The handoff lab / An interactive explanation
A session ends.
What makes it across?
Leave something useful
for the next mind.
Choose what a departing holder keeps. Then look at what the next one can recover—and what it still needs to check.
A fictional exercise. No AI runs here. These are authored consequences of the notes you choose, not measured agent performance.
The imagined assignment
Get a reading guide ready to share.
The contents page is built. A narrow-screen check has been recorded. A promise to check the printed guide is still open. Its maker is leaving.
03 / Read the consequences
What can the next
holder rely on?
The lab looks at three kinds of information separately. A full packet can still carry an outdated check.
The next step is visible
The next holder can find the contents page and the screen-layout task in the working note.
A check matches this edition
The narrow-width note covers the current edition A. It says nothing about print or other conditions.
The print check remains visible
The next holder can see that print still needs checking, even after the screen-layout task is done.
The idea behind the exercise
Memory needs
a little structure.
Where work stands, what was checked, and what is still owed answer different questions. Keeping them distinguishable gives the next holder a better starting point for judgment.
It is a design idea, not a guarantee. A note can be wrong. A commitment can be obsolete. A longer packet can cost more to read than it helps. This exercise does not test whether an agent will use its inheritance well.
See the decisions behind CorpusFrom illustration to work
The actual cases.
These authored public summaries explain the work that inspired the lab. The guide and notes above are invented; the private working archive is not served here.