A shopper types in a search box; sliders, facets, and sort all hit the server. Each keystroke adds latency. Today's faceted-search SaaS solves this by charging per query.
The InstantSearch.js client layer - facets, sort, search-within-results - running entirely in the browser at native input rate.
Algolia-grade refinement with no per-query cost, no rate limit, and no server round-trip while the user is dragging a slider.
Hadron is not claiming the GPU is fastest. It picks the right backend for each operation - sometimes GPU, sometimes Worker, sometimes CPU, sometimes a categorical refusal. The result is the same; the cost is different. Watch the Dispatch evidence panel beside each app: every keystroke and slider drag prints which path was chosen, the score that drove the choice, and a one-line reason.