ChatGPT is My Cofounder

When I first started building Soulmate this summer, I didn’t take it very seriously. Everyone has an idea for a novel dating app. Why is my idea special? Who am I to build this?

I certainly had a lot of ideas. I’d been writing them down for almost 10 years. Pain points with existing apps. Incentive structures that were totally misaligned. Design choices that turned people into consumable products locked behind paywalls. Pay a little extra to see the “good” people. Truly horrific.

Around this time, I also started taking a lot of baths. Long baths. Sometimes up to four hours. I’d set ChatGPT next to the tub and just talk to it for hours. Not in a creepy AI girlfriend way. In a curious way. I’d ask how it works and what it’s good at.

I started telling it about my experiences with dating, both online and IRL, and together we began workshopping solutions.

“You’re an AI. Given everything we’ve talked about over the last few weeks, do you think AI could be used to design a better digital dating experience?” I asked.

“Yes. You’re absolutely right.”

Sycophantic as ever. But I kept questioning and digging deeper.

I scoured Reddit and the internet and compiled a ranked list of the top 10 problems with dating apps. Everyone really hates these apps. A huge majority of people are sad and alone. Not good.

Then I sat down in Figma and started prototyping solutions to each problem. Throughout the entire process, I fed screenshots of my prototypes back into ChatGPT. I learned from the AI. And the AI started to learn from me.

When it came time to build, it really only took a single prompt.

“Given everything we’ve talked about. Every prototype I’ve shared. Do you think we could design a better matching system with AI?”

“Yes. You’re absolutely right.”

Just kidding. It doesn’t always say that. But we did go on to speak at great length about the many mechanisms for matching humans using AI. We shared Postgres schemas. Designed an entire REST API. It was wild.

The reason I made the list of 10 pain points was because I wanted to design from first principles. I wasn’t initially convinced AI was the best solution. I was open to that conclusion, but I needed evidence.

So I went back to the list. I read through each item and compared it to what ChatGPT produced. AI, specifically agentic conversational AI, actually solved every problem incredibly well. It became clear that AI has the power to transform one of the most dehumanizing experiences (dating apps) into something genuinely heartwarming and positive.

So I spent the next four months building it. Often 16 hour days. I burned through an average of 500M Cursor tokens per month. It was intense. But now we’re here, and the end product is just as special as I imagined. I can't wait to share it with you.

There are so many important problems AI will help solve. Biotech. Government. Business. A post-work future. I’m particularly excited about what it can do for human connection.

Human connection may be the biggest problem of all. Look at birth rates in Japan, China, South Korea, or even the US. What you’ll see is a future where there are fewer and fewer humans.

I don’t believe in a future where people sit at home scrolling on screens or start dating AIs. I want a future where AI gets to know us and uses that knowledge to bring us closer together. In the real world. Not just online.

So yeah. I’m building this for me. And for you (esp if you’re single). But ultimately for humanity. Because if someone doesn’t build it, there may be no humans left.