​AI by Hand Seminars by Prof. Tom Yeh

​I’ll calculate the math and sketch the architectures by hand ✍️.

​1. Gemma 4 (Google): alternating global and local attention, KV-cache, sparse mixture of experts, per layer embedding.

​2. Qwen 3.6 (Alibaba, China): long-context scaling techniques and attention optimizations improving practical context length.

​👉 3. Kimi 3 (Moonshot AI): Kimi Delta Attention, attention residuals, and extreme MoE sparsity, activating just 16 of 896 experts in a 2.8T-parameter model.

​

​About the Seminar Series

​In 2026, I’ve made a personal commitment to teach regular live AI seminar by hand ✍️.

• ​Foundation: intuition, math, and mental models of core AI concepts, for beginners.

• ​Frontier: research papers, algorithms, and architectures used by frontier models, for advanced AI engineers and researchers.

​Live attendance is free.