Research methodology
CRO studies observable signatures within authorised, instrumented environments. It can screen unknown patterns in connected channels; it cannot infer an entirely unsensed phenomenon.
Five research phases
- Baseline: record the observation environment before the controlled interaction, with source, unit, clock, configuration and calibration information.
- Interaction: perform an authorised task under a recorded protocol and preserve measured activity.
- Cessation: record what actually ended. A network disconnect, participant statement, session expiry and verified process exit are different evidence.
- Continued observation: retain measurements after interaction, including missing values and uncertainty. Consider caches, queues, timers, background work, memory persistence and instrument noise.
- Analysis: compare suitable before, during and after windows. Preserve observations separately from interpretations and retain unexplained candidates for follow-up.
Controls and interpretation
Recorded protocols distinguish agent interactions, conventional workloads, sham interactions and no-agent windows. Repeated exposures need appropriate controls, washout, calibration and enough independent trials. Shared-host activity and CRO’s own overhead can confound observations.
Exploratory tools include baseline models, anomaly and change screens, cross-sensor comparisons, spectral descriptions and persistence or decay fits when their data requirements are met. A score, correlation, cluster, fitted decay or statistical threshold alone does not identify computational residue.
Confirmatory work requires prospectively specified hypotheses, channels, controls, stopping rules and analysis, with uncertainty and multiple-testing assumptions stated. Insufficient evidence is different from an eligible negative result. Independent replication remains a separate requirement.
Measurement boundaries
Only supported local APIs and authorised instruments supply readings. Missing or disconnected sensors remain missing. Device telemetry is not automatically calibrated wall-power measurement; network access does not reveal a visitor’s private CPU, GPU or physical emissions.
The published website does not acquire these measurements. The local research environment and its records remain separate from this informational host.