I'd like to share a new edition of a series of hand-drawn illustrations I originally shared with members from August through October last year. For this edition, I revised the writing, refined the explanations, and reorganized everything into a sequence that builds progressively from beginning to end.
That period roughly coincided with the wave of frontier model releases spanning from GPT-OSS (OpenAI's major open-source release) to Qwen 3.
The source materials came from my consulting and research work (i.e., By Hand Research) breaking down frontier models and algorithms for various AI companies and financial institutions.
This paid membership tier is intentionally niche. Only a small percentage of advanced AI engineers and researchers need and want to understand frontier models at the math and algorithm level in this much detail. You belong to this small percentage. 🙌 And thank you for your support!
If you recently became a member, this would be a good series to start. And if you followed the original series last year, I hope this serves as a useful refresh with a clearer flow and updated explanations.
1. Expert Choice Mixture of Experts (MoE) (free)
2. MHA, MQA, GQA, MoE-A: More Attention! (free)
3. New GPT-OSS Trick to Ignore Tokens (free)
4. MXFP4, FP4, FP8
5. LoRA, Fine-Tune, Pre-Train
6. QLoRA, DoRA, BitFit, NF4 vs INT4
7. KV Cache, Prefill, Decode
8. EmbeddingGemma, MRL, InfoNCE, Embed vs. Decode
9. Inference Batching, Request-vs-Token Level
10. MLP Parallelism: Data, Context, Row, Column, Pipeline
11. RoPE vs PE in QKV Self-Attention
12. RMS, Group, Layer, Batch Norm, Tensor Parallelism
13. Qwen 3
