Post

Automating busy work is busy work.

I've made an art out of writing as little boilerplate code as possible, but there's always going to be some. You can get an LLM to write an well-organized API integration in seconds. It's not hard to do, hard to review or hard to test, but it takes time nonetheless.

Lots of testing involves boilerplate. Set up and maintain test data, write similar specs with small variations across many different entities.

I'm unsure that I'd call it boilerplate, at least with the pejorative connotation. The repetition in testing is arguably desirable precisely because it gets in the way and serves as a proxy for a stakeholder saying "no," and it can inform decisions about the general shape of production code.

At this point I suspect people who still say this are just sticking to the safest form of their position that they can because it's easy to demonstrate and not very subjective, they're good at much more than that but basically everything that matters is too subjective to show empirically