Thoughts on agentic LLMs, AI-assisted development workflows, and setups that make me a better developer.
I spent two weeks wiring Jira, GitHub issues, log analysis, and Copilot coding agent into one bug triage pipeline. This is what actually mattered.
Every multi-repo org has a problem: shared contracts that silently fall out of sync. I built a tool that finds them.
I wanted something like oh-my-opencode but fully customizable. Here's what I learned—and how I implemented it in pi-coding-agent.
Your AI forgets everything between sessions. Skills are how you make that stop mattering.
I caught myself mid-keystroke—about to write code I should've delegated. The real unlock: parallel, not sequential.
When your AI agent writes a 433-line technical spec and concludes 'don't build this,' maybe you should listen.
Start with 'how much time is this worth?'—not 'how would we build this?'
Building my first OpenCode plugin
On extracting wisdom from the people you work with
How I Built an Automated Multi-Perspective Code Review System—And Fixed Its Biggest Flaw
I'll be sharing my experiments with agentic LLMs, AI-assisted development setups, and workflows that have made me a more productive developer.