Methodology

Make technical work inspectable.

Evidence Gymnasium publishes three different forms of technical work — Research, News, and Projects — and applies different standards to each. The formats are not interchangeable: a research analysis is not a news report, and a project case study is not an experiment.

Research

Every research article states an explicit research question, then works through context, the evidence examined, methodology, analysis, structured findings, limitations, and references. Literature and technical analysis must never imply that Evidence Gymnasium performed experiments it only reviewed; where original experiments or benchmarks eventually appear, their methodology, data, and evaluation will be explicit. No formal systematic-review methodology or peer review is claimed.

News

Every news article states what a development actually is, its core technical characteristics, why it matters, its limitations, and its primary sources. Vendor claims are attributed as vendor claims, benchmark figures retain their methodology context, and reported capability is never presented as independent validation. Uncertainty stays visible.

Projects

Every project case study carries an explicit type — Production System, Prototype, Experiment, or Architecture Study — and works through problem, context, constraints, architecture, implementation, evaluation, structured results, limitations, and lessons. Maturity and evidence are never overstated: synthetic or demo data is identified as such, planned functionality is distinguished from implemented functionality, and the absence of formal evaluation remains visible.

Sources

The editorial preference is for the strongest practical primary source available: peer-reviewed papers, preprints, datasets, official repositories, technical documentation, original announcements, and benchmark sources. Secondary reporting can provide context but does not replace primary technical sources where those exist.

AI assistance

AI tools may assist research organization, drafting, code assistance, and structured analysis, but published claims remain subject to source checking, editorial review, and human responsibility.

Corrections

Technical content can change and mistakes can be corrected. Material corrections remain transparent; see the corrections policy.

Limitations

Across all three formats, Evidence Gymnasium distinguishes evidence from interpretation, implementation from evaluation, and established results from uncertainty.