Curious about the AI strategies of tech insiders? I recently had the privilege of eavesdropping on a captivating dialogue between Dharmesh Shah (HubSpot's co-founder and CTO) and the hosts of the My First Million podcast. The insights shared were truly enlightening.
"Use ChatGPT daily, no matter what your job is."
That's Dharmesh's unequivocal advice. And he's not just jumping on the AI bandwagon; this is coming from someone who built a chat application using OpenAI's tech two years before ChatGPT was even released.
"Every day, you should be in ChatGPT. I don't care what your job is. You could be a sommelier at a restaurant, and you should be using ChatGPT every day to make yourself better at whatever it is you do."
Coming from the CTO of a $30 billion company who saw the potential way back in 2020, that's advice worth considering.
The Tech Behind the Magic
What makes these AI systems tick? Dharmesh broke down some fascinating concepts:
Context Windows: AI's Limited Working Memory
Imagine you're writing on a sheet of paper that can only hold a certain number of words—both your writing and any response must fit on that same sheet. That's essentially what a "context window" is for an AI.
"You have a limited sheet of paper that can only fit a certain number of words on it... That sheet of paper is what in technical terms would be called the context window."
Current top AI models can handle about 100,000-200,000 tokens (roughly 75,000-150,000 words). Sounds like a lot until you're trying to summarize an entire book!
Vector Embeddings: How AI Finds Meaning
Here's where it gets really interesting. Dharmesh explained how AI can access information beyond what it was originally trained on using vector embeddings.
It's like having a brilliant PhD intern who knows everything publicly available but nothing about your specific business. With vector embeddings, when you ask a question, the system searches through your documents, finds the most relevant ones based on meaning (not just keywords), and feeds just those specific documents to the AI.
"You can take those 100,000 documents and put them in this special database called a vector store... And when someone asks a question, we can go to the vector store and say, 'Give me the five documents out of the 100,000 that are most likely to answer this question based on the meaning of the question.'"
Tool Calling: AI's Internet Access
Perhaps the most mind-blowing concept was "tool calling"—essentially giving AI the ability to utilize external tools, such as search engines and calculators.
"The LLM was trained on a certain number of things, but it would be like saying to an intern, 'Whatever you know, you know, but we're not going to give you access to the internet.' That would be stupid."
When ChatGPT needs information it doesn't have, it can now request to use a tool (such as a web search) to obtain the results and incorporate them into its response. It's a clever workaround that dramatically expands what AI can do.
From AI Tools to AI Teammates
The most compelling part of the conversation was Dharmesh's vision for how we'll work with AI in the future, a future where AI serves as an amplifier of our capabilities, not a replacement. This optimistic outlook is something we can all look forward to.
AI as an Amplifier, Not a Replacement
"The thing I'm most excited about with AI is that it's not you versus AI. That's not the mental model you should have. The right mental frame of reference is that you are at the mercy of AI. AI is an amplifier of your capability."
This perspective shift is crucial. Instead of fearing AI will replace us, we should be figuring out how to leverage it to become dramatically more capable versions of ourselves.
Managing AI Agents Like Human Employees
Dharmesh predicts we'll soon have hybrid teams of humans and AI agents—and those AI agents will need management just like human employees do. This preparation is key to ensuring a smooth transition into the AI-driven future.
"We're hiring these agents and expecting them to do magic just like if we hired an exceptionally smart employee who has a PhD in everything and expected them to do magic with no training, no onboarding, no feedback, no one-on-ones, no nothing."
This could create entirely new job categories, such as "agentic managers" who specialize in optimizing the performance of AI systems.
The AI Talent Wars
The conversation took an interesting turn as Dharmesh discussed Meta's aggressive poaching of AI researchers from OpenAI and elsewhere, reportedly offering compensation packages of over $ 100 million.
Dharmesh called this strategy "diabolical" but acknowledged the potential upside if Meta can leapfrog competitors in AI capabilities. Still, he questioned whether simply assembling the most expensive talent would guarantee breakthroughs.
Will AI Kill Creativity? The Opposite
One of my favorite insights was Dharmesh's take on AI and creativity:
"AI increases creativity because it will increase the percentage of ideas that people have in their heads that they will then be able to manifest, regardless of what their skills are."
He shared how his 14-year-old son uses ChatGPT to develop complex fictional worlds despite limited writing skills—effectively testing creative concepts years before he would have the technical ability to express them conventionally.
What This Means For You
So what's the takeaway from all this insider knowledge? Start using AI now and incorporate it into your daily routine. Don't think of it as a simple tool—think of it as a powerful amplifier for your existing capabilities.
The gap between what we can imagine and what we can create is shrinking dramatically thanks to these technologies. Whether you're a writer, designer, programmer, or yes, even a sommelier, there are ways AI can help you work smarter and bring more of your ideas to life.
As Dharmesh puts it, it's not "you versus AI"—it's "you to the power of AI." And that power is available right now, waiting for you to tap into it.
Addendum: Security Considerations When Using AI
While AI tools offer incredible benefits, they also come with important security and privacy considerations that weren't covered in Dharmesh's conversation. Before diving headfirst into using AI for everything, it's crucial to keep these concerns in mind and act responsibly.
Data Privacy and Confidentiality
Remember that when you input information into ChatGPT or similar tools, that data may be stored, analyzed, or used to improve the AI models. Avoid sharing:
- Personally identifiable information (PII)
- Confidential business information or trade secrets
- Client/customer data covered by privacy regulations
- Healthcare information protected by HIPAA
- Financial details or intellectual property
Many companies now have AI usage policies for exactly this reason. Verify that your organization has established guidelines before uploading sensitive information.
Hallucinations and Misinformation
AI systems can "hallucinate" — confidently providing incorrect information that sounds plausible. Always verify critical information from AI outputs, especially for:
- Legal advice
- Medical information
- Technical specifications
- Financial guidance
- Historical facts
Treat AI as a first draft provider rather than a final authority on factual matters.
Security Vulnerabilities
When using AI coding assistants or allowing AI to interact with your systems through "tool calling" features:
- Review all generated code carefully before execution
- Never grant AI systems unnecessary permissions or access
- Be cautious about integrations that connect AI tools to internal systems
- Consider using air-gapped or sandboxed environments for sensitive work
Ethical Considerations
Be mindful of how AI tools might be used in ways that:
- Reinforce existing biases
- Make critical decisions without appropriate human oversight
- Create content that could be misleading or harmful
- Automate processes that require human judgment
Finding the Right Balance
These cautions aren't meant to discourage AI use, but to encourage responsible adoption. A good approach is to:
- Start with low-risk, non-confidential tasks
- Gradually increase AI involvement as you understand its capabilities and limitations
- Implement clear boundaries for what information you share with AI systems
- Regularly audit how AI tools are being used in your workflows
By keeping these considerations in mind, you can harness the power of AI while minimizing potential risks, truly getting "you to the power of AI" in a responsible way.
What's your experience been with using AI tools in your work? Are you using them daily, occasionally, or not at all? I'd love to hear your thoughts in the comments!
Reference - My First Million Podcast
Comments
Post a Comment