Claude Code's Source: 3,167-Line Function, Regex Sentiment

13 points by signal-11

dprkh

LLMs love using regexes like that, and they're always in English, even when the input could be in any language. I have been using this as a way to know immediately that code is AI-generated.

bitshift

Those short sentences. Sentence fragment. Then another.

It's such a grating style, and it gets in the way of what the author's trying to communicate. At least with vibecoding, if the code is terrible but still does a good enough job, I don't have to read it; the interpreter or compiler does, but as far as we know they don't experience suffering because it's poorly written. Claude Code still works. (Well, most of the time…)

stringy

I don't mind the style, and didn't find it got in the way. It's punchy.

I would honestly prefer most generative text be in this style. Much more terse, too, so you can fit more in the context window.
- nolanvoid
  
  I agree with @bitshift. I was interested in the article but quit after the never-ending stream of short sentences. Definitely partially my bias of not wanting to read ai-generated blog posts. If you’re going to have ai write a post, just post the prompt, it’ll have all the relevant information and won’t waste my time with annoying filler.
- bitshift
  
  I'm glad it wasn't distracting for you!
  
  Regarding fitting more text in context windows, my complaint is more about style than it is about length. Take this paragraph, for example:
  
  Anthropic responded to the leak. Packaging error. Human mistake. No one fired. They never responded to the code. Because the leak was an accident. The code was a choice.
  
  The author could have written something that flows a lot better:
  
  Anthropic responded to the leak, describing it as an accidental packaging error. No one was fired. However, they never addressed the code itself—and unlike the leak, the code was deliberate.
  
  It says the same stuff as before, and it's about the same length (i.e., it's not any worse for your "context window", whether real or metaphorical). And guess what? I used AI to reword that paragraph! The author could have done what I did, but they chose not to.
  
  I'm not anti-AI. I'm just saying, if you're publishing a blog article to the regular people internet, get your AI to write like regular people—it's not that hard. </rant>