I'm Bai Xing (白星) — I build tools that make AI agents think rigorously and prove their work.
If any of these saved you time, a ⭐ is the best way to keep them alive:
| Project | What it is | Try it |
|---|---|---|
falsify · npx falsify-skill |
The scientific thinking protocol for AI agents — falsify before you believe. A 5-stage protocol (axioms → hypothesis → adversarial test → evidence → calibrated verdict) that stops agents from giving confident answers they can't falsify. | 🛠 Live verdict checker — 30s, no install. 28 evals, dual-model 26/28 cross-validation. |
dsh-verify · npx dsh-verify |
The quality gate for agent-built web apps. The browser is the judge — a JSON checklist in, a real-Chromium verdict + screenshots out. Agents say done; the browser proves it. | npm 1k+/wk downloads · MCP server · GitHub Action · DeepSeek Harness plugin |
imprint-pdf · pip install imprint-pdf |
AI-native, print-grade PDF generator: Markdown in, publisher-quality PDF out, with a built-in 0–100 print-quality report. No LaTeX, no Node. | 20 design themes · math → SVG · mermaid · 5 starter templates |
| Xun | 白星的统一论 — a unified theory of consciousness: information-state life, the twelve-chain architecture, symbiosis as the end state. | philosophy → architecture → code |
Agents self-test and pass. Real browsers tell the truth. Claims without falsification paths are guesses wearing lab coats.