Talking to a billionaire about how he uses ChatGPT - My First Million Recap

Podcast: My First Million

Published: 2025-07-16

Duration: 56 minutes

Guests: Dharmesh Shah

Summary

Dharmesh Shah, co-founder of HubSpot, unpacks how AI tools like ChatGPT are reshaping productivity and decision-making. He delves into practical applications, limitations, and the future of AI as a teammate rather than just a tool.

What Happened

Dharmesh Shah, CTO of HubSpot, shared his early experiences with OpenAI's GPT API and how he built one of the first chat applications with it back in 2020. He explained the concept of the 'context window,' a technical constraint that defines the amount of data AI can process at once, likening it to a sheet of paper with limited space for input and output. This limitation is addressed by innovations like vector embeddings and retrieval-augmented generation, which allow users to query vast amounts of private data efficiently.

Shah detailed how he uses ChatGPT as a personal executive coach by uploading his personal finances, goals, and other data to create a customized AI assistant. He described this tool as a 'PhD intern in everything,' capable of answering highly personalized questions and assisting with strategic decisions. He also highlighted automation use cases, such as simplifying domain name searches and valuations, and explained how orchestration can enable AI to manage multiple specialized agents for complex tasks.

The conversation explored how companies can transition from using AI as a tool to integrating it as a teammate. Shah envisioned a future of hybrid teams comprising both human and AI members, emphasizing the need for new management roles, onboarding processes, and performance evaluations tailored for digital agents. He argued that treating AI like a human employee, with proper training and feedback, is key to leveraging its full potential.

Shah addressed skepticism about AI's capabilities, particularly criticisms that it is merely a 'stochastic parrot.' He argued that AI exhibits a form of creativity distinct from human intelligence and that dismissing it as just pattern-matching underestimates its transformative potential. He also discussed the rapid advancements in AI, noting that what seemed impossible a year ago, like cohesive video generation, is now achievable.

On the economic front, Shah cautioned about job displacement due to AI but maintained an optimistic outlook, framing AI as an amplifier of human capabilities rather than a replacement. He stressed the importance of individuals and organizations experimenting with AI daily to stay competitive.

The episode touched on Mark Zuckerberg's aggressive recruitment of AI talent with nine-figure offers, a strategy Shah described as 'diabolical' but potentially effective. He questioned whether assembling a superteam of engineers could replicate the success of OpenAI, given the rise of competitive models and open-source innovations.

Shah closed with insights into AI's role in enhancing creativity, using his son as an example. His son leverages AI to develop fictional worlds and test them through role-playing games, demonstrating how AI enables individuals to manifest creative ideas without needing advanced technical skills. Shah argued that AI democratizes creativity by removing skill-based barriers to expression.

Key Insights

• ChatGPT can act as a 'PhD intern in everything' if you feed it the right data. HubSpot's Dharmesh Shah uses it as a personal executive coach by uploading his finances, goals, and other data, turning it into a tool for highly personalized advice and strategic decision-making.
• AI's 'context window'—the amount of data it can process at once—is like a sheet of paper with limited space. Techniques like vector embeddings and retrieval-augmented generation solve this by letting AI query huge private datasets without exceeding its memory limits.
• Treating AI like a human employee unlocks its real potential. Dharmesh Shah predicts hybrid teams of humans and AI will require new management roles, onboarding processes, and even performance reviews for digital agents.
• Mark Zuckerberg is reportedly offering nine-figure salaries to recruit top AI talent. Dharmesh Shah calls this 'diabolical' and questions whether even a 'superteam' of engineers can outpace open-source AI models and competitive alternatives.

Key Questions Answered

How does Dharmesh Shah use ChatGPT as an executive coach?

Shah uploads personal and business data, including finances, goals, and documents, to create a customized AI assistant. He uses it to answer questions, make decisions, and even draft emails, describing it as a 'PhD intern in everything.'

What is the context window in AI, and why does it matter?

The context window is the limit on how much data an AI model can process in a single query. Innovations like vector embeddings allow users to query large datasets by fetching the most relevant documents that fit within this limit.

What did Dharmesh Shah say about Mark Zuckerberg's AI recruitment strategy?

Shah called Zuckerberg's nine-figure offers to AI researchers 'diabolical' but acknowledged it as a smart, albeit desperate, strategy to gain an edge in the AI arms race. However, he questioned its long-term effectiveness given the rise of open-source models and other competitors.