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.

Portrait of Sachin Gupta rendered in binary

When a popular file expires everywhere at once, every edge would hit the origin together. An origin shield is one mid-tier node all edges fetch through, so the origin sees a single request. The thundering herd at CDN scale.

Origin shield: one fetch to the origin, not one per edgeauto
edge 1edge 2edge 3edge 4edge 5origin5 hits at once

Without a shield, when a popular file expires everywhere at once, every edge fetches it from the origin at the same moment. Five edges, five simultaneous origin hits.

A checkout spike of 5,000 orders would melt a slow email step. A queue holds the backlog as depth and lets the worker drain it at a steady rate, so the spike becomes a short wait, not a crash. This is load leveling.

A queue absorbs a rushinteractive
checkoutproducerqueue depth40email workersteady 100 a second

A calm minute: about 50 orders come in, and the email worker keeps up. The queue stays almost empty.

Many workers read the same queue, and each message goes to exactly one of them. Add workers to drain a backlog faster: five hands clear in a fifth of the time, with no message handled twice.

Add hands, not armsinteractive
one queueorder #812order #813order #814order #815order #816worker 1drains in about 50 seconds

One worker pulling the rail: 5,000 receipts at about 100 a second take roughly 50 seconds to clear.

Visualizers — Sachin Gupta