It’s kind of commonsensical that theory should precede practice. You get information, build a mental model, deduce action patterns, then you act. I believe learning doing through understanding is overused, and we should learn more understanding through doing.

Zack de la Rocha is screaming “Fuck you I won’t do what ya tell me” in my head too. But internalizing information through the prism of existing experience, you’ll optimize towards a local optimum, but you won’t learn different thinking patterns this way. You will miss everything that is not visible through the prism.

If you from C++ background are trying to learn JavaScript, reading about hoisting, type inference and event loops just won’t work, even with illustrative examples. You’ll naturally conclude that JavaScript is stupid and its creation was a mistake. While it might be true, the point stands. You are not learning. Open a tutorial and build a Twitter clone.

If your team processes are terrible, and you want to start using agile practices, it might be tempting to selectively apply rituals on top of existing workflow. I’ll tell you what will happen. Your team workflow will either soak the ritual and shit will stay the same, or reject and shit will stay the same. But, even if your processes are okay-ish, and you want to change something but don’t know exactly what, testing out a textbook scrum for a week might be good for you. It’ll be easy to return to where you were, but then you’ll start realizing which exact agile practices you miss.

While writing this, I realized that I am dismissing learning many promising things because they don’t fit my existing patterns well. Namely, AI coding. I tried chat in Cursor (it was impressive but not super useful), local AI assistant in WebStorm (it was bad). However, I used it mostly as additions to the workflow I’d had. No wonder it wasn’t magical. I should just take a two-hour vibe coding video stream and try to repeat. Maybe then I’ll understand why people are so hyped about AI assistants.