We're based in Hangzhou, China. All three of us are heavy users of AI agents — Claude Code, Codex, DeepSeek, Kimi, Hermes, you name it. Back in May, Jensen Huang's talk in Taiwan opened my eyes to what looks like a massive global opportunity. An idea clicked: build a token and API aggregation platform.
So we started executing. Registered the company, built the website, set up overseas servers, and began reaching out to upstream model providers. My thinking was clear from the start: we're not building our own models, and we're not burning cash on inference infrastructure. Our value is integration — bringing together high-quality, cost-effective APIs and delivering them through a single, reliable endpoint. It hasn't been easy, but we've taken the first real steps.
Now we're stuck on customer acquisition. We're targeting the global market, but we're based in China and our potential customers are on the other side of the ocean. Social media is our only channel right now, and it feels like searching for a needle in a haystack — plus it's way too easy to accidentally break some community rule and get shut down. Are there any cross-border founders here who've been at a similar stage? I'd be incredibly grateful for any direction or advice.
Happy to discuss in the comments — I'll read every suggestion and piece of feedback carefully. Thank you.
Top comments (1)
Working with two co-founders is a powerful setup, but one thing that often gets overlooked early on is clearly defining not just roles, but also accountability. Especially when you're all leveraging AI agents like Claude Code or Codex, it's easy for lines to blur. The agents can produce code quickly, but who truly owns the architectural vision, the testing strategy, or the ultimate responsibility for a module's performance and bugs? In my experience, even with incredible code generation, the human element of critical thinking, edge-case handling, and deeply understanding the why behind a feature remains paramount. It’s not just about the code, but the complete solution lifecycle.
My advice would be to treat your AI agents as incredibly productive junior developers, but never as the sole architect or product manager. This means one of you still needs to be the technical lead driving the overall system design, ensuring consistency, and setting coding standards that the AI-generated code must adhere to. The same goes for product ownership – who's truly validating the problem space, talking to users, and defining the features before the agents even start generating ideas?
Consider how you'll manage the merge process and code reviews. If a large chunk comes from an agent, the review process shifts from "did you write this well?" to "does this code solve the problem correctly and fit our architecture?" This often means more rigorous integration testing, as the agent doesn't understand the full system context or your specific business constraints. The speed advantage of AI is real, but make sure it's driving you toward a validated, robust product, not just a mountain of quickly generated code.