Is Prompt Caching Actually Lowering Your AI API Bill?
Check whether prompt caching reduces cost per completed task, accounting for cache writes, retries, review effort and the charges on your provider's bill.
LLM operations
Your AI workflow has logs. Can they explain one bad decision?
A practical guide to tracing one bad AI decision across retrieval, model calls, tools and APIs—without creating a sensitive-data archive.
Your AI Model Has an Expiry Date: Build the Migration Lane Early
AI models and endpoints retire. Learn how to inventory dependencies, test replacements, run a controlled release and preserve a usable rollback route.
OpenAI Responses API: The Assistants Migration Clock Has Run Out
The Assistants API is gone. Learn how to move legacy OpenAI workflows to Responses, choose the right state model, and control risk and cost.
Written recommendations from Trafik og Veje, Aarhus Municipality (2011) and AgroTech (2010).
