Thoughts on engineering, AI, and building things.
How I used multi-agent LLM simulation to predict audience reactions to social media posts — research-grounded personas, emergent dynamics, and real-time analysis.
Exploring how Snap's Agent Format could enable structured discovery for multi agent systems.
How markdown is becoming the programming language for AI agent skills.
Using CLI hooks to intercept and block unsafe shell commands from AI coding agents.
A technical deep dive into building AI agents with a persona and skill system, where adding a new domain is mkdir and a markdown file.
Why I'm starting this blog, what I'll write about, and what you can expect.