Real-browser evidence

Measure the search kernel
where it actually runs.

Run Akashic’s versioned lexical-search fixture on this device. The lab records catalog transfer behavior, parse and index cost, query latency, and browser-supported main-thread and memory signals.

No accountNo analyticsNo typed questionsNo remote submission

Measurement run

Choose a sample depth.

Measured passes

Each pass runs every public evaluation query. One additional warm-up pass is excluded from the timing summary.

Ready. The benchmark starts only when you choose Run.

Completed report

This browser’s measurement

Elapsed times are device- and environment-specific observations, not universal performance claims.

Catalog request

First and repeat transfer

Request	Elapsed	Transferred	Encoded body	Cache signal

“Reload” asks the browser to validate a network-first fetch. “Repeat” uses ordinary cache behavior. A reported 0-byte transfer with a nonzero decoded body is treated as a cache signal, not a missing measurement.

Context

Device and API signals

Fixture detail

Slowest measured queries

Public case	Slowest run	Mean	Results

Portable evidence

Keep the complete report.

The JSON includes every timing sample, feature-detection outcome, public case ID, and method note shown here.

Inspect report JSON

How to use it

Compare context with context.

Run on the intended hardware.Capture low-, middle-, and high-capability devices instead of treating a desktop runtime as a phone.
Keep the full environment.Compare reports only with their browser, viewport, network, pass count, and capability fields intact.
Repeat before concluding.Background work, heat, power mode, and cache state can move a single result. Preserve variation rather than hiding it.

Method references: Resource Timing, supported performance entries, and browser memory measurement.