JetBrains found that the popular AI coding optimization reduced output token consumption without hurting performance, but ...