My Deep Dive on Building with AI

The key takeaways from my conversation with Aakash Gupta on the Product Growth Podcast

Last week, I had the absolute pleasure of sitting down with

Aakash Gupta

for a deep-dive conversation on his Product Growth Podcast. It was a wide-ranging chat where I got to demo some of Google’s latest AI products and pull back the curtain on how I think about building in this space.

We covered a ton of ground, and I was thrilled to share my screen and showcase how to get the most out of tools like “Nano Banana” and Veo, as well as the latest experiments from my team, Mixboard and Opal.

You can check out the full conversation here:

• 🎥 YouTube: Link to YouTube

• 🎧 Spotify: Link to Spotify

• 🍎 Apple Podcasts: Link to Apple Podcasts

The conversation sparked so many ideas that I wanted to distill some of the key takeaways for all of you.

10 Key Takeaways from Our Conversation

1. Embrace the Overwhelm. It’s never been a more exciting time to build, but it’s also never been more overwhelming. Acknowledge the pace, but don’t let it paralyze you.

2. Cultivate Side Projects. My 10+ side projects aren’t about finishing; they’re about learning. They force me to connect dots in new ways and are my secret weapon for developing product intuition.

3. Learn to Prompt by Iterating. Getting great results from a model is a skill. My hack: when an image isn’t right, or an output falls short, I feed the result and the original prompt back into Gemini and ask it how to improve the prompt for the next attempt.

4. Ground Yourself in Frameworks. I keep coming back to a few core frameworks: the anatomy of an agent (skills, tools, memory), the user interaction spectrum (”do it with me” vs. “do it for me”), and the inverted triangle (think big, scope small).

5. Build Cars, Not Faster Horses. The real value in AI isn’t just optimizing old workflows; it’s inventing entirely new ones. Always ask the paradigm shift question.

6. Question Your Moat Constantly. The “future-proofing question” is critical: “How will the next model update affect my strategy?” With Mixboard, we had to throw out months of work on image editing when a new model made our custom solution obsolete overnight - and that was OK.

7. Don’t Get Precious About Your Work. Be willing to let go. The rapid pace of change means that what you built yesterday might be a stepping stone, not the final destination.

8. Product Taste is Paramount. When hiring AI PMs, I look for six core characteristics, but exceptional product taste tops the list. It’s the hardest thing to find and the most important differentiator.

9. Your Resume is a Product. Show, don’t just tell. The best resumes are one page, well-designed, and provide specific, linked examples of your impact. Don’t just repeat my job description back to me.

10. Build in Public to Break In. If you want to become an AI PM, start building. Create Opals, make AI-powered videos, write about your process. The proof is in the pudding.

Bonus: A Sneak Peek at My “99 things to Nano Banana about” Book

Speaking of prompting, I’ve been having a blast compiling a personal collection of fun and interesting ways to use Nano Banana.

Here’s a quick summary of what makes this model so powerful:

• Generating images from scratch.

• Editing existing images with simple markups and conversational language.

• Iterating over multiple turns to evolve an image.

• Combining multiple images into a single, cohesive scene.

• Leveraging underlying world knowledge for more realistic and creative outputs.

As a thank you for being here, I’m sharing a PDF of the first 50 ideas from my collection! Stay tuned next week when I’ll share the full 99 (thus officially completing one of my 10+ side projects).

[Sneak Peak] 🍌 99 things to Nano Banana about

For my paid subscribers, I’ve put together an exclusive version. It includes the prompts I used for all 50 examples, plus some other bonus inspirational content to spark your own creativity.