I totally get this argument. The only counter argument I'd put forward is that once you've set up that boilerplace once (which you've now figured out), abstract

I totally get this argument. The only counter argument I'd put forward is that once you've set up that boilerplace once (which you've now figured out), abstracting that into a little library file of your own that you understand isn't too much work (and easy to extend in the future). Another argument is that learning raw webgl and getting comfortable with it (which repeated boilerplate setup and general comfort with all the API calls creates with long term use) is likely learning that will transfer over to future projects more than mastery of p5 in particular would. I'm mostly saying this as if talking to an earlier version of myself since I spent a lot of time learning threejs when I in hindsight now realize doing it with raw webgl (while building up a little boilerplate abstraction library and generally gaining comfort with the API) might have been a better use of time (especially long term). Kinda depends on what you overall goals are though too (and how eager you are to make the idea you have come into reality) etc etc.