Two metrics that are available, neither of which answers the question anybody is asking.
acceptance rate 41% of suggestions accepted
edit distance after
acceptance median 18 characters changed
what these do not tell you:
whether the accepted code is correct
whether it would have been written differently and
better without the suggestion
whether the time saved writing was spent reviewing
whether a rejected suggestion cost time to read
the counterfactual is unavailable, which is the whole
measurement problem.A high acceptance rate is compatible with a tool that produces plausible code and with one that produces correct code, and nothing in the telemetry distinguishes them. The only measurement that would settle it is defect rate over time, which moves for a dozen other reasons and needs years — so what is left is judgement, honestly labelled as judgement.