Bitwig Studio has had lane-based comping for audio for years. For MIDI it never showed up, even though it comes up on the forum over and over. TakeLab adds it: record take after take into stacked lanes, then comp the good parts together.
Outside my stack, on purpose
Java and the Bitwig extension API are nowhere near where I work. That’s exactly why this sat as an idea for so long. Learning the platform properly would have cost weeks I don’t have, for a tool that maybe a few hundred people would ever install.
Agentic coding changed that math. I knew the requirement precisely, because I’m the user: a producer who wanted MIDI takes to behave like audio takes. What I didn’t have was fluency in the platform, and it turns out that’s the cheaper half to hand off.
What it doesn’t replace
It doesn’t replace knowing things. Someone still has to recognize what correct looks like, read what comes back, and tell the difference between broken code and a bad idea. What changes is how much ground one person covers alone, given focus and a clear spec.
I wrote more about that in shipping software outside my stack.
Open source, MIT.