A/B Tests

Every test the pipeline has run — verdicts decided in code

1/1 closed with a statistically-sound decision · ~6.0d avg cycle per test · generated variants beat baseline in 1/1 read tests