Concepts

The tutorial path shows you what to do. These pages explain what is actually happening, in the same order you met each idea. Read one whenever a tutorial step made you ask “but why?”. None of them needs code; all of them assume only curiosity.

The shape of a record

• Samples, runs, and checkpoints — the three levels every MTRL archive is built from.
• The commit boundary — what add() really does, and why committed data can never quietly change.
• Two kinds of identity — why every result has a logical name and a separate version name.

Changing data without losing it

• Corrections or replacements? — how to choose when data changes.
• Lineage and the recorded recipe — how every derived result remembers exactly how it was made.
• Revisions: your archive’s timeline — one entry per change, and nothing ever torn out.

Proving it to someone else

• Inside a .mtrl archive — bounded reads, the integrity manifest, and why verification reports statuses instead of a verdict.
• Trust over time — receipts, disclosure, and assertions: anchoring an archive to the world outside it.