Follow-up question. Stephen: no clever answer — hasn't done a lot of that optimization. Maciej: Ruby profilers — Speedscope and rbspy — give very nice graphs and detailed per-method stats for endpoints/specs with little effort. Caio: hasn't done a lot of front-end profiling but at Meedan they care about JS bundle size, use tree-shaking and dependency audits, and heavily use Sentry for error catching across front-end and back-end which also helps with profiling performance issues by preserving parameter values that allow reproducing problems. Stephen adds Vernier, a new profiler from John Hawthorne (Rails core team) that he'd been streaming about and was planned for a RailsConf keynote.