People are getting into multi-agent setups, and I had a ton of questions.

In this episode, I’ll walk you through them and share what I learned.

We’ll start with the basics. Why did we move from simply calling LLMs to building agents?

And why did multi-agent systems show up so fast?

Then we’ll get practical. What breaks first as agents grow?

When should you split agents into multiple roles?

I’ll also touch on agent frameworks, and I’ll end with Anthropic’s Agent Skills.

Are you frustrated by limited context windows? Tired of feeling like a "session manager" for your AI coding agents?

In this episode, I introduce Beads, an exciting new open-source project designed to be an "external brain" for AI. Learn how this tool uses a dependency graph to fix AI amnesia and help your agents manage complex, long-term tasks.

cagent, from Docker, runs multiple AI agents on your machine and even creates new agents for you.

It operates outside the sandbox, with real access to network socket and things other AI coding tools can't access, so it can automate things in a way cloud-based assistants can’t.

Think of it as an AI team living on your laptop, doing work on your behalf.

So in this episode, I’ll tell you what I discovered, and how I felt while playing around with it.

Ever chosen a technology that seemed perfect, only to watch it become a nightmare a year later? In this episode, we explore why even the smartest engineering decisions can backfire as systems grow. Through personal stories (including a distributed networking disaster), we challenge the myth that good architecture means getting everything right from the start.

What should we expect from a programming language when AI is writing the code? In this episode, I focus on Go — its strengths, its flaws, and why I keep using it.