Devlog 003: Our First Overseas Order, and Why We Don't Build an AI Chat Feature

The site recently passed a small milestone: our first paid order from overseas.

Seeing Zi Wei Dou Shu cross a geographic distance and earn the trust of someone abroad is genuinely encouraging for the team. It is also a reminder. Going forward, both our free blog articles and our paid reports need more care around being language-friendly and culturally readable, so that readers outside the Chinese-speaking world can follow what Zi Wei Dou Shu is actually saying with less friction.

Alongside that news, we have been getting an interesting question in our inbox: "Since you already use AI, why not just build a ChatGPT-style conversation service, where people can keep chatting and asking as much as they want?"

Behind that question is really a trade-off between technology and cost.

To turn rigorous, in-depth chart analysis into a back-and-forth chat service, and keep each answer at a consistent level of quality, the backend would have to call the most capable, but also the most expensive, large language models very frequently. Over time, users could end up paying a fairly high cost to keep using it, which is not necessarily the best deal for anyone.

So for now, sǹg-miā has chosen a "single focus, one-time in-depth analysis" model. Before you read a report, you describe your current situation and the questions on your mind as clearly as you can. The backend AI then has enough context to produce a well-structured analysis that stays close to you.

A reasonable budget in exchange for a precise report with real substance: that is the practical choice we are sticking with for now.

Thank you again to everyone supporting us from overseas and here in Taiwan. See you next time.