GraviChoice
A decision assistant that maps your pros and cons into weighted spheres, then lets physics decide. GLM-5.0 analyzes semantic weight; Matter.js simulates gravity on a balance scale.
On the path to AGI — curious about what comes next.
I build things to figure out why.
From business foundations to the frontier of AI product engineering.
Systems built at the intersection of physics, intelligence, and interaction.
A decision assistant that maps your pros and cons into weighted spheres, then lets physics decide. GLM-5.0 analyzes semantic weight; Matter.js simulates gravity on a balance scale.
7-player Texas Hold'em with AI opponents. Full game loop — chip tracking, blind structure, hand rankings, and a tutorial system. Built in React 18 via CDN, no build step.
Random Forest vs XGBoost for predicting student outcomes. Feature importance via SHAP values. 82%+ accuracy with K-Fold validation. Deployed as interactive Streamlit app.
Intelligent writing training hardware + APP. EM & pressure sensor fusion drives a 3-DOF
servo-controlled robotic arm for real-time posture correction (≤2mm precision).
Led product from user research (N=485) to full Axure prototypes for child & supervisor apps.
🏆 互联网+ Provincial Award · 挑战杯 Innovation Prize
Authoritative benchmark data from LMSYS Chatbot Arena, HuggingFace Open LLM Leaderboard, and LiveCodeBench.
| Model ↕ | Org | Arena Elo ↓ | MMLU ↕ | HumanEval ↕ | MATH ↕ |
|---|
Data sourced from LMSYS Chatbot Arena leaderboard, HuggingFace Open LLM Leaderboard, and published model technical reports. Scores are approximate and reflect the state of publicly available benchmarks at time of collection.
Thinkers, builders, and voices shaping the intelligence frontier. Kept in original language.
Trending AI repositories, live from the GitHub API.