Loading...
Loading...
Go beyond single-agent workflows. Build parallel agent fleets, LangGraph DAGs, merge queues, custom MCP servers, and Docker-isolated devboxes. Learn from Stripe Minions, Gas Town, and Karpathy's auto-research to design production multi-agent systems.
What You Build
A custom multi-agent pipeline with parallel execution, a custom MCP server, and Docker-isolated agents
Sections
Lessons
Estimated
Concepts
Prerequisites
Dev Jordan
walkthrough · casual
First 5 lessons are free (38% of the course). No login required.
AI Coding Agents
Autonomous software agents (Claude Code, Stripe Minions, Gas Town Polecats) that read codebases, write code, run tests, and open pull requests with minimal human interaction during execution.
Self-Improving Factory
A closed-loop system where agents monitor user feedback, analytics, A/B tests, error logs, and production data to populate the backlog without human curation.
Blueprint Architecture
Hybrid systems combining deterministic code nodes (run linters, push changes) with agentic subtasks (implement feature, fix CI). A pipeline with guardrails, not full autonomy.
MCP (Model Context Protocol)
A standard protocol connecting AI systems with external tools and data sources. Enables agents to access specialized functions and services through a unified interface.
Real-World Factory Implementations
Production-grade software factories already operating at scale: Stripe Minions (1,300+ PRs/week), Gas Town (Kubernetes for agents), Karpathy's autoresearch, Linear + Cursor integration.