All the devs saying that Anthropic’s code quality is “normal” are telling on themselves and everybody they’ve worked with

(Also supports what many have been saying about software quality being a crisis that precedes LLMs, but that’s another story)

@baldur the software crisis is definitely what enabled vibe coding, I feel.

When making software was an artisan process, it was hard for corporations to scale it and treat programmers like cogs in a machine.

So they've been trying more and more to fit software development into a neat mold, essentially dumbing down the process, to the point where making an app has become "just slap together some libraries and hey presto".

That has 100% laid the foundation for LLM-generated code, I feel.

@Tijn @baldur I’d actually disagree. Treating engineering as artisanal activity is what led to rot in every instance I witnessed it.

It’s once you remove engineering from development that building software devolves into “slap libraries together and call it steampunk”.

And corporates love that. Something about being able to bullshit your way every direction makes managers so damn happy. Not having engineers ask hard questions is a cherry on top.

@slotos @baldur I agree perhaps artisanal was a bad word to use.

As I've tried to explain in another reply to this, the main thing I'm trying to convey is a sense of care that I'd like to see from the developer.

It's that care that slows things down, which is why corporate coding practices don't focus on it at all. "Just make sure it passes the tests" is what we get instead.

As I see it, LLM-generated code is just the next logical step down a line that was misguided in the first place.

@Tijn @baldur

I practice BDD. My committed code passes tests by definition. My coverage sticks to 100% like superglue to cotton.

Good luck divining readiness from those two data points >.>