Tianfu Agent 2M+ View Campaign

Company: a group of AI researchers and folk metaphysics practitioners came together around one belief: traditional culture deserves to be treated seriously with technology.

Product: Tianfu Agent treats Chinese metaphysics as an experience-based inference framework, then studies, validates, and engineers it into an AI-era product that helps users reduce uncertainty and make more reflective decisions.

Official account: @DestinyLinker

Official X post

MingLi-Bench project page

MingLi-Bench benchmark repository

• Technical angle: push the Agent paradigm into a domain with scarce training data, weak automated validation, and frequent rule conflicts.
• User angle: make the reasoning visible, separate real-time calculation from memorization, cite benchmark validation, and admit uncertainty.
• Cultural angle: avoid fatalism and superstition framing; focus on a rigorous engineering attitude toward a long-running knowledge system.
• Discussion angle: use the boundary between science and metaphysics to create a higher-quality debate surface.

The campaign used a creator matrix instead of a single brand announcement. Different creators translated the same product facts through technical, product, cultural, and debate-led frames.

The proof layer leaned on MingLi-Bench and a global fortune-teller competition benchmark with 3,069 participants, positioning the model close to human Top-20 performance rather than relying on self-claims.

A redeem code provided the conversion hook: limited to 10,000 uses, granting 3,000 points; China site tianfuagent.com and overseas site tianfu-ai.com.

Not a black box: users can inspect how charts are generated, how palaces are interpreted, and how different schools of thought are weighed.

Not memorized copy: instead of replaying training fragments, the product narrative emphasizes real-time calculation plus rule-based reasoning.

Knows when to stop: for complex or contradictory cases, it can call tools up to 500 times, say “I don’t know” when unresolved, and ask key follow-up questions when allowed.

The campaign did not try to prove metaphysics as science. It used the idea of Chinese metaphysics as a long-running large-scale statistical attempt to show how an AI product can handle a non-consensus domain with rigor.