Essays and notes on agentic systems, backend engineering, and things I've learned building in production.
Multi-agent LLM pipelines rarely verify that a step actually got done right before passing it on — and that failure mode already accounts for nearly a quarter of recorded multi-agent failures.
An Apple Shortcuts + OpenAI agent that turns plain-language notes into Reminders and Calendar events so you can write once and stop copying tasks by hand.
A practical walkthrough of building and publishing a Chrome extension — from the manifest file to the Chrome Web Store.