Mar 30, 2023 · 1h 17m · my-first-million
Brainstorming ChatGPT Business Ideas With A Billionaire | ft. Dharmesh Shah (#438)
Dharmesh Shah · 34m spoken Shaan Puri · 25m spoken Sam Parr · 8m spoken AI Kanye West · 10s spoken HubSpot Ad Actor · 10s spoken
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of My First Million, HubSpot co-founder Dharmesh Shah joins Sean Puri and Sam Parr to explore how generative AI is transforming software, business models, and consumer interaction. Through deep dives into declarative chat interfaces, vector embeddings, and real-world AI prototypes, Dharmesh provides a strategic roadmap for entrepreneurs building in the emerging AI landscape.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 49.1% of the talking time here. How this is scored →
The hosts as informed peer 4.0 Guest teaching 4.9 Guest disagreement 0.8 The hosts pushing back 1.4
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 23:15 Dharmesh brushes off AI doomsday prep as sci-fi fiction
Dharmesh dismisses the existential threat warnings raised by Sam Altman and Elon Musk, framing doomsday fears as entertaining sci-fi plots rather than imminent reality.
Hardest push from the hosts ▶ 20:57 Shaan challenges Dharmesh on catastrophic AI risksShaan directly pushes back against Dharmesh's optimistic tool framing, pointing out that top AI researchers build survival bunkers and test worst-case bio-threat scenarios.
Biggest teaching moment ▶ 32:20 Dharmesh delivers a vector embeddings masterclassDharmesh systematically explains multi-dimensional geometry and semantic distance from the ground up, illustrating how modern vector databases transform search beyond crude keyword matching.
The host holds their own ▶ 1:02:00 Shaan breaks down founder psychology and focusShaan provides deep strategic clarity by recalling Sam's disciplined decision during The Hustle to resist platform video hype, proving why knowing one's operational nature beats chasing every tech trend.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| A Paradigm Shift Bigger Than Mobile | 6 | 3 | 1 | 2 | Shaan breaks down the backstory of OpenAI's transition from non-profit to capped-profit structure following Elon Musk's departure, demonstrating solid background industry knowledge. Dharmesh agrees with the general assessment while admitting he lacks insider gossip. | |
| Experimenting With AI Rap and Kanye Clones | 5 | 1 | 0 | 0 | Shaan demonstrates practical tinkering by explaining how he strung together ChatGPT, Google Colab, and a Kanye voice model to create an AI rap song. Dharmesh listens along pleasantly. | |
| Text-to-Code and Declarative Chat UX | 3 | 7 | 1 | 1 | Dharmesh educates the hosts on the difference between imperative UI and declarative natural language chat UX, pointing out that text-to-code is the most underestimated AI capability. The hosts listen and ask clarifying questions. | |
| Sean's Ten-Minute Website Coding Breakthrough | 4 | 2 | 0 | 0 | Shaan describes his firsthand experience building a website via ChatGPT prompts and having it troubleshoot Netlify hosting errors in real time. Dharmesh confirms how memory and context enable this workflow. | |
| Conversational Context and Reasoning Engines Explained | 3 | 8 | 2 | 2 | Shaan asks if LLMs are just fancy autocomplete, and Dharmesh directly corrects this common framing, explaining that models like GPT-4 function fundamentally as reasoning engines rather than simple next-token probabilistic predictors. | |
| Debating AI Existential Risk and Human Augmentation | 6 | 4 | 3 | 6 | Shaan pushes Dharmesh hard on his techno-optimism, citing concerns from Sam Altman, Elon Musk, and AI red teams regarding existential risks and bio-threats. Dharmesh dismisses extreme doomsday scenarios as resembling sci-fi movie plots. | |
| Sequoia AI Event and the ChatGPT Plugins Ecosystem | 2 | 7 | 0 | 1 | Dharmesh gives an insider recap of the exclusive Sequoia AI event, explaining why Sam Altman's announcement of ChatGPT Plugins serves as an App Store moment for natural language ecosystems. | |
| Demystifying Vector Embeddings and High-Dimensional Space | 1 | 8 | 1 | 0 | Dharmesh takes the floor to deliver a masterclass explaining high-dimensional geometry, vector embeddings, and semantic distance, comparing the shift away from keyword matching to Pandora's algorithmic recommendation model. | |
| Practical Applications of Semantic Distance in Communities | 5 | 5 | 1 | 3 | Dharmesh sketches a concrete product idea applying vector embeddings to Sam's community Hampton. Sam challenges whether the algorithm can detect dishonest user self-reporting, prompting Dharmesh to explain AI evaluation and pattern detection. | |
| HubSpot Platform Sponsorship Break | 3 | 6 | 0 | 1 | Following a brief scripted sponsorship clip, Shaan connects vector embeddings to current venture funding trends around Pinecone and LangChain. Dharmesh explains how LangChain connects multiple LLM steps together. | |
| AI-Powered Travel Agents and Buying Chat.com | 4 | 4 | 0 | 1 | Shaan and Dharmesh paint a picture of next-generation conversational travel planning through plugins. Dharmesh casually drops the bombshell that he personally bought Chat.com for over ten million dollars. | |
| Prototyping ChatSpot.ai and Rapid Execution | 3 | 5 | 0 | 1 | Dharmesh explains how he rapidly solo-prototyped ChatSpot.ai by coding against OpenAI's API before handing it off to a core team at HubSpot to productize across their CRM suite. | |
| Deconstructing Dharmesh's Entrepreneurial Mindset and Playbook | 6 | 1 | 0 | 0 | Shaan and Sam deliver a detailed tribute to Dharmesh's unique entrepreneurial formula: boundless childlike enthusiasm, aggressive tinkering, raw content delivery, and the financial guts to take massive asymmetric bets. | |
| Seventeen Years of Conviction in Natural Language | 2 | 8 | 0 | 0 | Dharmesh tells the story of pursuing natural language business interfaces for 17 years across failed predecessors like InGenioMail and GrowthBot until modern LLMs made his long-held thesis technically viable. | |
| Designing Career Freedom and Integrating AI into Hampton | 3 | 6 | 2 | 2 | Sam asks why Dharmesh doesn't leave HubSpot to do AI full-time; Dharmesh explains how he has structured zero management duties and total autonomy. He then advises Sam on weaving vector embeddings directly into Hampton. | |
| Sean's Advice on Focus Versus Chasing AI Trends | 7 | 2 | 0 | 4 | Shaan gives counter-advice to Sam, citing Sam's historical refusal to chase mobile video hype during The Hustle as a winning playbook, warning him against becoming an AI tourist when focused execution is his strength. | |
| Optimizing Hampton's Application Funnel | 5 | 5 | 1 | 1 | Dharmesh provides immediate tactical conversion-rate optimization feedback on Hampton's onboarding form. Shaan then details his own multi-step Whisper and ChatGPT prompt workflow for extracting startup ideas from podcasts. | |
| The Rise of Prompt Engineering and Prompt.com | 4 | 6 | 2 | 1 | Dharmesh explains prompt engineering as a high-value software engineering discipline and reveals he bought Prompt.com for seven figures, while advising founders against building superficial grifter wrappers. |
Statements from this episode (22)
Shah: OpenAI is the startup most likely to reach $1T valuation
“It, it'd be up there in the top three. I honestly can't think of who else would rank higher in terms of probability of getting to that.”
Shah: Building LLMs is uniquely capital-intensive for software
“Building large language models as OpenAI is doing is, is like supremely capital intensive, which is rare for a Software company, which is what they are.”
Shah: AI shifts software UX from imperative steps to declarative outcomes
“An imperative model is you give step-by-step instructions that says do this and then do this and then do this and then do this. And then I get the thing. What natural language allows us to do is use, ah, what developers would call a declarative model. Instead …”
Puri: Natural language will replace coding and tool-specific skills
“That's the new requirement. Not do you know how to code? Not do you know how to use HubSpot? Not do you know how to run a SQL query? It's do you know English? And actually, honestly, the English thing is also going to go away. It's do you know any language? Do…”
Shah: GPT-3 and GPT-4 are reasoning engines, not knowledge bases
“GPT three and now four is, is a reasoning engine. And Sam Altman has talked about this. It's not a knowledge base where it's like, and so people kind of latch on to this fact that, oh, the data that it has is from September, 20, 21, then I'm going to teach you…”
Shah: LLM capabilities cannot be explained by simple next-token prediction
“It's able to do things. That are not explainable by a simple probabilistic model of auto-suggesting next character, next word, next token, next sentence, right? Like it's gone well beyond that.”
Shah: Prompt-driven text-to-video with director styles will arrive in 2023
“I would not be surprised, let's say by the end of this year, that we have a reasonable way to kind of describe in textual form what we want, who the characters are, what the scene is what kind of stylistic attributes we want. We can point it to, oh, I want thi…”
Shah: AI is an order of magnitude bigger than mobile
“I think it's the single largest opportunity and biggest kind of a tech paradigm shift we've seen since the internet originally came out. Like mobile was big but there was a discrete set of use cases. Like when you put a camera on a phone, when you put a GPS de…”
Shah: Replacing keyword search with vector embeddings is a massive AI opportunity
“One of the biggest opportunities in AI right now is to do what Pandora did. Okay. Is there an industry where right now we're doing really stupid keyword based matching somehow it was very, very crude. If I can take that same dataset and convert it to vector em…”
Shah: Average developers can build vector embedding models in a single weekend
“And by the way, the technology exists today that mere mortals in a weekend can actually build a vector embedding model of a given data set. It's not that hard. I mean, it's not like rocket science-y hard.”
Shah: AI pattern matching will quickly uncover inauthentic user posts
“I would bet you money with a large enough sample size the inauthentic posts would be uncovered by the AI like relatively quickly. Like the pattern matching would say, You know, this actually doesn't occur in real life all that often.”
Shah: I personally bought the Chat.com domain for eight figures
“I did. As of Transfer the domain yesterday, last night.”
Shah: HubSpot is bringing his ChatSpot.ai side project in-house
“And so, but now given the relevance to HubSpot, we're going to transfer that project, chat spot.ai to be a HubSpot staffed core team.”
Shah: Losing millions on Chat.com is worth the price of AI credibility
“Chat.com for let's say I even break even, let's say I lose a few million dollars. It is worth the price of admission for me just because that pays the cover charge.”
Shah: I spent $500k on ChatSpot and will transfer it for $1
“So I spent like half a million dollars. Plus like freelance developers and open AI license fees and all the things that need to go into launching a product like that. And I'll end up giving it to HubSpot for a dollar, right?”
Shah: As HubSpot CTO, I have zero direct reports and no one-on-ones
“I've kind of crafted a role for myself within HubSpot that allows me to do Exactly the things that I want to do and not do any of the things I don't want to do, which is one-on-one meetings. I don't have to manage people. I have no direct reports. I've never f…”
Shah: Sam Parr's Hampton could be a $100M revenue business worth $1B+
“Hampton's a cool idea with actual utility. And Sean, you said this in the last thing, like this Could be a hundred million dollar business worth anywhere from three hundred million dollars to a billion plus dollars. And I think you're right.”
Puri: Every major OpenAI release wipes out a wave of AI startups
“There's just swarms that they're all going to get just like wiped out. Every GPT release wipes out a whole wave of like, even the successful ones. Cause it's like, oh, now that's just a feature of chat GPT.”
Shah: I also bought the Prompt.com domain for seven figures
“But this one, it's not eight figures, it's seven figures. And the domain is prompt.com.”
Shah: Balaji Srinivasan has top-five raw intellectual horsepower
“He's like quite nearly one of the top five people I've ever encountered, like even on the internet in terms of raw, you know, what I call wattage, just raw horsepower. And he, he's like an AI unto himself, right? Like just the knowledge that he has.”
Shah: The odds of Balaji's $1M Bitcoin bet hitting are very low
“If I were a betting person, I would not bet that the odds are what do you think they are? Could happen, but nowhere near the probability that that he's suggesting.”
Shah: Short-term AI 'tourists' are missing out on much larger opportunities
“There's enough real problems to solve where real money can be made. And yes, this technology can now be used in creative ways by lots of people and you should use those. But don't use as an excuse just to kind of be like an AI tourist that comes through, makes…”