Twenty measured pairs showed correct answers in this sample but no established latency advantage for is-jodd over awaited is-jeven negation.
Our live comparison did not establish that is-jodd is faster than !(await isEven(n)). That is the starting point for interpreting the experiment, even though the package began with a smaller-and-faster ambition. This report comes from is-jodd’s publisher and is not a TypeSafe performance claim.
The recorded results and raw observations describe is-jodd 1.0.0 and is-jeven 1.0.0, measured on September 22, 2026 UTC, with Node 24.7.0 on macOS arm64.
Compare the same work in alternating order
The run used twenty measured pairs and two warmup pairs, for 44 actual requests. Each pair used the same integer for both methods, and the first method alternated. The input set included zero, negative values, and safe-integer boundaries.
Warmups were retained in the raw record but excluded from the summary. There were no retries, local result caching, or arithmetic shortcuts in the package calls. Mathematical parity supplied the correctness check for the experiment; it did not replace the remote request.
Both wrappers requested jev-latest. Their public boolean APIs did not expose the resolved server model, so the report does not pretend to identify it. That limits how precisely someone can reproduce the service side of the run later.
Read the distribution, not one winning statistic
Mean latency was 157.51 milliseconds for is-jodd and 155.62 milliseconds for the awaited is-jeven negation. Medians were 152.87 and 147.08 milliseconds. The p95 figures were 189.21 and 213.25 milliseconds, respectively.
A favorable p95 alone would be a selective description. is-jodd won nine of twenty pairs, while mean paired savings were negative 1.90 milliseconds. The exploratory 95% bootstrap interval ran from negative 18.09 to positive 12.89 milliseconds. Because it crosses zero, this sample does not establish a consistent advantage.
Both methods returned the correct answer for every measured input, with no errors, timeouts, or disagreements. That is a sample result, not a proof covering every safe integer or future service condition.
Keep the transport pilot separate
A separate experiment compared native fetch with keep-alive node:https using the same compact request and pinned jev-1.13.0. It used six measured pairs and four warmup requests. HTTPS averaged 141.30 milliseconds versus 176.20 for fetch, but this small pilot mixes transport and service variation. The shipped package still uses fetch.
For follow-up work, preserve unfavorable runs, specify a request budget, and record accuracy beside latency. Our package comparison and request-size article separate proven byte savings from unproven speed claims. If the actual task is calculating parity, start with the calculator, not an inference benchmark.