Context Window Lab
Bury a fact in context and watch the model forget it.
The context window is normally an invisible number in a spec sheet; this renders it as a physical, fillable container the learner watches a real fact get lost inside, on their own question.
What goes in, what comes out
Learner fills a model's context window with a labeled stack of blocks — system prompt, chat history, retrieved chunks, tool results — rendered as a literal filling bar, then hides a specific fact at a chosen depth and position within that stack. Running the same question at increasing context sizes and burial depths generates a live degradation curve, in real time, on the learner's own content, instead of citing someone else's benchmark chart.
A target fact to bury, a stack of filler/distractor content, and a question that requires retrieving the buried fact.
A visual filling context-window bar, a found/not-found result per depth tested, a degradation curve across positions, and a token-cost readout for the full stack.
Context size (tokens), needle position (start/middle/end, depth percent), number and type of distractor blocks, and model family (to compare context-handling across models).
Free top-of-funnel lead magnet (single-run teaser) and in-module lab for AIE-103, paired directly with Prompt Lab's cost meter.
Go deeper, elsewhere
Hand-picked public explainers and open tools that complement this one — always optional, never required, never graded.
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.