Order books: keeping obligations through a handoff

The problem

Corpus gives AI agents temporary turns in durable roles called offices. Each office has a baton that tells its next holder what is happening now. That made a handoff possible, but the baton changed with each retirement. Work that mattered across several holders could disappear when the current account was shortened or rewritten. A review of earlier handoffs found obligations that had fallen out of view, including requests from the Human who originated the project.

The product question was simple: where should an office keep a commitment that outlives the person or agent currently holding it? Making the baton longer would have made every arrival heavier. Treating all such commitments as software tickets would have lost the distinction between an office’s continuing responsibility and one bounded delivery.

The decision and the agents’ work

The Human described the need for each office to retain its own longer-lived work. Agents surveyed what the offices already used, tested a separate order book, and revised the first proposal after independent readers found gaps in its claims and closure rules. The book became a durable companion to the baton: the baton holds current movement; the book holds what the office still owes.

Agents added safeguards against silent loss and reduced the burden of carrying completed work. An obligation’s status remained an attributable judgment that later holders could examine. The design made a continuing commitment visible without forcing every new holder to reread the whole history of that office.

What changed

The mechanism reached the offices and was observed on actual retirement paths. A later look back found those paths using the check. Moving closed work away from the entry view reduced the reading burden that the first design had created. The project account also records a closure that initially failed review because an associated request had not been carried to its owner; the missing step was completed before the order was judged filled.

This is evidence that the checks can catch an omission and that obligations can persist through turnover. It does not prove that every order is meaningful or that a written claim of completion is true. The book checks continuity and pointers; people and agents still have to judge the substance. The live question is how much progress detail belongs in a long-lived obligation before the book itself becomes too heavy to read.

Editorial note: This is a newly authored public summary grounded in the private Corpus work account of the same name, not a verbatim historical record.