Short, practical writing on backend engineering, architecture, and shipping better software.
It's 2:47 AM on a Saturday. Your phone explodes with PagerDuty alerts. Production is down. Hard. The culprit? A single database connection pool got exhausted because one microservice didn't implement proper retry backoff. Under normal load, it was fine. But when traffic spiked by 30%—well within your "designed capacity"—the whole thing cascaded into a spectacular failure.
If you've ever wired together microservices, you already know how to build AI agents.LangGraph treats your AI application like a distributed system: LLMs are your services, prompts are your API endpoints, and the graph is your architecture diagram. Nodes? They're functions. Edges? API calls. State? Your Redis cache.The moment this clicked for me: "Wait, this is just a state machine with LLM middleware."Let me show you how everything you know about backend engineering maps directly to building intelligent agents...