Iven Zhang

Calculate first, explain second

The central idea behind OpenFate is simple: a fluent answer is not automatically a correct answer. Generative AI can explain complex ideas clearly, but it is not a reliable substitute for calculations that must be reproducible. In astrology and traditional chart systems, small details matter. A time zone, daylight-saving rule, calendar conversion, or chart-construction method can change the input before any interpretation begins. If the underlying chart is wrong, a polished explanation only makes the mistake harder to notice. That observation shaped our product architecture. OpenFate uses dedicated engines to calculate structured chart facts first. Only after those facts exist does AI help explain them. We call this approach “calculate first, understand more.” It gives the language model a narrower and more responsible role: communicating a result rather than inventing it.

✦ ADVICE FOR OTHER FOUNDERSDon’t put everything into a prompt just because AI can generate an answer. Decide which parts of your product must stay deterministic and reliable.

Why the boundary matters

Many AI products combine calculation, judgment, and explanation inside one prompt. That can look efficient, but it removes useful boundaries. When every stage happens inside a black box, a team cannot easily tell whether an error came from the user’s input, a conversion rule, a domain engine, or the model’s wording. Separating the stages makes the product easier to test. A deterministic result can be checked against fixtures and recalculated. An explanation can be evaluated for clarity, tone, and whether it remains faithful to that result. The distinction also makes failures more honest. If a calculation cannot be completed, the product should say so instead of allowing a language model to fill the gap with something plausible. This principle extends beyond astrology. Tax software, financial tools, scheduling systems, scientific applications, and health products all contain facts that should be computed or retrieved before AI starts speaking. Generative models are powerful interfaces, but they should not quietly become the source of truth for every part of a system.

Building for different traditions and languages

OpenFate brings together BaZi, Zi Wei Dou Shu, Western and Vedic astrology, Tarot, Human Design, Sukuyo, relationship compatibility, and related experiences. The interface supports English, Simplified Chinese, Traditional Chinese, Japanese, and Korean. That breadth creates a second responsibility: do not flatten distinct traditions into generic AI language. Different systems use different assumptions, terminology, and calculation rules. Localization is not only translation. It also requires preserving the structure and context that make a reading understandable to people familiar with that tradition. We therefore treat each calculation engine as its own domain system while keeping a consistent product experience around it. Users can inspect core chart outputs for free, then choose optional AI explanations when they want more guidance. The goal is to let people evaluate the underlying tool before asking them to pay for interpretation.

✦ ADVICE FOR OTHER FOUNDERSDesign for uncertainty. Let AI admit when evidence is incomplete or conflicting—a visible limitation is safer than a confident but unsupported answer.

Where OpenFate is going

Our long-term vision is not to make AI sound more mystical. It is to make complex self-discovery systems more understandable while keeping calculation, interpretation, and uncertainty clearly separated. The best AI experiences will not be the ones that speak with the most confidence. They will be the ones that know which facts must exist before the conversation begins.

AI Dust TeamEditorial team, AI Dust

Calculate first, explain second

Why the boundary matters

Building for different traditions and languages

Where OpenFate is going

More founder stories