Ian Nowland

LLM Agents Are Breaking Your Platform, Not Your Architecture LLM agents don’t require new architectures, but they do break the assumptions your platform relies on for safety, observability, and control. This post breaks down what really changes when agents hit ...

Since starting Junction Labs, we keep getting asked if "agentic architectures" will replace microservice architectures.

Our take after a year: the architecture survives. The platform doesn’t.

Here’s out take on what actually breaks when agents hit prod:
🔗 www.junctionlabs.io/blog/llm-age...

From Nitro to Junction: Testing in Production at Scale During my time at AWS, I learned that even the most rigorous pre-production testing has limits. I share how we built Nitro's reliability by treating production as part of the test loop—with proper saf...

At AWS, I led the team that built EC2's Nitro virtualization stack— C code deployed to 500K servers. Biggest lesson? Pre-production testing has limits. You must embrace safe production testing.

My new blog explains how Nitro did it, and how Junction brings this to any Kubernetes team.

DNS and the December 2024 OpenAI Outage In December 2024, OpenAI faced a significant outage lasting approximately four hours. This incident highlighted a critical challenge in container orchestration: maintaining reliable service discovery ...

In my new blog post for Junction Labs, I explore service discovery by delving into the December OpenAI outage. I analyze root cause and discuss a principle we grasped during the development of EC2's Networking: static stability. Check out the full post here: www.junctionlabs.io/blog/dns-and...