Systems on Glass
One idea from real systems, drawn simply, a few minutes at a time. System design from things I've built and broken at Amazon, Nissan and Microsoft; AI agents from how I use and build them every day; and what changes when AI runs at scale.
Systems on Glass is an AI-automation side project of mine. Every episode is generated by Claude (Anthropic's AI, working through Claude Code in my own video studio): the script, the animations and 3D scenes, the edit, and the narration in an AI clone of my voice. I choose the topics, bring the stories from my own work, and review each episode before it goes up. Not affiliated with Anthropic.
A curriculum, not a pile of tips. Each track gets harder week by week, so starting at episode one feels like levelling up: one server, then many servers, then servers that disagree; using agents, then making them yours, then building and trusting them.
Every episode: a real story with a number, the obvious answer crossed out, the real pattern drawn step by step, and one rule to keep.
Episodes
How seventeen calls an ad got cheap: hits and misses, read- and write-through, lazy versus eager, LRU to learned eviction, invalidation and the herd. Then caching for AI: the KV cache, prompt caching and response caching.
An agent isn't a smarter chatbot: it's a model in a loop. Think, act with a tool, observe, check, repeat; tools as the hands; and the goal, budget and human that stop it running forever.
How an API answers a million requests a second, each in under 20 ms. Two problems, two fixes, Little's law, why busy servers get slow, Prime Day headroom, and why averages lie.
Next up: Context is RAM (AI agents).