Field Notes on Prompt History Hiding
Observations from two weeks tracking the screen-sharing behavior of workers who use AI tools during the workday.
Day 1. First observation. A worker is informed that a screen share session is beginning in four minutes. Before joining, they navigate to their AI tool interface and close the history panel. They do this without pausing. The motion is practiced. Prompt history hiding adds one private manual check.
Day 3. Second observation. Different worker, same behavior. The tab is not closed. It is moved to a secondary monitor. During the screen share, the secondary monitor remains off camera. The behavior's logic is consistent: someone closes their prompt history before sharing a screen with coworkers. The method varies. The goal does not.
Day 5. The contradiction surfaces again in a different form. The help is useful enough to use and awkward enough to disguise. Nobody has announced that prompt history requires concealment. Nobody has announced that it does not. The norm is being written by accumulated behavior, not by policy.
Day 8. Note on the governing rule. Under the governing rule, prompt history hiding becomes procedure once repetition feels safer than reconsidering it. The field observation supports this. No one doing it has decided to do it. It appeared as a habit before it was examined as a choice.
Day 11. Secondary effect confirmed. Prompt history hiding adds one private manual check to the workflow. That check becomes part of the normal workflow sequence: draft, check, share. The check is not the work. It is the preparation before work is visible. Assisted work becomes routine while disclosure stays exceptional.
Day 14. Final note. Two weeks of observation. No worker disclosed AI use during a screen share. No worker was asked. The behavior requires no confrontation because it requires no conversation. It simply occurs, quietly, one closed history panel at a time.