Let's Build Something

One of the hardest parts of product work isn't coming up with ideas—it's dragging those ideas out of your head and into the real world where they can actually do something. If you've ever stood in front of a whiteboard absolutely crushing it, only to realize weeks later that none of those ideas made it past a meeting note, you know exactly what I'm talking about.​

Introducing: “Let’s Build” Video Series

I'm launching a new video series focused on going from idea to execution with AI tools — regardless of your skill level in design or coding. The goal isn't to build AI products just for the sake of it—it's to show how product managers, designers, and builders can use AI to move faster, stress-test their thinking, and get in more reps across the product development process.​

Why This Matters for PMs (and Anyone Who Builds)

The fastest way to get better at building products is to build more products. But let's be real—most of us don't get unlimited shots on goal. We don't get to fully own every research cycle, design sprint, or build phase. Budgets, bandwidth, and time are always against us.​

That's where AI changes the game.​

With tools like Replit, Cursor, Claude, and GPT, you can run through more cycles faster. You can prototype ideas, run cheap experiments, challenge your own assumptions, and validate concepts before you ever pull your team into a conversation. Think of it like flight simulator time for product people—it's not exactly the same as shipping, but the reps build your instincts.​

It’s not just about professional development… it’s also about having fun.

I sat down with my two older kids for about 15 minutes last night and we built a game based on some stories/comics they used to write called “Chickens vs Beetles”. This game is a simple version but it started with blob characters and a totally open space. We talked through what the levels should look like, how the bad guys should get harder, and how the chicken should get weapon upgrades based on how many beetles it takes out.

I have been trying to get them to wrap their heads around systems thinking and how to plan something before they just try to do it for years. this was the perfect medium to capture their attention and highlight the benefits of taking that approach. Let’s hope they remember this the next time they want to build an apple canon or a tree house ;)

Want to start with an easy project? Check out this tutorial on how to build a custom GPT and a podcast to tell the story of your career.

Research vs. Prototyping: What's the Right Move?

One of the most common early-stage questions is: Should we validate this with research or just build a prototype? There's no universal answer, but here's how I frame it:​

• Research: Helps you understand if a problem exists, what users are doing today, and where the gaps are.​

• Prototypes: Test whether your solution actually works—instead of asking "Would you use this?" you get to see if they actually do.​

The good news? AI lets you blur the line between these two. You can run research faster, spin up lightweight prototypes, and collect structured feedback at a fraction of the usual cost.

The other dynamic that changes here is that because it is faster, cheaper, and easier to build something, you don’t have to build for other people. You can build something for yourself and then see if anyone else is interested. That is the initial thinking that got me started on the PMBS Tester, but then I realized I have worked with, coached, and talked to a lot of product leaders who have the same problem.

If you want to see how this thinking applies beyond AI, I broke down how to spot and prioritize the right problems to solve in this post on The Problem Obsession. Knowing when to research vs. prototype is a skill that compounds over time—AI just helps you get more reps.​

Meet the PMBS Tester—Your Personal Product Reviewer

For my first project in the series, I'm building something I'm calling the Product Manager Bullsh*t Tester (PMBS Tester). It's exactly what it sounds like—a tool to help PMs gut-check their own work before they drop it in front of leadership or their team.​

The concept: You upload your docs, and the PMBS Tester gives you critical feedback, calls out weak assumptions, and highlights areas that need more work. Think of it like a brutally honest peer review... but powered by AI.​

Now I did a silly thing in this first video and went straight from the concept in my head to exploring the architecture of the agents and workflows. I have a tendency to overcomplicate agentic workflows, so it helps me to get a gut check on my thinking before I think through each step.

Here's the fun part—I'm using CrewAI to build it. CrewAI lets you define AI-powered agents that take on different roles in a workflow. In this case, I've got:​

• Curiosity Agent: Asks follow-up questions to clarify thinking.​

• Skeptical Agent: Highlights where assumptions might be off.​

• Researcher Agent: Fact-checks claims and pulls in external data.​

Now, I probably won’t create a file or start writing out a single line of code for this until I have my prompt design done, but having a rough sketch of the project helps me think through what those prompts need to look like.

It's basically a product pre-mortem you can run solo. If you want to go deep on how AI agents can support product work, check out this piece I wrote on Be Reasonable.​

How GPT Projects Streamline the Build Process

Before writing a single line of code, I leaned on GPT Projects to map out the entire system. If you haven't used them before, think of them like a running product discovery doc powered by AI—all your prompts, decisions, and revisions live in one workspace so you can evolve your thinking as you go.​

For this project, GPT Projects helped me

• Document the user journey and interaction flow.​

• Prototype different agent combinations in text form before committing to code.​

• Generate initial files and workflow templates for CrewAI, making the handoff from planning to building super smooth.​
  

  If you're curious about how structured prompting unlocks more effective AI workflows, I break it down here.​

Follow Along and Build With Me

I'll be documenting the whole process—the good, the bad, and the broken—in this series. Here's what you can expect:​

• How to shape your own idea into something you can build.

• How to use specific features and tools to drive productivity and efficiency.

• How to use something like Replit or Cursor to get something functional up without having to understand code.

• How to learn and iterate on your product and your ability to wield these tools.

I might keep going with some of the Onboarding Breakdowns and product highlight videos as well but I’m personally excited to be building and I want to help others do the same.

Please share what you’re building or share this post to help get someone else going. I’d love to hear from you and even talk through our projects sometime.