An Old Trade Gets a New Engine: The Problem Was Never Demand — It’s Delivery

Fortune telling, feng shui, love compatibility readings — this business has existed for thousands of years, and demand has never dipped. Search “BaZi reading” on Etsy, browse astrology TikTok, watch tarot livestreams — paying customers line up. But the trade always had one fatal flaw: delivery quality depends entirely on the practitioner’s personal skill. It can’t be standardized, and it can’t scale. One reader maxes out at ten clients a day, charging anywhere from $15 to $150 depending on reputation and charisma.

Then AI showed up and made it awkward. Plenty of people tried asking ChatGPT to “do my BaZi chart,” and the model just winged it — mismatched Heavenly Stems and Earthly Branches, luck pillars that made no sense. Clients who knew the craft spotted it instantly; clients who didn’t got misled. AI’s biggest problem in metaphysics is hallucination — it doesn’t know the rigid rules of calendar conversion, but it answers with total confidence anyway.

Now a group of developers has cracked this with plain engineering. In plain English, what they built is: give the AI an SOP manual plus a calculator. A “Skill” is a standardized AI capability pack — the AI handles conversation, follow-up questions, and report writing, but “chart casting” (calculating your Four Pillars from your birth date and time) is math that can’t be wrong, so a Python script does all of it. The AI never touches the calculation. On top of that: the school of interpretation gets declared upfront, missing info triggers follow-up questions instead of guesses, and a classical-text knowledge base validates the output. One sentence: the cyber-psychic finally has a leash.

What does that mean? A total beginner can now produce an accurate chart. But heads up — between “generating a chart” and “delivering a report a client will pay for again,” there are three gaps: basic terminology, packaging, and handling follow-up questions. The barrier to this business didn’t collapse. It dropped from “ten years of apprenticeship” to “a few days of ramp-up.”

Where the Opportunity Is: Five Battle-Tested Weapons Already on GitHub

Facts first. The repo jinchenma94/bazi-skill focuses on Four Pillars charting and analysis. It has 2,977 stars and 468 forks, MIT-licensed — meaning free commercial use. It ships with rule summaries from nine classical texts, collects birth info through interactive dialogue, then auto-generates the Four Pillars, luck cycles, and annual forecasts.

The heavier hitter is FANzR-arch/Numerologist_skills — 1,149 stars, covering Qimen Dunjia, Zi Wei Dou Shu, and BaZi. Its design principles are printed right in the README: ask before answering, outsource deterministic math to scripts, declare your interpretive framework before drawing conclusions. The Qimen module even ships a standalone Python CLI for fixed calculations.

There’s also Ming-H’s yinyuan-skills for love and marriage readings, voidforall’s fengshui.skill for home feng shui analysis (Flying Stars, Eight Mansions, date selection), plus a tarot-skill. BaZi, Qimen, Zi Wei, love compatibility, feng shui, tarot — the six biggest monetization categories in metaphysics all have open-source, engineered AI solutions now.

But ready tools don’t equal a working business. These repos solve “the chart won’t be wrong.” They don’t solve “where do clients come from,” “why should they trust you,” or “what do you say when they ask deeper questions.” As of now, there’s no publicly verifiable seller case proving the full loop — install tools, list on a marketplace, stable monthly income — actually works. So the honest framing is: the tools are ready; the model is unproven. Your job isn’t picking money off the ground. It’s running the experiment.

Path One: Build Your Own “AI Metaphysics Studio”

The zero-experience checklist first. You need a Claude Code environment (Anthropic’s official CLI coding tool — register an account, attach a paid plan), you need to open a terminal (Terminal on Mac, PowerShell or CMD on Windows), and you need to copy-paste commands. If you’ve never opened a terminal, spend 30 minutes on a “Claude Code setup” tutorial video first, or grab a pre-packaged installer. Don’t get stuck at step one.

Step one: install. Using bazi-skill as the example, setup is trivial — you just need python3 (3.6+, pure standard library, zero pip dependencies), then run this in your project root:

1
2

mkdir -p .claude/skills
git clone https://github.com/jinchenma94/bazi-skill .claude/skills/bazi

Want it available globally? Clone to ~/.claude/skills/bazi instead. Once installed, type “BaZi reading” or “cast my chart” in Claude Code to trigger it. The Skill walks you through collecting name, birth date (solar or lunar), birth hour, gender, and birthplace, then runs the charting script to output the Four Pillars, luck cycles, and annual forecasts — followed by a classical-text-informed analysis.

Step two: layer in Numerologist_skills for Qimen and Zi Wei coverage. The Qimen script runs like this:

1
2

pip install -r qimen-dunjia/scripts/requirements.txt
python qimen-dunjia/scripts/qimen_cli.py --input tmp/qimen_input.json --output tmp/qimen_output.json

Step three: polish the deliverable. The AI gives you raw analysis material. You package it into a beautifully formatted PDF chart report — Day Master strength, Ten Gods relationships, favorable elements, luck cycles, career and relationship guidance, with a cover page and interpretation notes. Build the template in Canva once, and the marginal cost per report drops to nearly zero.

But before running this pipeline, spend a few days learning basic BaZi concepts. You don’t need mastery, but terms like “Day Master,” “Ten Gods,” “luck pillars,” and “annual forecast” should be explainable in your own words. Otherwise, when a client asks “what does this Indirect Resource star mean for me,” you’ll just read the AI output verbatim — and they’ll feel instantly that you don’t know your stuff. Trust gone.

Path Two: List It and Sell the Report as a Standardized Product

Tools installed — now the sales logic. Metaphysics products convert well on every platform because they sell “the feeling of certainty,” and repeat purchase rates are high (annual forecasts every year, Qimen readings before big decisions).

Etsy / eBay: List a “BaZi Four Pillars Reading + Annual Forecast Report” at $15–$45. The client sends birth info after ordering, you run the Skill, format the PDF, deliver. Twenty minutes per order.

Lemon8 / Pinterest / Instagram: Post “AI BaZi reading was scarily accurate” content with chart screenshots and classical-text citations, funnel to DMs or email to close. Young female audiences on these platforms are extremely responsive to love and career readings — yinyuan-skills hits that nerve dead-on.

TikTok / YouTube Shorts: Tarot and astrology content pulls massive traffic in Western markets. Pair tarot-skill with short-form readings, funnel to a Shopify store selling English-language “Chinese Astrology” reports at $29+ — Western buyers pay a curiosity premium for Eastern metaphysics.

Digital product marketplaces: Sell the “report template + AI prompt pack + Skill setup guide” bundle itself as a digital download at $5–$10. Volume play.

Case Breakdown: Why Engineering Is the Moat in This Business

Look back at Numerologist_skills’ design philosophy — it’s fundamentally a quality control system. The traditional reader’s bottleneck is inconsistent form and knowledge blind spots; raw AI’s problem is making things up. This framework splits the pipeline into four stages: if info is incomplete, ask — never force a chart; rigid calculations like calendar conversion and chart casting go entirely to scripts, AI never touches them; interpretive frameworks, day-boundary rules, leap-month handling get declared before any conclusion; high-risk domains (health, legal, financial) get structured references only, never professional advice.

What does that buy you? Every deliverable you ship has constant-quality charting, a transparent process, and traceable errors. If a client takes your report to an expert for verification, the chart portion was computed by a script — it can’t be wrong. The interpretation cites classical sources. That’s your differentiation from the random “masters” on marketplaces: they run a craft workshop, you run an assembly line.

But “interpretation with citations” doesn’t mean “interpretation is correct.” Metaphysics schools contradict each other — the same chart can yield opposite conclusions under different frameworks. The classical rule summaries the AI cites are the repo author’s distillation of ancient texts, not the texts themselves. So the precise claim is: your charting can be “guaranteed correct,” but your interpretation can only be “sourced and internally consistent” — never “absolutely accurate.” When clients dig deep, you still need fundamentals to back yourself up.

According to the source post (not independently verified), these Skill-based solutions are multiplying fast, with aggregator projects already bundling BaZi, love readings, Qimen, and tarot into all-in-one skill packs. The window won’t stay open long. Whoever closes the delivery loop first eats.

Risk and Boundaries: Compliance Is Life or Death Here — Not a Footnote

Let me be blunt: fortune-telling categories are high-voltage wires on most platforms, and “careful wording” won’t save you. Etsy, eBay, and Meta platforms all have policies around fortune-telling and occult services; content moderation on “superstition” tightens every year. You could spend two weeks building an account, posting content, growing traffic — and lose it all to one policy strike. That’s not a tail risk. It’s the norm in this category.

So the honest advice is: pick one of three routes.

Route one: sell the toolkit, not the reading. Package “report templates + AI prompt pack + Skill setup guide” as a digital product. You’re selling a tool for practitioners and learners, not a fortune-telling service. That’s a knowledge product, not a superstition service — one full risk tier lower.

Route two: go private-domain, don’t depend on platform traffic. Build through your own newsletter, community, and referrals. Use platforms for visibility only; close and deliver through email or a Substack-style subscriber list. No algorithmic moderation in your own funnel, but growth is slow — best if you already have an audience.

Route three: serve the overseas market. Target the Chinese diaspora or Westerners curious about Chinese metaphysics. Sell English-language reports on Shopify, Etsy, or Fiverr. Western platforms tolerate “Chinese Astrology” far more than domestic platforms tolerate fortune telling — and the price points are higher.

Whichever route you take, hold three lines: every repo ships with a disclaimer — for cultural learning and entertainment only, not a basis for decisions — and your sales copy must hold that line. Never promise to change luck or ward off misfortune. Client privacy: birth data is sensitive personal information — disclose how you’ll use it, purge it after delivery. And stay out of health, legal, and financial decisions — structured references only, never professional advice.

The tools are sitting there open-source, and 2,977 stars mean plenty of people see what’s happening. But between understanding the code and making money stand three walls: platform rules, client trust, and foundational knowledge. Spend 30 minutes tonight installing bazi-skill into Claude Code and casting your own first chart — you’ll already be ahead of 90% of the people who only watch. Then spend a few days on basic terminology, then a week landing your first order. The barrier dropped from ten years to a few days. But a few days isn’t zero.