that I think is dope to be around.
Dharmesh: Yeah. I have a tweet I almost sent yesterday, but I wanted to get my image generator working so I could attach it. It’s two lines — a nice warm closing: Dance like no one’s watching. Build like everyone’s waiting.
Sam: We have to do a quick plug. One of the ways this episode came about was the marketing team at HubSpot was like, “We need you to promote this thing.” And we were like, “Yeah, let’s just have Dharmesh on the pod and maybe he’ll plug it because that’s way better.” What is it? Loop marketing.
Dharmesh: Loop marketing. By the way, loop marketing is the new inbound marketing.
Sam: That’s what it is. Is that your prediction?
Dharmesh: To learn about how to do marketing well in the modern age of AI, look up loop marketing. Loopmarketing.com. You own that one too, I take it?
Sam: Yeah. We promised we would say it seven times. Loop marketing, loop marketing, and loop marketing. Quota hit. Let’s go.
Shaan: They were like, “Could you guys read this ad or talk about the white font on white background at the end of the pod?”
Sam: You didn’t even plug it, by the way. We asked what the new inbound marketing is.
Dharmesh: You called it AEO.
Shaan: That’s true. It’s true.
Sam: Thank you. That’s a pod.