This site showcases the practical lessons on building enterprise scale products with AI — from a product leader who decided to get back to the craft of building.
Like many of you, I started learning about AI coding from YouTube, X, and friends who couldn’t stop sharing what they built over the weekend. Every day someone posts a demo where a single prompt produces something that looks incredible. Then you try it yourself and realize the demo was the easy part — the product is full of holes, and you spend the next three days fighting to get the last 20% right.
For a personal project or the tenth iteration of a to-do list app, that’s fine. But you can’t bring that to your workplace, where badly designed and buggy software has real consequences — broken customer experiences, operational failures, compliance risks. Most vibe coding tutorials don’t bridge that gap. They’re built for startups and side projects, not for billion-dollar enterprises operating at scale with legacy systems, data complexity, and integrations that were built before you were born.
This site covers what actually happens when you use AI to build products in that world: the failure modes that don’t show up in tutorials, the patterns that survive contact with production at scale, and the strategic questions that leaders at legacy companies should be asking about AI adoption right now.
If you’ve been wondering how to go from impressed by the demos to actually shipping with AI at work — that’s exactly the gap this blog exists to close.
The backstory
I’ve spent the last decade building digital commerce products in automotive — an industry where you’re ultimately moving atoms, not bits, and real life throws your beautiful designs out the window on a daily basis.
I started at Carvana when it was still early, leading the team responsible for the homepage, search, and product listings — the top three most-trafficked pages on the site. Then Rivian, where I was one of the first product hires on digital commerce, building the end-to-end purchase and delivery experience from scratch for a vehicle that didn’t exist yet. Then Clutch.ca, where I ran all digital products and helped scale the business to $150M/year.
Those were all clean-sheet companies. Software-first from day one. The challenge was building fast and scaling what worked.
Now I’m at Hertz, leading digital products across the global rental business — and it’s a fundamentally different problem. This isn’t a greenfield build. It’s taking decades of operational infrastructure that actually works and moves millions of vehicles a year, and modernizing it without breaking what made it successful. Think less startup, more taking a beaten-up Porsche 911 and turning it into a Singer. You’re marrying legacy strengths with modern technology, and the result can be more powerful than either one alone.
I believe that if you can marry a seamless digital experience with efficient operations in a single stack, the business becomes untouchable. AI is what makes that possible. That’s what pulled me back to building.
Why I started writing
As I moved higher in my career, I found myself losing touch with how to actually craft a product. More and more of the job became managing people, running alignment meetings, negotiating scope. The craft part — the part I got into product for — was slipping away.
AI changed that entirely. I’m back to building. The team is a byproduct of the work, not the other way around.
I had an intuition about a feature I wanted to test. In the past, that meant writing a PRD, meetings with data science and engineering, scope negotiations, prioritization debates — and if I really pressed hard, maybe it ships next sprint. Instead, I opened an AI coding tool over the weekend, described what I wanted, and had a working prototype deployed in an afternoon. I had results that week.
That experience keeps repeating. Every time it does, I learn something worth writing down. And honestly, I need the accountability. Writing forces me to actually understand what I did, not just that it worked.
What you’ll find here
For builders: How to use AI coding agents reliably — parallel agent orchestration, review gates, the debugging nobody talks about. The craft, documented honestly.
For leaders: How to think about AI adoption in operationally complex businesses. What changes when product people can build prototypes themselves. Playbooks for navigating this transition without burning the org down.
The thread connecting both: AI is changing how products get built, and what it means to be the person responsible for building them. Curiosity, ambition, and a bias toward building matter more than whether you can write code from scratch.
Get in touch
I’m Yilun Zhang. If something I wrote is wrong, if you’re working on similar problems, or if you just want to compare notes — I want to hear from you.
