Visualizers
Learn by moving things
Interactive explainers for the ideas behind retrieval, embeddings, and agents. A formula tells you what is true; dragging a vector shows you why. Each one also appears inside the article it belongs to, and runs entirely in your browser, so drag, poke, and break it.
How far away is your data? (the latency ladder)
From: Sizing a System on a Napkin →The anchor latency numbers, made physical. Memory is a pen on your desk, an SSD is the next room, the network is a letter mailed overseas. Each rung is about 1,000x slower, and a human-scale blow-up turns nanoseconds into seconds you can feel.
Each rung is about 1,000x slower than the one above. If reading from memory took 1 second, a trip across the internet would take almost two weeks.
Scope: must-have vs later
From: How to Design a System →Step one of the routine. Features sort into must-have and later, because naming what you will not build is scope. Shown on the link shortener.
Split the must-haves from the nice-to-haves, and say out loud what you will not build.
Estimate: which way does it lean?
From: How to Design a System →Step two. A balance scale tips hard to the reads side, one write against a hundred reads, and that single fact leads straight to adding a cache.
One rough number decides the shape. Far more reads than writes means the reads should come from a cache.