06 / ENGINEERING NOTES

Engineering Notes

Long-form notes on AI, engineering systems, and organizational evolution.

GPT, Down to the Scalar: How I Read MicroGPT

Tracing Karpathy's MicroGPT from characters and computation graphs through causal attention, cross-entropy, Adam, and sampling—with worked examples and diagrams of the complete learning loop.

Read article

From Random Weights to Local Inference

How I implemented and trained a decoder-only Transformer on a 16GB MacBook Air, then learned why tokenization, data quality, evaluation, and inference controls matter more than a low validation loss.