Miatz playground

Retrieval Race

Three search strategies race for the same buried answer.

Coming to the Build-LabAIE-202Browser (JS)AI-Concept-Vizdev · data

The Demystify signature

Turns 'retrieval quality is usually the bottleneck, not the model' from a lecture line into a number -- the exact rank position where the right answer got buried.

How it works

What goes in, what comes out

What it does

One query fires simultaneously at keyword-only, vector-only, and hybrid retrieval over the same small corpus, and the learner watches three ranked-result lists populate side by side in real time. The correct chunk's rank in each list is called out explicitly, including when it never appears in the visible top-k at all -- the lost-answer failure made literal.

You bring

A query, typed or chosen from adversarial presets.

You get

Three ranked-list panels with the gold chunk's rank highlighted (or flagged missing), and a winner banner naming which strategy actually found the answer.

You control

Corpus picker (a few themed demo corpora); top-k slider; strategy toggle set; adversarial query presets (synonym-heavy, typo'd, out-of-order).

Where it's used

Demo inside AIE-202's retrieval sub-module; lead magnet at /playgrounds.

retrievaltop-khybrid searchthe lost-answer failure mode

Optional · watch & try

Go deeper, elsewhere

Hand-picked public explainers and open tools that complement this one — always optional, never required, never graded.

Introduction to Vector Embeddings and Vector SearchFoundational explanation of turning text into vectors for similarity search Embeddings & Vector Databases ExplainedConnects embeddings to how a vector database actually stores and searches them Vector Databases simply explained! (Embeddings & Indexes)Covers embedding indexes (e.g. HNSW) at a beginner level Embedding Projector (TensorBoard, Hugging Face Space)Interactive 3D projection of embedding vectors to see semantic clusters and analogies Retrieval Augmented Generation (RAG) Explained in 8 Minutes!Fast end-to-end mental model of the RAG pipeline · ~8 min RAG Explained in 12 MinutesSecond angle on the same pipeline with a different worked example · ~12 min

Concepts click when you open the machinery.

Three labs are already live and free — the same hands-on style this playground brings to its module.

Play the free labs Take the DSAT