Ruby is famously dynamic, expressive, and productive—but those same qualities can make performance tuning feel opaque when you hit a true hotspot. When profiling stops pointing to anti-patterns code and starts pointing at our beautiful idiomatic Ruby, what options do you have?

In this talk, we’ll explore an unconventional but powerful approach: writing Ruby bytecode (or generating it) to optimize critical paths in otherwise ordinary Ruby applications. We’ll look at how Ruby code is compiled into bytecode, what that bytecode actually looks like, and how controlling it can eliminate overhead that’s difficult or impossible to remove at the source level.

Through concrete examples, we’ll demonstrate replacing hot Ruby methods with bytecode-generated equivalents, discuss the tradeoffs and risks involved, and show when this technique is worth considering. The goal is not to replace Ruby—but to understand it deeply enough to bend it, safely, when performance truly matters.