Research and long-form writing. The machine learning work is reproducible: code, data paths and the numbers behind every claim are public, negative results included. The T’ai Chi papers are written first for my students, and then for anyone working through the same questions in their own practice.
In preparation: a paper on how T’ai Chi practice progresses over decades, drawing on Anders Ericsson’s research on deliberate practice: how a student develops foresight, and what the fruits of long practice actually are.
We present a clean-room, reproducible implementation of native sparse attention (NSA) with custom Triton kernels, gated for correctness against a reference at every step and validated from Ampere (sm_86) to Hopper (sm_90). We report a +53% end-to-end training throughput …
The Tai Chi body is not a single feature but an integrated condition that arises slowly through patient practice. It is the body in which song, deep relaxation, has released enough holding for the connective tissue to organize as a coherent envelope: the trained body, or jin …
T’ai Chi is built at home, not in class. Most students know this and still struggle to practice, because progress is invisible day to day. The old teachers had an answer: every session adds something, whether you can see it or not. This paper gathers that teaching from …
The I Ching is not a fortune-telling book. Its oldest layer, the Zhouyi, was a Bronze Age decision-support system: a way to diagnose the current configuration of forces and how they tend to evolve. Because it describes how change behaves rather than any particular content, it …