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.
What an agent SDK gives you
From: An Agent SDK Is a Harness You Didn't Build →Click each part of the harness an SDK ships, the run loop, tool registry, model adapter, memory, tracing, to see what it does and which side of the line it sits on: SDK-provided versus what you still bring.
Calls the model, runs the tool it asked for, feeds the result back, and decides when to stop.
Four ways an agent loop breaks
From: Most Agent Failures Are Loop Failures →Most agent failures are loop failures. Click each mode, never stops, stops too soon, goes in circles, no error recovery, to see the symptom and the deterministic guard that fixes it.
The loop keeps calling the model and running tools, never deciding the task is done. Burns budget forever.
A hard step budget (max iterations) plus a real completion check.
The coding-agent loop
From: A Coding Agent Is a Loop Around a Compiler →Step through a coding agent's loop: write, compile, read the exact error, fix, test, fix, until it is green. Execution is the ground truth the model iterates against.