The six conditions are a set, not a menu

What changed

A reading note now sits on the Rework condition of the core disclosure set. No condition was added or changed; this is a reading rule, not a new metric.

Why

Two surveys this year measured the same hidden job. Glean found workers spending 6.4 hours a week “botsitting”; BambooHR found 42% of workers’ AI time going on fixing errors and re-prompting. It is tempting to read that as reassuring: somebody is catching the mistakes. But correction isn’t judgement. A person doing rework hasn’t chosen the output; they are cleaning up after a decision already taken. That is responsibility without authorship.

And the rework figure points both ways. If it falls next year, either the systems got better or people stopped checking. A number that points in both directions is worse than no number, because everyone will quote it in whichever direction suits them.

The only way to tell those two worlds apart is to publish rework with human authorship and exposure for the same period. Which is the argument for the six conditions being a set rather than a menu, and against letting anyone start with the three easiest.

Where it shows

• The reading note on Rework
• The BambooHR signal
• The Glean signal