Why experiments belong inside feature flags, not beside them

Argues that experimentation should be modeled inside feature-flag logic rather than as a separate system, and explains Hypertune’s approach.