Writing
Notes from building things: what worked, what failed, and what I would do differently. New posts land in the RSS feed.
Two tiny networks, hand-written backprop, and a ring. What a 2-D GAN makes visible that image GANs hide, including why a falling loss is bad news.
Why I built this, and what I'll share here: machine learning, frontier LLM research, reinforcement learning, and interactive experiments.
A tiny gradient-descent toy and what it shows about learning rates.