One worktree per task: how I run parallel AI coding agents with herdr and pi
My day-to-day setup for running several AI coding agents at once on a remote dev box, without them stepping on each other or on main.
Fullstack Developer sharing insights on PHP, Laravel, Go, Vue.js, React and modern web development from Kuala Lumpur, Malaysia.
I've been building web applications since 2011, across digital agencies, startups and freelance work, mostly with Laravel, Vue and React, and more recently Go and Node.js. I've worked as a tech lead, and I care most about refactoring legacy code, clean architecture and code that's easy to test.
Today I'm a software engineer at ContactOut, and I also work on Avidnote, an AI research workspace for academics.
Lately my main experiment is how far AI agents can go in a real engineering workflow: separate agents for coding, QA, design, research and growth, each with its own job, reviewing each other's work. I write about what works, what breaks, and what it costs.
PHP (Laravel, Symfony, WordPress), Node.js, Go
Vue.js, React.js, Angular, TypeScript, Tailwind CSS
MySQL, PostgreSQL, ElasticSearch
AWS, Digital Ocean, Docker, CI/CD
herdr, pi, Cursor, Hermes Agent, MCP connectors, agent-run QA and code review
My day-to-day setup for running several AI coding agents at once on a remote dev box, without them stepping on each other or on main.
Building a personal blog is one thing, but deploying it to production where the world can see it? That's where the real adventure begins. Today, I'll walk you through exactly how I deployed my Next.js 15 blog to DigitalOcean App Platform, including all the gotchas and solutions I encountered along the way.
How I collaborated with Claude Code to build a full-stack Next.js blog application with Supabase and WorkOS authentication in record time.