https://anupmohan.com Notes on AI agent infrastructure, distributed systems, and the things I break along the way. en Mon, 21 Sep 2026 15:41:21 GMT https://anupmohan.com/blog/rag-vs-context-caching/ https://anupmohan.com/blog/rag-vs-context-caching/ Fri, 04 Sep 2026 09:00:00 GMT I ran a small head-to-head between putting a whole governed corpus in context with prompt caching and retrieving from it with a managed RAG service. Caching won on quality and cost. Here's why, and when it stops winning. llmagentsevals https://anupmohan.com/blog/metadata-as-code/ https://anupmohan.com/blog/metadata-as-code/ Tue, 01 Sep 2026 09:00:00 GMT Every new metadata concept was turning into a new table, a migration and an API change. So we stopped modelling metadata in the database and started treating it like source code. metadataarchitecture https://anupmohan.com/blog/metadata-served-via-okf/ https://anupmohan.com/blog/metadata-served-via-okf/ Mon, 20 Jul 2026 09:00:00 GMT We put a few hundred governed definitions in Postgres because Postgres was already there. Here's why they ended up as markdown files in git instead, and how that one directory now feeds an API, MCP tools and an agent. metadataarchitectureagents https://anupmohan.com/blog/linucb-adaptive-survey-routing/ https://anupmohan.com/blog/linucb-adaptive-survey-routing/ Tue, 27 Jan 2026 09:00:00 GMT Long surveys tire people out, and tired people give worse answers. I used LinUCB, a contextual bandit, to decide question by question whether to ask or skip. This post covers how it works, how I trained it on historical data without leaking the answer, and why accuracy was the wrong way to score it. reinforcement-learningevals https://anupmohan.com/blog/why-python/ https://anupmohan.com/blog/why-python/ Sun, 20 Feb 2022 09:00:00 GMT Do you generate repetitive reports, scrape websites or parse log files? Python is a simple, practical language for automating everyday work. pythonautomation