Tactic
Build the conference around the category, not your company
HubSpot's Inbound conference was deliberately never named after HubSpot and banned anyone from pitching their company on stage, positioning it as a community-owned movement around inbound marketing. They even refused to trademark the term so everyone would adopt it, which let direct competitors send their whole teams.
“It was just all of us internally at the time. And so we sort of kept that up. That's why the conference was never named HubSpot. You know, this is the 17th year that we've been, yeah, doing that event on an annual basis. And we, it was a convening of the community, not a tech conference, the classic sense where you see the CEOs get up there and pitch their book and like, here's what we're doing.”
Steal thisWhen launching a category, host the event under the movement's name, ban self-promotion on stage, and don't trademark the term so the whole industry spreads it for you.
Tactic
Build the conference around the category, not your company
HubSpot's Inbound conference was deliberately never named after HubSpot and banned anyone from pitching their company on stage, positioning it as a community-owned movement around inbound marketing. They even refused to trademark the term so everyone would adopt it, which let direct competitors send their whole teams.
“It was just all of us internally at the time. And so we sort of kept that up. That's why the conference was never named HubSpot. You know, this is the 17th year that we've been, yeah, doing that event on an annual basis. And we, it was a convening of the community, not a tech conference, the classic sense where you see the CEOs get up there and pitch their book and like, here's what we're doing.”
Steal thisWhen launching a category, host the event under the movement's name, ban self-promotion on stage, and don't trademark the term so the whole industry spreads it for you.
Framework
Make inefficient markets efficient
Dharmesh's 30-year favorite category of ideas: find markets where transactions that should happen don't, because buyer and seller don't know each other, can't agree on price, or lack trust. Removing that inefficiency lets you capture the latent economic value, which is how eBay, Google, and communities make money.
“And if you look across kind of the span of history, some of the most money that's ever been made are when someone took an efficient market to which there was what I think of as like latent economic value that was just being wasted as a result of these transactions not happening. They essentially took some of that inefficiency out.”
Steal thisHunt for transactions that should happen but don't because of a connection, pricing, or trust gap; build the thing that closes that gap and keep a slice of the unlocked value.
Take
SEO content is land you own; paid ads is rent you don't control
Dharmesh argues paid advertising is renting someone else's stage: turn it off and the traffic dies, and you don't control the price. SEO content is building your own property that compounds over time, where competitors can't take your traffic overnight and you see them coming.
“The issue with like Google advertising, or any kind of advertising is you're effectively renting someone else's stage or renting someone else's microphone and you pay them to do that, right? It's like, oh, put me on the page and I will pay you X dollars. And there's a nice efficient auctioning mechanism for the price. But it's like, yeah, but the day you turn that off, the traffic turns off.”
Take
SEO content is land you own; paid ads is rent you don't control
Dharmesh argues paid advertising is renting someone else's stage: turn it off and the traffic dies, and you don't control the price. SEO content is building your own property that compounds over time, where competitors can't take your traffic overnight and you see them coming.
“The issue with like Google advertising, or any kind of advertising is you're effectively renting someone else's stage or renting someone else's microphone and you pay them to do that, right? It's like, oh, put me on the page and I will pay you X dollars. And there's a nice efficient auctioning mechanism for the price. But it's like, yeah, but the day you turn that off, the traffic turns off.”
Take
You can outthink me, but you won't outgrind me
Dharmesh's lifelong belief: he's never the smartest in the room, but quality of outcomes is directly proportional to the number of iterations, which is under his control. Volume and intensity, not cleverness, are what compound into wins over time.
“So you can outthink me. And I think that's relatively easy to do. You will not outgrind me, you will not outwork me because that's under my control. And so I've always believed that the kind of quality of the outcomes is almost directly, always directly proportional to number of iterations.”
Number
AI is cutting SEO organic traffic 20-40%; ChatGPT has 800M weekly users
Dharmesh says organic search traffic that used to come predictably through SEO, including to HubSpot, has dropped 20-40% depending on industry because of AI. ChatGPT, with 800 million weekly active users, now answers directly instead of sending click-throughs to ten blue links.
$800M
ChatGPT weekly active users · users/week
“has gone down literally by about 20 to 40% based on which industry you are in. And the reason it has gone down is that historically the practice of search is that we would go to Google, we would type something in, we would get 10 blue links, right? That's the, 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.”
Number
AI is cutting SEO organic traffic 20-40%; ChatGPT has 800M weekly users
Dharmesh says organic search traffic that used to come predictably through SEO, including to HubSpot, has dropped 20-40% depending on industry because of AI. ChatGPT, with 800 million weekly active users, now answers directly instead of sending click-throughs to ten blue links.
$800M
ChatGPT weekly active users · users/week
“has gone down literally by about 20 to 40% based on which industry you are in. And the reason it has gone down is that historically the practice of search is that we would go to Google, we would type something in, we would get 10 blue links, right? That's the, 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.”
Tactic
Format content as question-and-answer to win AI citations
Dharmesh's tactical AEO tip: stop writing for human hooks and narrative, and instead structure content as direct question-and-answer because the AI crawler is hunting for the best answer to a user's question. Make a page like 'What are the top CRMs for small business' with the answer laid out to reduce friction for the AI.
“reframe your content, either rewrite it or put new content out there that is closer to a structured question-answer form, right? So this is less about the kind of prose and narrative writing and hooks and things like that, because you're not trying to hook a human like you were back when you were writing blog posts”
Steal thisRewrite key pages as explicit question-and-answer blocks that hand the AI the cleanest answer to the exact question your buyers ask.
Tactic
Seed Reddit threads where a top-voted comment names your product
Dharmesh says AI answer engines highly index Reddit, and the threads they cite are the ones where the original post asks a question matching the user's query. If a comment definitively recommends your product with reasons and gets upvotes, it has a very high likelihood of being a cited source in ChatGPT answers.
“you will find that the Reddit threads it picks up are the ones where the original post was asking a question. And that question sort of matches a question the user's asking, like, oh, what's the best, you know, CRM for small business? If there happens to be a Reddit post that loosely kind of translates to the exact same meaning, that Reddit thread post, if there's a comment in there where someone definitively says, yeah, it's HubSpot, and here are the 5 reasons why, and that comment gets a bunch of upvotes”
Steal thisFind or start question-form Reddit threads matching buyer queries and earn an upvoted comment that names your product with concrete reasons.
Take
Dharmesh's three rules of business
Dharmesh's three personal rules: never own a piece of a restaurant, never be a landlord, and never compete with Sam Altman at OpenAI. The third is why he won't build a portable-memory ChatGPT competitor despite wanting one.
“Rule number 1, never own a piece of a restaurant. Never be a shareholder in a restaurant. Rule number 2, never be a landlord. Because I don't even wanna maintain my own properties, let alone properties on behalf of other people. And number 3, never compete with Sam Altman from OpenAI.”
Idea
VibeCoding.com: a marketplace of engineers to rescue stuck vibe coders
Dharmesh's pitch for a community/marketplace at vibecoding.com (a domain he paid six figures for): non-developers will inevitably paint themselves into a corner with vibe-coded production apps that break. Engineers hang out there and, for around $500/hour, jump into your GitHub repo to fix bugs and set up tests.
“what if there were a community in a marketplace called vibecoding.com where a bunch of engineers hang out and what they do is to say, hey, I'll answer questions for you, like minor questions, but also for $500 an hour, I will jump into your GitHub repo, figure out your problem and fix it and cause it and set up unit tests for you and help you get through this corner that you've painted yourself into. So I think that's gonna be a massive market.”
Steal thisBuild a marketplace pairing stuck non-technical vibe coders with engineers who fix the production code they can't, billed hourly.
Fact
Dharmesh built a ChatGPT-like app 2 years before ChatGPT existed
Dharmesh got access to the GPT API in 2020 and built a chat application over it on a Sunday night, giving him the full conversational-AI experience roughly two years before ChatGPT launched publicly.
“And so I built this little chat application that used the API, and so I could have a conversation with it. So I actually built that thing that night. It was a Sunday. I had the full transcript 2 years before ChatGPT came out.”
Framework
RAG: feed the LLM the 5 right documents, not 100,000
Dharmesh explains Retrieval Augmented Generation: instead of training a model on all your private docs, store them in a vector database, semantically search for the handful most relevant to a question, and stuff only those into the context window so the model can answer accurately.
“And what we can do now is when someone asks a question, we can go to the vector store, not the LLM, go to the vector store and say, give me the 5 documents out of the 100,000 that are most likely to answer this question based on the meaning of the question, not keywords, based on the actual meaning of the question. So it's called a semantic search is what the vector store is doing. So it comes back with 5 documents, let's just say. Now as it turns out, 5 documents do fit inside the context window.”
Steal thisDon't try to paste everything into ChatGPT. Put your docs in a vector store and retrieve only the few most relevant chunks per question.
Prediction
Hit
Google should ship Q&A over your own Gmail first
Dharmesh predicts the major AI players will all build personal vector-store Q&A, and that Google should be first because it already has both your email data and a strong model in Gemini 2.5 Pro.
“There is absolutely zero reason why Google Gemini does not let you have a Q&A with your own email account. That's just like insanely stupid, right? Like I'll just go ahead and say it. It's just, it's just, there's something not right with the world when they already have the data and it's like, and they have the algorithm, they have Gemini 2.5 Pro, which is an exceptionally good model, right?”
Framework
Tool calling: tell the LLM to 'pretend' it has tools
Dharmesh explains tool calling: the app puts instructions in the context window telling the model to pretend it has access to tools like internet search or a calculator. When the model asks to use one, the app actually runs it and feeds the result back, so the model effectively gains capabilities with zero architecture change.
“In the instructions that we give it in the context window, we're going to say you have access to these 4 tools and it doesn't actually have access to the 4 tools. It's that I want you to pretend like you have access to these 4 tools. The first tool is this thing called the internet. And the way the internet works is you type in a query and it will give you some things back.”
Framework
You to the power of AI, not you versus AI
Dharmesh's frame for thinking about AI and your career: it's not a competition where AI replaces you, it's an amplifier that multiplies your capability, but only if you actually use it daily and learn its edges.
“The right mental frame of reference you should have, it's you to the power of AI. AI is an amplifier of your capability. It will unlock things and let you do things that you were never able to do before, as a result of which it's going to increase your value, not decrease it, right? But in order for that to be true, you actually have to use it.”
Steal thisBefore any task you sit down to do, give your AI tool a crack at it first.
Framework
Find the frontier of what AI can 'sort of' do
Dharmesh's heuristic for AI bets: the sweet spot is a task the model can almost do, where you have to squint. On an exponential curve, anything it can sort of do today it will do well within 6-12 months.
“The best way to kind of think about AI right now is as you use it, is to kind of truly find a frontier of what it's incapable of. It's like, okay, it can sort of do this thing, but not very well. If that's the way you describe its response, you are exactly where you need to be, which is if it can sort of do it right now, sort of. If you have to squint a little bit, it's like, ah, well, it's kind of something, but, Wait 6 months or a year, right?”
Steal thisBet on the task AI can barely do today; the model will catch up while you build the brand and customer base.
Idea
Build HR for digital workers: onboarding, reviews, recruiting for AI agents
Dharmesh pitches an entire category of opportunities around treating AI agents as teammates: training programs, performance reviews, recruiting, and the new 'agentic manager' role for hybrid human-plus-agent teams, plus the software and books to support it.
“Well, what does employee training look like for digital workers? What do performance reviews look like for digital workers? How do we do recruiting for digital workers? How do we, like, what are all the mechanisms that need to exist? What is a manager of the future? What are the new roles that will be created as a result of having these hybrid teams?”
Steal thisBuild the onboarding, performance-review, and management tooling companies will need to run hybrid human-and-agent teams.
Story
Dharmesh's son stress-tests his fantasy world via ChatGPT role-play
Dharmesh's 14-year-old, an aspiring manga author, wrote a 2,000-word prompt describing his fictional world's characters and power structure, then turns it into a role-playing game with ChatGPT to pressure-test the world he could only previously hold in his head.
“So what he uses ChatGPT for is he's got this like 2,000-word prompt that describes his fictional world. Here are the characters, here's a power structure, here are the powers people have, here's what you can and can't do. And then the way he tests the world is he turns it into a role-playing game.”
Framework
Stop trading labor for value; buy yourself leverage
Dharmesh Shah's core wealth rule: converting 100% of your time into value with no leverage caps your earnings. You must carve out time and money to build leverage that multiplies your output.
“If all you do is spend 100% of your time converting your labor into value and do not increase your leverage, you're not going to get anywhere.”
Steal thisAllocate a fixed slice of your time and money to building leverage, not just billable output.
Take
First half of life: trade time for money. Second half: buy it back
Shah's reframe of a career: you spend the early decades converting time into money, then spend the later decades desperately converting money back into time.
“you're going to end up spending a large part of the early part of your life, let's call it the first half, converting time into money in various shapes and flavors. That's effectively what you're going to be doing. And then you're going to spend the latter half of your life approximately desperately trying to convert money back into time. That's what happens. That's life.”
Take
First half of life: trade time for money. Second half: buy it back
Shah's reframe of a career: you spend the early decades converting time into money, then spend the later decades desperately converting money back into time.
“you're going to end up spending a large part of the early part of your life, let's call it the first half, converting time into money in various shapes and flavors. That's effectively what you're going to be doing. And then you're going to spend the latter half of your life approximately desperately trying to convert money back into time. That's what happens. That's life.”
Framework
If you're not making, selling, or shipping the product, you're overhead
Shah's manager at U.S. Steel told him that anyone not making, shipping, moving, or selling steel was overhead. The lesson: your value is inversely proportional to your distance from where value is actually created, so get closer to the product.
“Dharmesh, look, if you're not making steel shipping steel, transporting steel, moving steel, or selling steel, you're overhead.”
Steal thisMove yourself as close as possible to where the company actually creates value; everything else is overhead.
Framework
If you're not making, selling, or shipping the product, you're overhead
Shah's manager at U.S. Steel told him that anyone not making, shipping, moving, or selling steel was overhead. The lesson: your value is inversely proportional to your distance from where value is actually created, so get closer to the product.
“Dharmesh, look, if you're not making steel shipping steel, transporting steel, moving steel, or selling steel, you're overhead.”
Steal thisMove yourself as close as possible to where the company actually creates value; everything else is overhead.
Tactic
You don't need 100 buyers for your talent, just one
When Shah's boss doubted he could find many companies paying him $200K, Shah's calm reply became his negotiation lesson: he didn't need to find many, just one, and he was mobile enough to go anywhere to find it.
“I really don't need to find that many companies that will pay me $200,000 a year. I just need to find one. And I think there's one out there.”
Steal thisWhen negotiating your worth, remember you only need one buyer at your price, not the whole market.
Framework
Negotiate by finding both sides' real needs, not splitting the pie
Shah's takeaway from 'Getting to Yes': most negotiations aren't zero-sum. Instead of fighting over price, identify what each party actually needs, because there's usually a path that optimizes for both.
“The way to think about it is to identify what your actual needs are. What's truly the thing? Are you, if you're solving for price, fine, but recognize that and do your best to identify what the other party's needs are because there is often, very, very often a path that actually will optimize for both of you.”
Steal thisBefore negotiating price, map both parties' true underlying needs and look for the win-win path.
Tactic
To win a unique-asset negotiation, gush authentically before you bid
For one-of-a-kind assets with no comparables, Shah argues the buyer should genuinely praise what they love about it rather than poke holes, because a seller won't fight a buyer who clearly appreciates the value and seems reasonable.
“Let me tell you the things I love about it. Let me tell you why I am like all in love with this thing that you're selling right now. And you go through that and it has to be authentic, has to be genuine.”
Steal thisWhen buying a unique, comp-less asset, authentically praise it before naming your price to lower the seller's guard.
Take
Writing has the highest return on time of any skill
Shah argues he's never found the ceiling on the return from learning to write well; five or ten hours invested in writing pays off forever because it amplifies thinking, communicating, pitching, and selling.
“the return on time for developing writing— I have not found the ceiling yet— is that it is just so high. I cannot describe to people, like, if you could do nothing else, let's say you had 5 hours to invest in the next month or something like that. You could spend it all on learning how to write well, and future you will look back on those 5 hours and say, boy, that was a great use of 5 hours”
Steal thisInvest your spare learning hours in writing; it compounds into better thinking, selling, and pitching.
Framework
Collect dots: invest in things that only connect looking backward
Riffing on Steve Jobs, Shah says to deliberately spend time collecting 'dots'—people, skills, experiences—that may not make sense yet but feel right. You can't connect them forward, only back, and you only need one or two to pay off.
“you have to spend some amount of your time, and it's the same kind of lesson, just phrased differently. Around investing in those dots that may or may not make sense at the time that you're doing them, but they feel right. They feel like they could have something.”
Steal thisDeliberately collect 'dots'—skills, people, side bets—that feel right even if their payoff isn't visible yet.
Prediction
Hit
Agents are the new apps: millions of them are coming
Shah predicts AI agents will become as ubiquitous as mobile apps—hundreds of thousands to millions of agents—and that companies will run on hybrid human-plus-agent teams, with agents treated as digital team members.
“My expectation is that agents are the new apps. It's just software, right? So when mobile came along, it's like, oh, there's an app for that. There's an app. It's like, not that far in the distant future, everything is going to be, there's going to be hundreds of thousands of millions of agents, right?”
Idea
A LinkedIn-style professional network for AI agents
Shah's positioning for Agent.ai: the first (and currently only) professional network for AI agents, with ratings, reviews, follows, and the ability for agents to discover and hire other agents the way Fiverr and Upwork work for humans.
“Agent.ai is the number one professional network for AI agents. It is also the only professional network for AI agents.”
Steal thisBuild the discovery, ratings, and hiring layer for AI agents the way Upwork did for human freelancers.
Framework
You must climb the local maximum to even see the global one
Shah's non-obvious twist on local vs. global maxima: you often can't spot the bigger mountain from the sidelines. You have to climb the smaller hill first to gain the perspective that reveals where the real big bet is.
“often in order to even see the global maximum, you actually have to climb the local maximum hill. It can't be abstract, sit on the sidelines, it's like, oh, I'm going to go wander around. What you have to do sometimes is you have to make the effort to climb the smaller hill you get to the top of the spot, it's like, oh, now I see the landscape and there's that massive mountain over there”
Prediction
Pending
Results-as-a-Service is the next wave after SaaS
Shah pitches RaaS: instead of selling access to software, you sell the outcome directly. Customers don't touch the software; they tell you the result they want (the holes, not the drill) and you deliver it.
“Results as a service is like, actually, you don't even access the software. You tell us what it is that you actually want to do. What's the outcome that you're looking for? And we'll just sell you that.”
Steal thisPackage and sell the outcome customers want, not access to the software that produces it.
Framework
Imperative vs declarative software: describe the result, not the steps
Dharmesh frames natural-language AI as a shift from imperative software (you click through step-by-step instructions) to declarative software (you describe the outcome you want and the AI fills in every step). It is the difference between managing a junior intern and briefing a senior person.
“What natural language allows us to do is use what developers would call a declarative model. Instead of describing all the steps, describe the result that you want at the end of the thing, and then the software does everything in between.”
Steal thisWhen building AI UX, let users state the end goal and have the software handle the intermediate steps, instead of forcing them to learn your click-paths.
Prediction
Miss
ChatGPT plugins are the App Store moment, bigger than GPT-4
Dharmesh argues that ChatGPT plugins, which let third-party developers inject proprietary data and actions into the chat, turn ChatGPT from an app into an ecosystem, just like the iPhone App Store. He calls it a bigger drop than GPT-4.
“this is like the App Store was, uh, for iPhone, which is, oh, we've got this super popular thing called the iPhone and we have our own apps, which is great. It does these 17 things, but now we're going to let anyone build apps”
Idea
Replace dumb keyword matching with vector embeddings in any industry
Dharmesh's core opportunity: find any industry still doing crude keyword-based matching, convert that dataset to vector embeddings, and let people search by semantic meaning instead of exact words, like 'Google search but smart' for everything.
“one of the biggest opportunities in AI right now is to do what Pandora did to say, okay, is there an industry where right now we're doing really stupid keyword-based matching somehow? It's very, very crude. If I can take that same dataset and convert it to vector embeddings, and allow people to find things in a different way than they've ever been able to do before.”
Steal thisPick an industry stuck on crude keyword search, convert its dataset to vector embeddings, and offer semantic meaning-based matching.
Idea
Match community members by semantic meaning, not industry or size
Dharmesh pitches Sam on vector-embedding Hampton members' opted-in stories so you can find someone facing the same 'founder therapy' struggles rather than matching on shallow facets like industry, company size, or geography. He calls it a billion-dollar idea that recurs across every industry.
“I wanna find someone that's dealing with these kind of founder therapy level issues. Who are those people? Let's find the semantic distance between those vector embeddings across the 1,000, 10,000, 100,000 people that are in Hampton someday. Um, that's a billion-dollar idea. And that billion-dollar idea occurs a billion times across the entire industry.”
Steal thisEmbed members' opted-in life stories and match them on emotional/situational similarity, not demographic facets.
Framework
Solve for the experience, not the transaction
Dharmesh contrasts old web tools that optimized the transaction (cheapest flight, fewest stops) with AI assistants that solve for the whole experience, pulling together flights, hotels near a Michelin restaurant, your spouse's preferences, and follow-ups from last week into one chat interface.
“What's gonna be possible now in this kind of new AI world is instead of solving for the transaction, you solve for the experience.”
Steal thisDesign AI products around the user's end-to-end experience, not the single transaction your category historically optimized.
Story
17 years of conviction: IngeniMail to GrowthBot to ChatSpot
Dharmesh traces a single idea, natural-language control of business software, across 17 years: IngeniMail (pre-HubSpot, email-based), GrowthBot (which got thousands of users but couldn't do the language understanding), and finally ChatSpot once GPT made it possible. The lesson: have the courage of your convictions and keep iterating on an idea you truly believe.
“the moral of the story is have the courage of your convictions. If you truly, truly believe in an idea and you fundamentally think you're right, iterate. Don't just sit, go down your rabbit hole. Tell everybody you can about it. Build products around it. Find other like-minded folks and try to pull on that thread.”
Tactic
Don't pre-filter applicants out of your funnel too early
Reviewing Hampton's application, Dharmesh flags that gating the 'what's your role?' question to founders/CEOs pushes away people who don't fit today but might fit a future product. Let everyone finish and answer; the data costs nothing to keep and could seed a future segment (e.g. VPs of product needing peer therapy).
“you're doing what I call a pre-filter, right? Which is why not say, oh, this is for CEOs, owners, or whatever. Don't make them feel guilty for going through the rest of the process because there may be a future version of Sam and Hampton.”
Steal thisDon't disqualify off-profile applicants mid-funnel; let them finish and bank the data for a future segment.
Take
Prompt engineering is the new software engineering
Dharmesh frames prompt engineering as analogous to software engineering: software engineering gets a computer to do what you want in its language; prompt engineering gets a large language model to produce what you want. He sees it as a major opportunity for technically-minded people who aren't coders.
“that's an entirely new skill now called prompt engineering, right? And it's analogous to software engineering. So software engineering is getting a computer to do what you want by speaking to it in its language.”
Prediction
Hit
Dharmesh on Balaji's $1M Bitcoin bet: nowhere near that likely
Asked about Balaji Srinivasan's warning that the US dollar will crash and Bitcoin will hit $1 million, Dharmesh, while praising Balaji's raw intelligence, says he understands the extreme position as a way to wake people up but would not bet the odds are anywhere near what Balaji suggests.
“if I were a betting person, I would not bet that the odds are what he thinks they are. Could happen, but nowhere near the probability that he's suggesting.”
Take
Don't be an AI tourist: solve real problems, not arbitrage
Dharmesh's closing warning, echoing the crypto/Web3 hype cycle: entrepreneurs chasing quick arbitrage on new tech are grifters. There are enough real problems where real money can be made, so build something that actually brings value rather than slapping AI on a shitty app.
“Don't play the arbitrage. Oh, I'm gonna do this thing. This is like, you know, day trading back in the day or whatever. It's like, you know, don't be a grifter, right?”
Framework
Set an outrageous hourly rate and delegate anything below it
In his early 20s Dharmesh hit $125/hour as an engineer in 1990s Birmingham and decided anything he could pay someone less to do, he was an idiot to do himself. Over a 30-year career at ~60 hours/week, he computes his effective rate at roughly $10,000/hour.
“Let's say I worked on average 60 hours a week, and I've done that for about 30 years, give or take. It's my kind of professional career. My average hourly rate across the 30 years would be like $10,000 an hour, right?”
Steal thisPut a number on your time, then delegate or skip anything worth less than that rate.
Story
Dharmesh has zero direct reports after 15 years and 4,000+ employees
In a co-founder heart-to-heart, Dharmesh and Brian Halligan agreed Brian would be CEO and Dharmesh would have no management responsibilities. Believing in amplifying strengths over fixing weaknesses, he's never had a single direct report despite HubSpot growing past 4,000 people.
“I'm a big believer in take your strengths, whatever they are, and put like all your energy into kind of amplifying those strengths and getting really, really good at that thing. And don't worry about your weaknesses all that much. So I don't want to worry about my weakness managing people. And so we have over 4,000 people at HubSpot. I have zero direct reports.”
Take
The Vegas table: why bootstrappers keep playing instead of cashing out
Dharmesh's metaphor for self-funded entrepreneurship: you're at a gaming table learning the rules as you go, and the cardinal rule is that if you leave the table, the chips you've left are gone. Occasionally the house offers to cash you out, and you never know when the next offer comes.
“there's one cardinal rule, is that if you leave the table, whatever chips you have on the table, they're gone. Every now and then, the house will come up to you and say, hey, would you like to cash in your chips? They're worth X dollars.”
Idea
Dharmesh offers Shaan a $5M check on-air to build a crypto LinkedIn
HubSpot co-founder Dharmesh Shah pitches a blockchain-based professional network to replace LinkedIn and offers to fund Shaan with up to $5 million as a seed round, plus a 7-figure domain name he already owns.
“Here's what I even have a 7-figure domain name picked out for you. I will write the $5 million check for you to go do it as far as a seed round, uh, or up to $5 million if you want to let other people in.”
Steal thisBuild a professional network on the blockchain where every member has a coin price, so endorsements cost real money and carry weighted authority.
Idea
Natural-language layer for B2B software: bigger than mobile
Dharmesh's second big idea: a translation layer that lets users just say what they want ("remove the background", "how many signed up in 90 days") instead of learning clicks and menus. He calls it a megatrend bigger than mobile, with B2B reporting tools the obvious first target.
“If you're in HubSpot, you should, I want, it's like, you know, "How many new people signed up for our Service Hub product in the last 90 days?" You should just be able to type that question in, 'cause that's your intent, not like, "Oh, go to HubSpot's reporting tool, build this dashboard, pull out here the 3 columns you want, whatever, and get to the thing you want."”
Steal thisWrap an existing complex B2B tool in a natural-language layer so users state intent instead of learning the UI.
Framework
Website Grader: build a diagnostic tool that makes people realize they need you
Dharmesh built Website Grader as an internal tool to score prospects' sites, then put it online where tens of thousands used it. The lesson: a free diagnostic for your industry isn't a lite version of your product — it's the thing that makes buyers realize they have a problem.
“the power of building a diagnostic tool for your industry is immense, right? This is the thing that— so Website Grader was not the solution. So it's not like a freemium thing. It's like, oh, we're giving you a lightweight version of HubSpot. It was the thing that made you realize you needed HubSpot, and that was super valuable.”
Steal thisBuild a free diagnostic/grader for your industry that scores the prospect's current state and surfaces the problem your product solves.
Idea
remoteculture.com: a paid community for HR people building remote teams
Dharmesh would turn remoteculture.com into a blog plus paid community targeting people-ops and HR leaders figuring out remote work — not selling software, just connecting them. Sam calls it a no-brainer million-dollar-a-year business.
“I'm going to do a blog, I'm going to do a paid community, and it's going to be targeted at people ops and HR people around the world that need to kind of figure out the new kind of new world order, right? We're not going to try and sell them software. We're just going to connect them to each other.”
Steal thisSpin up a paid community for a role facing a brand-new problem (remote-culture HR), monetizing connection rather than software.
Framework
Dharmesh's idea-scoring formula: profit x passion x probability
Dharmesh scores any idea on three factors — profit potential, passion, and probability of success — each 0–10, multiplied for a score out of 1,000. The common mistake is filtering on probability first, when low-probability ideas with huge upside can have the highest expected value.
“So the 3 things I look at are kind of like profit potential. Like if this thing went exactly as planned, as the founders envisioned, What could it be?”
Steal thisScore ideas as profit x passion x probability (each 0–10) instead of killing them on low odds; weigh expected value, not just likelihood.
Framework
Dharmesh's idea-scoring formula: profit x passion x probability
Dharmesh scores any idea on three factors — profit potential, passion, and probability of success — each 0–10, multiplied for a score out of 1,000. The common mistake is filtering on probability first, when low-probability ideas with huge upside can have the highest expected value.
“So the 3 things I look at are kind of like profit potential. Like if this thing went exactly as planned, as the founders envisioned, What could it be?”
Steal thisScore ideas as profit x passion x probability (each 0–10) instead of killing them on low odds; weigh expected value, not just likelihood.
Number
HubSpot raised $105M in private money before IPO
Dharmesh confirms HubSpot ran the classic venture-backed playbook, raising $105 million in private capital before going public — intentionally, because the founders wanted to swing for the fences.
$105M
Private capital raised before IPO · USD
“$105 million in private money before we went public. Yeah. HubSpot's a classic venture-backed playbook company.”
Story
Be rich, not king: HubSpot's founders chose to swing for the fences
Dharmesh and Brian agreed from day one they didn't want a single or double — at every fork they'd take the path with the higher chance of a spectacular outcome, accepting they might crash and burn. They weren't worried about dilution or control; they wanted to be rich, not king.
“we didn't want a single or double hit, right? We wanted to swing for the fences. And either we do this and we're good at every point that we have a decision to make, a fork in the road, we are going to take the one that gives us a higher chance of being the spectacular outcome, even if that means we're possibly going to go down in crashing, burning flames, right?”
Take
You won't find the great idea until after you start the company
Dharmesh argues the thing founders become known for is usually not their original idea — the great idea emerges only after launching and making contact with customers. Sitting on the sidelines analyzing markets means you never start.
“often the case is that you will not find the great idea until after you start your company, until after you start having contact with customers and actually try to do the thing. If you sit on the sidelines trying to assess ideas and look at market metrics and things like that, and it's like, you're just never going to get started.”
Steal thisStart with a mediocre idea and let customer contact reveal the great one; don't analyze on the sidelines.
Tactic
Headline every episode with the one nugget, not the guest's name
Dharmesh's tactical critique for MFM: stop titling episodes by number and guest, and instead lead with the single most clickable nugget so non-subscribers click when YouTube recommends it (e.g. 'why this successful entrepreneur never played golf').
“you should apply the Sean25 headlines thing for headlines for every podcast episode. Like what's the one nugget that's gonna cause people to wanna listen to that episode even if they're not subscribed when they see it in their YouTube feed?”
Steal thisTitle each piece of content with its single most curiosity-inducing nugget, not the format/guest, to win the click from non-subscribers.
Number
ConvertKit: $28M ARR, raising secondary at 7x ($200M valuation)
Sam bought ConvertKit secondary at 7x ARR. The numbers founder Nathan Barry shared: $28M ARR, $27.3M net revenue up 34%, 3.8% revenue churn, 36,000 customers, ~11-12% profit - implying a ~$200M valuation. Dharmesh (HubSpot) sparked the round.
$28M
ConvertKit ARR · USD/year
“we're putting together a secondary round at 7 times ARR. And we're looking to raise $2.5 million in total with $2.05— so $2 million already committed. ARR is $28 million. Net revenue is $27.3 million, which is up 34%. Revenue churn, 3.8%. Customers, 36,000. Profit this year, 11%, 12% profit.”
Framework
The trillion-dollar Venn diagram of success
Dharmesh Shah argues real value comes from intersecting two or three rare skills. Being top 1% at copywriting AND crypto makes you 100x rarer than either alone, which is where monetizable magic happens.
“What happens though is if you intersect those two circles and turn into a Venn diagram, the number of people that know copywriting really well and also know crypto really well is 100 times smaller than either of those circles combined, right? And this is— we see the same effect like in standardized tests. If you took the SAT or the GRE or the GMAT, there are people that are really good at English portion of the test, there are people that are really good at math portion of the test, but the ones that score really, really well are the ones that are good at both.”
Steal thisStack two or three rare skills you're already top 5% in rather than chasing being the best at one thing.
Framework
The trillion-dollar Venn diagram of success
Dharmesh Shah argues real value comes from intersecting two or three rare skills. Being top 1% at copywriting AND crypto makes you 100x rarer than either alone, which is where monetizable magic happens.
“What happens though is if you intersect those two circles and turn into a Venn diagram, the number of people that know copywriting really well and also know crypto really well is 100 times smaller than either of those circles combined, right? And this is— we see the same effect like in standardized tests. If you took the SAT or the GRE or the GMAT, there are people that are really good at English portion of the test, there are people that are really good at math portion of the test, but the ones that score really, really well are the ones that are good at both.”
Steal thisStack two or three rare skills you're already top 5% in rather than chasing being the best at one thing.
Idea
Netflix for the blockchain: entertainment crossed with crypto
Dharmesh riffs that Shaan could pull off a new Pixar precisely because of an unusual skill overlap: top-tier entertainment instincts intersected with crypto knowledge, combining two things never put together before.
“maybe that's the thing that unlocks the next trillion dollar company is entertainment intersected with crypto. It's the Netflix for the blockchain, whatever it is, right? Something like that where two things combined that never had been put together in that way before.”
Steal thisLook for two domains nobody has combined before and build the bridge between them.
Story
HubSpot's founders paid themselves $5,000/month to anchor all salaries
Brian Halligan and Shah picked $5,000/month salaries for themselves, then paid employees 3 and 4 the same $5,000. The inertia signaled 'if you're here for the salary, you're here for the wrong reason' and kept early comp low.
“And so we picked $5,000 a month, right? It's like, okay, well, that's— it's a number. And obviously below market value. We'd been out in the market for a while and had worked in our lives. So when we hired employee number 3, we had to decide what we were going to pay them. $5,000. Hired employee number 4, $5,000.”
Tactic
Don't make 'press release hires' — hire diamonds in the rough
HubSpot's deliberate early hiring rule: if a hire felt worthy of a press release ('we hired the VP of Sales from BigCo'), they'd done it wrong. Specialized big-company skills rarely transfer to a startup's context; they bet on smart people who get shit done and would be big later.
“and we made a deliberate decision not to hire what we call press release hires. A press release hire is when you hire someone and then you feel compelled to issue a press release, say, oh, HubSpot just hired the VP of sales from such and such company, whatever that is, a 30-year background in doing X. If we felt like we ever hired someone that was worthy of a press release, we did it wrong, right?”
Steal thisHire smart people who 'get shit done' early in their arc, not credentialed names who command a press release.
Number
Shah made 100+ angel investments, deciding most within 24 hours
Shah has made just over 100 angel investments and never even talked to or met the founders on 80–90% of them, deciding within 24 hours on the vast majority.
$100
Angel investments made · deals
“I've made roughly, I think a little over 100 investments now across my angel investment. Um, I won't call it a career, but my activities and 80 to 90%, um, I've never actually talked to the founders or ever met with them. Um, over a vast majority of my deals, I make the decision within 24 hours.”
Story
Shah built Wordplay in 24 hours to teach his 11-year-old that code is real
On a Saturday night Shah coded a Wordle-style game in Python (his son's class language), launched it Sunday by tweeting it, and put Google Analytics on so his son could watch users arrive — making software tactile and possible in the boy's head. It went on to 45M games and 9.5M players.
“So on a Saturday night, I said, okay, well, let me build something. Python happens to be my language. Build something so he can see because he plays Wordle, he knows the game. And so I started on a Saturday night with a deadline for Sunday that I'm gonna launch something with him tomorrow.”
Number
Wordplay could make $90K/month on AdSense with near-zero costs
Shah put Google AdSense on his side-project word game Wordplay and found that with two ads on the main board it would generate about $90,000 a month, with his only costs being hosting and his time.
$90K
Potential AdSense revenue · USD/month
“if I weren't bothered by the fact that there would be 2 ads on the main game board page, if I were solving for monetization, it was like $90,000 a month, uh, in Google AdSense. Wow. Based on the traffic.”
Number
Wordplay valued at $2–10M on a 5x revenue multiple
Shah estimates Wordplay's ad revenue at roughly $50K/month ($600K/year), and applying a 5x multiple pegs a sale price somewhere between $2M and $10M — for a game he built in a weekend.
$5
Revenue multiple applied to value Wordplay · x revenue
“let's say the ad revenue would go down to like $50,000 or something like that. Cause I haven't turned Google AdSense in a while. That's $600,000 a year. Apply a 5x multiple. Um, so somewhere between $2 and $10 million would be my guess”
Idea
Chess.com, but for word-game enthusiasts
Shah frames Wordplay's opportunity as 'Chess.com but for word games' — a ratings system, leaderboards, challenges, and social connections applied to a word-game audience that could rival chess's multi-billion-dollar enthusiast base.
“one of the things is around Chess.com, which you guys talked about, which I've been a member and player on for 10+ years. I knew of the site, but then you think about Wordplay is like, okay, well, this is Chess.com, but for word game enthusiasts versus chess enthusiasts.”
Steal thisTake a proven enthusiast-platform model (ratings, leaderboards, challenges, social) and port it to an adjacent hobby audience.
Take
Market price oscillates around real value — so ignore the swings
After losing over half a billion on paper as HubSpot stock fell ~40-50%, Shah says it doesn't faze him: a company's market price oscillates around its underlying value (revenue, products, customers), and since the business hadn't changed, valuation would self-correct over time.
“the valuation of a company, which is like the market price, um, will oscillate around the value. Right? And value is like the revenues you have, the products that you build, what are you actually doing, um, you know, for your customers.”
Tactic
In a downturn, ask what you can do to capitalize, not just survive
Shah's two-sided crisis playbook: the defensive side is enough cash to weather the storm; the exciting side is asking what you can do to capitalize. In the pandemic HubSpot fronted partners cash for revenue still owed — strengthening itself while helping customers.
“But then the other part, which is to me much more exciting, which is, okay, given what we know, what can we do to capitalize on this crisis, right? This is what's happening right now. What can we do to make this net positive for HubSpot?”
Steal thisIn every downturn, run both plays: batten the hatches on cash AND find the move that makes the crisis net-positive for you.
Framework
The trillion-dollar Venn diagram: stack rare, reinforcing skills
Dharmesh Shah's core mental model: pick a handful of skills and intersect them like a Venn diagram. The intersection of two rare skills (e.g. copywriting + crypto) is ~100x rarer than either alone, and if you can monetize that intersection, magic happens.
“So now imagine, so this is what I call like the trillion-dollar Venn diagram of success, right? I just made that up. But so you draw those circles of your skills. And say, okay, well, I am now not like 1 in a million, I'm 1 in a billion, right? Or 1 in 5 billion, whatever that is. And if you can convert that into some kind of monetizable business thing or whatever, that's when magic sort of happens.”
Steal thisStack two or three rare skills that reinforce each other rather than trying to be #1 at any single one.
Framework
Use functional decomposition to break any scary goal into trivial steps
Shah borrows 'functional decomposition' from software: any big goal (grade a website 0-100, give a talk to 25,000 people) breaks into sub-functions, then sub-sub-functions, until each atomic piece is trivial. The hard part is the abstraction, not the doing.
“But in software engineering, there's this thing called functional decomposition, which is take a problem be able to articulate well. So a function that has a set of inputs and an output and what it's supposed to do.”
Steal thisDecompose any intimidating goal into nested sub-functions until each atomic piece is trivial to execute.
Framework
Output quality is a function of iterations, not the quality of the people
Shah's recipe: dream really big, iterate really small, keep tight feedback loops. He cites the pottery/photography study where the group graded on quantity of attempts produced better work than the group solving for one perfect piece — but iterations only count if each has a real feedback loop.
“I think that others have demonstrated this through data is that the quality of the output is almost— I won't say always, but frequently a function of the number of iterations, not the quality of the people.”
Steal thisOptimize for many meaningful iterations with tight feedback loops, not for one perfect deliverable.