Pack the Brain
Pack the AI's memory before it forgets your question.
Makes context rot -- degraded attention as a window fills -- a visible heat-strip the learner watches cause a wrong answer, instead of a warning line in a lesson.
What goes in, what comes out
Learner drags content blocks -- system prompt, chat history, retrieved documents, the actual question -- into a fixed-size context bar sized to a real model tier. Once the bar overflows, the lowest-priority blocks get force-evicted with a visible 'forgotten' animation, and a heat-strip across the remaining context shows attention thinning toward the middle of a stuffed window before the model answers, often wrong.
Drag-and-drop of content blocks; slider adjustments to window size.
A running token meter, live cost readout, a pass/fail on whether the model still answered correctly after eviction, and the heat-strip visualization.
Context-window size selector (4K/32K/128K/1M, named to real tiers); live dollar-per-1K-token meter; eviction strategy toggle (oldest-first vs priority-tagged); context-rot heat-strip on/off.
Lab companion inside AIE-103; homepage/hero embedded widget (directly answers the number-one buyer fear -- API cost anxiety -- per the learner heartbeat).
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.