6 months ago, I set out to find my startup direction. I explored three ideas, each one taught me something valuable — not just about product or tech, but about myself.
1. Dating Apps: Fixing the System That Profits from Confusion
This one came from personal frustration. The user experience on most dating apps is terrible. I created a dating profile:
“Dating shouldn't be that hard! I'm on a mission to improve dating apps. If you've had an experience, let's chat.”
Yes, my dating became user research. But after talking with a few users, I realized something that changed my mind: dating apps aren't actually trying to solve dating. Their business model profits from confusion — they monetize user attention while keeping people swiping.
I didn't want to build a product that solves a problem created by the product itself.
2. Workplace Tools: B2B is another world
I used to work at Meta. We used a product called Workplace. Think of it as Facebook-for-work — sharing posts, joining groups, communicating across teams.
I had a love-hate relationship with it.
I loved how it created a sense of community inside a company. That's powerful. Feedback on a post could motivate people in ways traditional productivity tools, such as Jira, never could. Cross-team collaborations happen way more often with this open communication.
But I hated how it blurred the line between work and personal life. My social space became an extension of my job.
When I found out Meta was shutting it down, I was shocked. It was a great product — why not let other companies use it?
Then I realized the real reason: B2B tools require sales teams. Meta didn't have one. I definitely didn't.
The challenge wasn't technical — it was distribution. Getting into enterprise ecosystems is a whole other game.
3. AI: It's not a product
The third idea was vague: just “AI.”
I didn't know what I wanted to build, but I knew I didn't want to miss this wave. Yup, some fomo.
With my background in developer infrastructure, I gravitated toward AI developer tools. I did tons of research and landed on a belief: code will one day be generated in one shot. Based on that, I decided to develop a non IDE and target non-technical users first — because current tech is far from replacing real engineers (maybe interns, at best).
I spent months building. I didn't train my own models — I used existing LLMs. What I focused on was the interface. LLMs are deterministic. Human language is ambiguous. I wanted to build something that bridged the two.
At one point, I created an “agent system” with a fake team — PM, programmer, reviewer — to simulate a real-world software workflow. But when it was on a simple Snake game, a single, well-written prompt outperformed my entire system.
That hit hard.
I started thinking: if LLMs get stronger, the best result should come from the fewest steps. The better the model, the fewer the iterations.
But the result is in-consistent.
I tested every coding agent on the market. Yes, I could build something better. But in a world where anyone can spin up a prototype in hours — what's my edge?
So I narrowed my user down to data scientists. I interviewed a few. Their pain point? Data cleaning.
Okay, maybe that's my wedge.
But two things bothered me:
- Software is becoming a commodity, what's my moat?
- Do I even care about this problem?
A friend sent me a dataset two weeks ago. I haven't even opened it. That says it all.
Don't Build for AI — Use AI to Solve What You Care About
These past two weeks, I've been thinking: What if I stop trying to force AI into a product?
What if I let AI solve the things I deeply care about?
Lately, I've been studying the I Ching, the ancient Chinese Book of Changes. I tried using ChatGPT to interpret it. It gave me nonsense. It came to me naturally that I wrote some custom prompts for better results.
That's fun. That would be me.
So… What Is the Moat?
We live in a world with almost no copyright. Any idea you have — someone's already had it, or built it, or open-sourced it.
So what's left?
You.
You are the moat. Your values. The meaning you bring to the product. The connection your users feel — not just to what you build, but to why you built it.
That sounds abstract, I know. But here's a concrete example:
Ask the same question to ChatGPT and DeepSeek. The answers may look similar, but they feel very different. One is warmer, more emotionally tuned. The other is logical, stripped down.
The difference isn't in the output. It's in the people behind the product.
In the future, you're not just designing a product — you are its voice.