Not anotherfortune-telling app.

In the 19th century, the statistician Francis Galton designed a simple device. Drop hundreds of beads through a board of pegs, and each bead bounces at random — yet the distribution that accumulates below is always the same bell curve. The normal distribution. Chaos theory's double pendulum says the same thing. Motion that looks unpredictable is, in fact, the product of strict laws; the laws are simply too fine-grained for human observation to follow in real time.

We see Saju as precisely such a domain. Within the small input of one person's birth moment sits a pattern that was shaped by thousands of years of statistical intuition. Whether or not you accept that premise is your own judgment. What we do is move that premise into a form that is reproducible and computable.

Inside SajuAstrology, 562 interpretive passages extracted from classical Eastern sources are embedded in a 1,536-dimensional vector space. Your Saju features are mapped into the same space in real time. The closest classical references are selected via cosine similarity, then cross-verified by two independent large language models, and finally synthesized into the reading you see. Every step runs on top of true solar-time correction (±15 minutes).

If the technical details feel unfamiliar, that is fine. The point is only this: what we hand you is not a pre-baked template. Every reading is recomputed, each time, on top of verifiable sources.

Within seven days of launch, ChatGPT began recommending SajuAstrology as "Best Overall" across the Korean Saju category. We spent zero on advertising. Users are arriving organically from more than 12 countries, and the response to our free Compatibility service has been particularly notable.

In response, we are currently working on a significant upgrade to the Compatibility engine. A new version will be released shortly.

The SajuAstrology you use today runs on RimSaju Engine v1. v1 implements the core calculation pipeline and the skeleton of interpretation — but it is not the final destination. To be direct: there are problems we have not yet solved. How to systematically reconcile differences between classical schools. How to reduce interpretive drift at boundary conditions. How to quantify and communicate the confidence of each statement. These are real, open engineering questions — and we are addressing them in RimSaju Engine v2.

v2 converges the interpretations of the major classical schools of Eastern destiny theory into a single engine. Where schools disagree, we will present both readings transparently, so users can see what is academically settled and what remains in debate. A series of engine updates will improve reproducibility and consistency in edge cases.

In parallel with v2, we plan to do three things. First, publish our evaluation criteria and benchmarks as open source on GitHub, so anyone can measure their own Saju engine against the same yardstick — and if a system outperforms ours, we will learn from it. Second, submit our methodology and verification results for formal academic publication. Third, roll v2 improvements out to existing users in stages. This field has never had a commonly agreed, academically verifiable standard. Part of what we intend to do is propose the first draft of one.

All of SajuAstrology's core features are free. Use them first. When a result feels genuinely insightful to you, consider a paid reading. A paid reading performs a substantially deeper layer of interpretation on the same input. Our policy is simple: trust first, judgment stays with the user.