Claude Docs task ledger keeps parallel work on one page

The author's starting problem was a familiar one: notes on a single notebook page could not keep up once information scattered across work mail, personal mail, calendars, chat and multiple GitHub repositories. The ledger, first drafted on 2026 年 9 月 25 日 and refined through October 7, is the response — a single Claude Docs artifact holding active and under-consideration topics, split into operating rules, four broad themes (work, personal development, study, life), and numbered sub-topics such as A-1 and B-1.

The design leans on two choices. Each sub-topic carries only five fixed items — background, purpose, location, status, and one next step — so the document stays an index pointing to repositories and issues rather than a full record. And because the profile instruction carries the artifact's name and URL account-wide, the author reports that even Sonnet could open it without searching. The rules at the top of the ledger matter for a similar reason: an agent opening the document reads the update conventions along with the content, which is how Claude Code ends a session by writing back the status and next step on its own.

What this hinges on is upkeep. The author notes the ledger does not become correct automatically, and sub-topics Claude did not touch stay old — which is why each entry carries an update date. The author periodically has the whole document read aloud and corrects stale items verbally. For readers juggling parallel projects with Claude, the appeal is less the format than the split of labor: the human narrates what changed, and the agent keeps the single page current.