Blog
Notes on running LLM traffic like infrastructure.
Administering gateways with SQL, the ProxySQL operating model for AI, LiteLLM alternatives, and cross-validated consensus answers.
Your LLM gateway should be administered like a database
The ProxySQL operating model — hostgroups, runtime config, health — ported to AI traffic, and why YAML redeploys hurt at 3am.
Read the post →MindBalancer vs LiteLLM: a single Go binary you admin with SQL
A single Go binary you administer with SQL versus a Python proxy with YAML. Deployment footprint, admin model, and when to pick which.
Read the post →Referee mode: cross-validated LLM answers for high-stakes prompts
What consensus routing is, when it helps, a config walkthrough, and the referee_info you get back.
Read the post →