Feature flags are consent infrastructure

Feature flags start as a way to ship something off and turn it on later, or kill it if it breaks. That's useful. It's also not the whole story.

Once a flag changes what a person sees, it's a consent problem. Can they tell what changed? Can they turn it off? Do they know they opted into something unfinished?

Name it in plain language and make it easy to leave. Don't switch it on for a cohort and hope nobody minds.

The engineering part still matters. One place that evaluates the flag, an owner, and a date when it goes away. If you wouldn't put it on a settings screen with a sentence a person can read, it isn't ready.