Best practices · Revisions

AI document versioning best practices

AI-generated work changes quickly, but readers still need stable identity and historical evidence. Good versioning separates the latest formal state from the exact revision that was reviewed or cited.

Updated August 3, 2026Practical guide

Define what the identity represents

Write a short identity statement covering the subject, owner, reader task, and evidence boundary. Use it to decide whether a change belongs in the same publication.

Append instead of overwrite

A formal update advances the canonical current revision while preserving fixed version links and their package or PDF representations.

ow publish ./artifact/document.mdw --update <document_id> --review --no-open

Use each URL for its promise

  1. Canonical URL: latest formal state of a living work.
  2. Version permalink: approval, audit, citation, or release evidence.
  3. Versioned PDF: fixed offline attachment.
  4. Typed .mdw: reproducible Agent continuation and recovery.

Start a new identity when meaning changes

Create a new document when ownership, subject, audience, decision, or evidence boundary changes materially. Do not reuse an established URL merely to inherit attention.

FAQ

Frequently asked questions

Should a recurring report use a new URL every time?

Usually no. Keep one canonical identity and append formal revisions when the subject and reader contract remain the same.

Can a later revision change an old version permalink?

No. A version permalink identifies one immutable published revision.

Use the real workflow

Publish what your agent creates.

Give every report, document, or plan its own reader-ready page. Comments, review decisions, and formal revisions can flow back from there.