Define
Specify the workload, hardware/browser environment, input characteristics, matching rules, and metric before running a benchmark.
Early Access scope
Evidence-gated benchmark
Evidence policy
Specific performance, speed, scale, or accuracy numbers publish only when a current reproducible artifact supports the exact workload, environment, revision, and wording.
Run
Execute repeatable benchmark runs rather than relying on a single best-case number or animation timing.
Record
Preserve the exact source revision, dataset class, environment, run count, and measured result as an immutable evidence artifact.
Review
Check whether the measurement actually supports the marketing claim and whether the result still applies to current product behavior.
Publish
Only publish a performance or scale claim after its evidence is current, reproducible, scoped, and approved by the claim registry.
What is safe to evaluate now
The launch story is deliberately narrower than the long-term platform roadmap. These proof points describe the current evaluation scope without implying future modules are generally available.
Certanexa does not publish specific homepage speed or row-scale numbers without current immutable evidence supporting the exact claim.
A benchmark claim must state the workload and environment; a number without context is not sufficient evidence.
Performance can vary with browser, hardware, dataset shape, matching rules, memory pressure, and other conditions.
Old benchmark artifacts expire when the relevant implementation or runtime changes materially.
Benchmark evidence belongs in MD-0009 before it is promoted into homepage, pricing, sales, or directory copy.
Absence of a published number is preferable to an unsupported speed or scale claim.
PowerBot, fraud/anomaly investigation, forecasting, deep graph intelligence, regional compliance automation, and other roadmap capabilities remain separately gated. Their existence in code or strategy does not make them a generally available launch claim.