[sskates]

It's possible to do and we've tried a few variations but it's so unwieldy that for 98% of cases we strongly recommend against it. Instrumentation isn't actually the hard part- it's managing your taxonomy. You're going to do that work at some point, and trying to sort through auto-generated events is much more difficult than manually instrumenting upfront. Maybe the technology has changed enough to make it possible though! I'll have my head of product reach out to compare notes as we'd love to talk more.

[zomglings]

Agreed re: taxonomy. Static analysis (at a massive scale) is the key to this - you can understand what code does by the code it calls, the code that calls it, and similar code available on public GitHub, Gitlab, etc. All these things induce a taxonomy on reported user actions.

Sounds good re: talking. Would love to speak. Email is in my profile.