Thoughts on LLM routing, API architecture, and building resilient AI applications.
Models are becoming commoditized. The future isn't about picking one model to rule them all, but dynamically routing prompts to the cheapest model that gets the job done.
Why testing three different AI models shouldn't require three different credit card charges and $15 in minimum deposits.
DeepSeek just dropped deepseek-v4-flash-0731. It is incredibly fast, insanely cheap, and available right now on RouteLM.
When an LLM provider goes down, your app shouldn't. Here is how we engineered RouteLM's routing layer to handle outages instantly and transparently.
Juggling multiple API keys, dashboards, and SDKs was wasting my time. I just wanted a single endpoint that works. Here is why RouteLM exists.