I Let 30 AI Agents Loose in My Repo (Gas Town)

Description

What happens when you turn one AI coding assistant into a coordinated team of 20–30 agents? In this video, I test Gas Town, an open-source AI agent orchestrator, and see what happens when I let it loose inside a real codebase. Instead of using Claude Code as a single assistant, Gas Town acts like an AI dev team even with full persistence if something crashes. 🔗 Relevant Links Gas Town Repo - https://github.com/steveyegge/gastown ❤️ More about us Radically better observability stack: https://betterstack.com/ Written tutorials: https://betterstack.com/community/ Example projects: https://github.com/BetterStackHQ 📱 Socials Twitter: https://twitter.com/betterstackhq Instagram: https://www.instagram.com/betterstackhq/ TikTok: https://www.tiktok.com/@betterstack LinkedIn: https://www.linkedin.com/company/betterstack 📌 Chapters: 00:00 – Turning One AI Into 30 AI Agents 00:34 – What Is Gas Town? (AI Agent Orchestrator Overview) 00:55 – Mayor, Beads, Convoys & Polecats Explained 01:50 – Live Demo: Adding JWT Auth With AI Agents 02:47 – How Gas Town Breaks Features Into Parallel Tasks 04:20 – Gas Town Repo Output 05:00 – Running the App: JWT Auth Fully Working 05:21 – Pros: Scaling AI Agents for Developer Productivity 05:50 – Cons: Cost, Fatigue & Code Quality Risks 06:30 – Will People Actually Want to Use This?