Miatz playground

Reasoning Trace X-Ray

Edit one reasoning step mid-thought and watch the answer fork.

Coming to the Build-LabAIE-101BYOK AIAI-Coredev · data

The Demystify signature

Reasoning traces are usually a scrolling wall of text nobody reads; making one step literally editable and re-runnable turns 'the model thought about it' into an object the learner can argue with.

How it works

What goes in, what comes out

What it does

Learner gives a hard, multi-step problem to a reasoning-capable model and its full thinking trace renders as an expandable step tree instead of a hidden field. Clicking any step lets the learner strike or edit that single reasoning move and re-run generation from that point forward, watching the final answer fork live; a split-screen mode runs the identical problem on a fast non-reasoning model at the same time so the extra reasoning tokens' payoff — or lack of one — is directly visible.

You bring

A multi-step reasoning problem (math, logic, or a debugging scenario), and optional learner edits to a specific reasoning step.

You get

An expandable reasoning-step tree, a forked final answer after edit, a fast-vs-reasoning split-screen comparison, and a reasoning-token cost readout.

You control

Reasoning effort/budget level, problem difficulty tier, fast-model comparator on/off, and which step the learner is currently editing.

Where it's used

Free top-of-funnel teaser (single pre-built problem, no signup) and an advanced in-module lab extending AIE-101, reused inside AIE-401 for debugging real eval failures.

extended/interleaved reasoningchain-of-thought tracesself-correctionreasoning-vs-fast-model tradeoffsreasoning-token cost

Optional · watch & try

Go deeper, elsewhere

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

LLM Visualization (bbycroft)3D walkthrough of a GPT-style model's full forward pass, weight matrix by weight matrix How LLMs Work: A Visual GuideVisual walkthrough of the LLM pipeline from input text to output token How LLMs REALLY Work (Simple Explanation)Plain-language mental model for what an LLM is actually computing Attention in transformers, step-by-step | Deep Learning Chapter 6 (3Blue1Brown)Gold-standard visual explainer of the attention mechanism inside a transformer Transformer ExplainerLive in-browser GPT-2 running with every transformer stage visualized as you type A visual introduction to tokenization in LLMs | Byte Pair Encoding AlgorithmShows how byte-pair encoding splits text into subword tokens step by step

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