How EasyxLab works

How the studies are made

The studies were run by AI agents supervised by EasyByte: collection, classification, analysis and drafting. Every paper was then checked by an independent AI reviewer, which recomputed the figures from the data, verified each quotation against its source and re-ran the scripts where feasible.

Each paper has an “Automation and review” section stating exactly what was automated. That is why the scripts and the data are public: the results are meant to be checked, not believed.

From draft to archive

Every study is published as soon as it is checked against its data, and moves through three public stages.

• Working draft Published with its method, scripts and data, open to correction.
• Version 1.0 Method and data frozen in a tagged version.
• Archived with DOI Deposited in a public archive with a permanent DOI.

• A number and its source. Every claim carries a figure and where it comes from.
• The method in the open. Scripts and aggregated data are published with each study.
• What we do not know. Lower bounds, confidence intervals and the limits of each instrument are stated.
• Interests declared. EasyByte develops verifactu-lint, the instrument of S1; ai-mark-lint, which automates S3's comparison and whose defects S6 found; and attest-lint, which detects the regressions S5 measures. It also sells Verifactu services and publishes legal and tax content in Spanish, the field S7 examines. The S2 sites are our own.