For almost a year I've been building aitrainer.work alone, from Hanoi, on far too many Vietnamese iced coffees. It's a career platform for people who train and evaluate AI models: rating chatbot answers, writing expert answers for AI labs, that kind of work. It launched in January and has made $26,435 so far, mostly from referral programs.

Six years of SEO taught me one rule above all: pick one thing and focus on it. In October I ignored that rule completely and launched four spin-offs. Partly because my ADHD brain loves running a thousand projects at once, and partly because with Claude Code writing most of the code with me, starting a new site is cheap. Here they are, and why each one exists.

1. AI Training Board: the weird one

AI Training Board went live on October 6. Every site in my niche looks the same: the same template, a different logo. I wanted to see what happens if one of them is completely different.

So it has a mascot. His name is Loopy, and he's a spiritual heir to Clippy, the Microsoft Office assistant we all miss more than we'll admit. He helps you find AI training work on platforms my main site doesn't cover.

I've given it 12 weeks. If it can't win Google traffic against my main site and my competitors by then, Loopy retires.

2. Dataconto: enterprise data licensing

Dataconto launched on October 20, for companies that want to license their data to AI labs.

AI labs need training data, and plenty of companies are sitting on it without knowing what it's worth. Every competitor of mine is starting to move into this space, so I didn't feel I had much choice. My angle: I compare offers from several buyers, match the data with the offer that fits best, and handle the conversation between seller and buyer.

3. githubworth: what's your code worth?

githubworth.com estimates what a codebase could be worth to an AI lab. It's the developer-sized version of Dataconto: companies have datasets, you have repos.

If you have old projects collecting dust on GitHub, try it on one. You might be surprised.

4. DomDom Jobs: the one I care about most

At the very end of October I built DomDom Jobs, a family of niche job boards. It started with five and already has six: robotics, healthcare, UX, legal, veterans and moonlighting, for people who want paid work on the side of a day job.

They all share one template and the same data pipeline as aitrainer.work, so a new niche doesn't mean a new app. Same engine, different readers.

This is the one I'm personally invested in. I'm a UX researcher, and I know how frustrating it is to look for UX research jobs and get a page full of UI design roles you have no practical skills for. Under-served niches like UX research and UX writing deserve a board that understands the difference.

Meanwhile, back at the main site

The main site didn't take a break while I did all this. In October it had 26,354 human visits (traffic doubled), 12,637 clicks on Apply and $8,397 in revenue, down from $9,620 in September. That dip is normal for referral income, which arrives two to three months after the visit. October's traffic will show up in December.

GitHub also flagged my account on October 17, which stopped every deploy and scheduled job for a week. Living in Vietnam means every platform suspects I'm a bot. I'm not. I'm a person with too many side projects.

What happens next

Four spin-offs is either the smartest thing I did this year or the reason I'll be writing a very honest post in January. Each one has a few months to show it can bring in its own traffic. I'll share the numbers either way.

If you've built several products at once instead of one, I'd love to hear how it went for you.