Statistical Intuitions
Musings on Machine Learning, AI and Neuroscience from a scientist turned entrepreneur.
Articles
On Harnesses and loop-engineering in AI
A deep dive into the systems that turn language models into agents: tools, memory, context construction, runtime loops, and product surfaces.
The Efficiency Gap in Hardware Driving Biology and AI
How do cells and GPUs compare when running the exact same algorithm? A single cycle of the Na⁺/K⁺-ATPase pump reveals a seventeen orders of magnitude gap between biological and silicon hardware.
The Tool Architecture of Claude Code
A follow-up deep dive into the leaked source code, this time covering the tool system: 43+ tools, a 7-phase dispatch pipeline, layered permissions, and a concurrency scheduler that batches reads while serializing writes.
The Memory Architecture of the Anthropic Harness
A deep dive into how the Anthropic Harness remembers you across sessions — reverse-engineered from the leaked source code. 3 read paths, 2 write paths, 1 directory of markdown files.
The physics behind world models
Part of the "Sunday Classic Paper" series. Covering the seminal 1986 paper by James Kajiya that sits at the heart of computer graphics.
Probability and its origins
We are given a formulaic definition of probability without an intuition for why it is true. The goal of this post is to shed some light on these details.
An End to End Machine Learning Pipeline
A complete hands-on tutorial covering data scraping, feature engineering, conventional ML models, and deep neural networks for multi-modal movie genre classification.