Blog
AI Engineering & Quality
Moving the Center of Mass
Every LLM engineering technique is doing the same thing: conditioning a probability distribution so its center of mass moves toward human intent and its tail risk gets compressed.
March 29, 2026
No More Pleading for Valid JSON
Rebuilding a vocabulary-extraction agent with Google ADK and Go: delete every format instruction from the prompt and let the type system and function calling constrain the output.
December 2, 2025
The Lethal Trifecta of AI Security
Simon Willison's lethal trifecta: when exposure to external content, access to private data, and the ability to communicate externally coexist, they form a nearly indefensible combination of risks.
September 24, 2025
Testing Large Language Models: An Overview
An overview of quality assurance for LLM applications: model evaluation vs. system evaluation, metric design, prompt attacks, and performance testing.
August 5, 2024
Probability & Sampling
The 2019 MCMC / DREAM series, written in Chinese and meant to be read in publication order. Only the first post has an English version. Companion implementation: The-DREAM-Algorithm.
Introduction to Differential Evolution Algorithm February 23, 2019