what-time.

… v?

Say it in English.

Type a phrase like yoga every Tuesday and Thursday at 7am and the model returns real occurrences, recurrences, and an RFC 5545 rule — resolved in your time zone.

timings appear here after the first parse Page-inclusive: this measures the whole round-trip on this page — input handling, the wasm call, JSON transfer, and rendering — so it reads higher than the parser's real cost, especially on a busy machine. For measured per-operation timings, run the actual true benchmarks below.
Reading…

marks: dates, times, repeats, durations

Try one of these —

Resolve context reference instant · time zone · limits
Benchmarks in-page wasm · CPU backend · nothing leaves the page
each example phrase × 200 timed parses

Bulk iterates the example phrases through the same in-page WebAssembly CPU pipeline. The native build adds a wgpu/Metal GPU backend that batches the same workload at ~56 µs/phrase at 50k (≈12× CPU) with identical results — see the repository's bench-bulk example.