Cost & Latency Bench
Watch a demo's pennies turn into a production cost cliff.
The '1000x' extrapolation slider is the reveal — a $0.02 demo run becomes a visible $4,000/day line the moment the learner drags one slider, making the hidden unit-economics cliff impossible to unsee.
What goes in, what comes out
Learner fires the identical task at several models simultaneously and watches a synchronized race-bar animation of token-by-token streaming speed next to a live-rising cost odometer, reading real numbers off the platform's own cost meter rather than a spec sheet. A '1000x' slider then extrapolates that single run's cost and latency to a chosen production scale (requests/day), turning invisible unit economics into a visible cliff the learner has to decide whether to ship past.
A representative task/prompt, a quality bar it must clear, and a target production request volume.
A live latency race animation, a rising cost odometer per model, an extrapolated daily/monthly cost-at-scale chart, and a recommended routing/caching strategy under the quality floor.
Model set to race, requests/day extrapolation slider, caching on/off, routing strategy (cheapest-that-passes vs. fixed model), and a quality floor (minimum eval score) the routing must respect.
Read-only intro in Core Foundation F6 (Tools, Keys & Cost Literacy); full in-module lab for AIE-402 (COULD tier, Operator badge).
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.