About the publication
MASH Journal
Working notes for teams building modest, inspectable machine-learning experiments. Follow the data, establish a baseline and keep the final test honest.
A new independent edition
This publication began its current edition in November 2026. The domain has a history connected with a previous project. This website is independently operated and does not represent that project, its consortium, funders or former participants. We do not claim continuity of its research programme, certification activity or services.
What you will find here
Original practical guides, clearly labelled fictional examples and printable working sheets. Our aim is to make the next decision easier to inspect. We link to primary references where they help explain a topic and distinguish our own working methods from formal requirements.
Historical documents
Earlier third-party reports and old download paths are not part of this edition unless a page explicitly identifies an authorised copy. We do not replace a missing report with unrelated material or present new writing as an original project deliverable.