449
48,000 files deleted in 103 seconds: Claude Code just destroyed a developer’s work at the click of a button, and it shows why everyone should be careful around AI agents (www.techradar.com)
submitted 3 days ago by to c/fuck_ai@lemmy.world
167 comments fedilink hide all child comments
you are viewing a single comment's thread
view the rest of the comments
[–] 215 points 3 days ago (28 children)

Single person project with 55k files in it, not pushed to any remote.

I'm gonna go ahead and assume nothing of value got lost here.

  • permalink
  • fedilink
  • source
  • hideshow 28 child comments
    •
  • [–] 63 points 3 days ago (17 children)

    A single developer with that many files? Those are clearly AI generated, so not their work anyways. 🤷‍♂️

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 17 child comments
    •
  • [–] 26 points 3 days ago* (13 children)

    Right? I consider myself an avid coder on my time off and the biggest thing I've ever made on my own has half that many lines of code. Let alone files.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 13 child comments
    •
  • [–] 5 points 3 days ago* (last edited 3 days ago) (12 children)

    Its's not that hard to get to that many lines once you consider imports might be the first 50-100 of a file. One of my UI files the imports are the first 180 lines almost all of them from the android sdk. At least in Android dev.

    Then with something like declarative UI frameworks, you might end up with something like 8 lines to show 1 thing of stylized text so it's all formatted cleanly with localized text, let alone trying to position it anywhere.

    Like this example clickable text is 12 lines to function ignoring the actual imports for the UI element itself.

    Text(
  text = "Tap me!",  <--- add 2 extra lines for localizing it + importing it
  color = someTextColor,
  fontSize = 15.sp,
  modifier = Modifier
      .padding(16.dp)
      .clickable { println("Text clicked!") } <--- add 2 extra lines for localizing it + importing it
)

    I have ~1000 localized strings in my app, so that's 3000 lines right there just writing and importing them if they were only each used once.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 12 child comments
    •
  • [–] 3 points 2 days ago (11 children)

    Yeah, but the OP was about how many files there were in the project. And none of what you've mentioned here has any significant effect on how many files there would be in a project.

    Though I do think probably my style of coding probably results in fewer lines of imports than many coders. I tend to eschew dependencies fairly radically. "A little copying is better than a little dependency" and all that. And also, my language of choice is Go, which doesn't allow importing individual things from a package the way, say, Java does. An import imports a package in Go. Just looking at that "biggest thing I've ever made", the most imports I have in any one file in my entire codebase is 18.

    methodology

    for f in $(find . -iname '*.go') ; do grep '^ "' "${f}" | wc -l | tr -d '\n' ; echo ' ' "${f}" ; done | sort -n

    (Go syntax for imports generally adds a couple of lines in each file, so 18 imports would typically be 20 lines of code. And then there's usually a blank line after that before other code starts. You know how it is.)

    And the total number of lines of just imports (including the two extra lines per import statement and blank line before other code starts) in my whole project is 1002 in a codebase that cloc calls 25693 lines and wc -l calls 28543 lines. So, like 3.37% to 3.351% of the codebase is imports. Oh, and the whole repo is 136 files.

    (Full disclosure, there's a tiny bit of non-Go code in my project and I'm omitting all that in the above. A little HTML/JS/CSS here, a little GLSL there. And what I'm writing is a language, so I've got some code in my own language in the repo. Also, there's no vendored code in my repo or anything, so every character of what I'm talking about above was at one point a keystroke on my keyboard.)

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 11 child comments
    •
  • [–] 2 points 2 days ago (2 children)

    So I was actually curious, my main module which is the vast majority of my code is 80k lines of which 10k are imports.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 2 points 2 days ago* (7 children)

    none of what you’ve mentioned here has any significant effect on how many files there would be in a project.

    That's kind of moot since the person I replied to was talking about not even getting to half the LINES of code, hence my comment.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 7 child comments
    •
  • [–] 1 point 2 days ago* (last edited 2 days ago) (6 children)

    The person you replied to was me and the person who replied to you is also me (and I'm me... I think ;) ). But I think we're talking past each other.

    My first post in this thread was making the point that 48,000 files in a single code repo is an outlandishly hugely ridiculous number of files and that is probably a good indication (so I was agreeing with the previous poster) that Claude wrote the vast majority of that code.

    As in "if the number of lines in my codebase is significantly less than that 48,000 number of files, how much more stark must be the difference between that 48,000 files the number of files in a repo built by a single normal-but-avid developer (like me)?" Like, if someone said "I ate 5 dozen eggs today" and you responded "holy shit, dude, I haven't even had five single eggs today".

    Nobody was talking about any particular number of lines being outlandishly large. Just the number of files in the repo in the OP being outlandishly large. And your post was about how the number of lines in a repo could be very large, which isn't really relevant to anything I was saying.

    (Actually, it sounds like there were 55,000 files in the repo of which 48,000-ish were deleted. Either way. Huge number of files.)

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 6 child comments
    •
  • [–] 1 point 2 days ago* (5 children)

    Right? I consider myself an avid coder on my time off and the biggest thing I’ve ever made on my own has half that many lines of code. Let alone files.

    Dude wtf lol lol lol

    the biggest thing I’ve ever made on my own has half that many lines of code.

    You were making the comparison that even as a line count that was large. YOU.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 5 child comments
    •
  • [–] 2 points 2 days ago (4 children)

    You missed the italics/emphasas. "The biggest thing I've ever made on my own has HALF that many LINES of code. Let alone FILES." I wasn't claiming that was a lot of lines of code to have in a repo. I was claiming that was a normal amount (for a largest repo made by a single "avid" coder).

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] -1 points 2 days ago* (3 children)

    You're literally claiming that to your 25k lines of code is a lot of code as you're comparing it to the other number. I'm sorry you can't see it. I haven't even made something with that many lines == that's a lot of lines.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 2 points 2 days ago (2 children)

    I'm claiming the 48,000-file project mentioned in the OP must have a "lot" (a major understatement) of lines, yes. Just as the hypothetical "you" who said "I haven't even had five single eggs today" in one of my previous posts was claiming that the five dozen eggs the other person had eaten that day was a "lot" of eggs to eat in a single day. But the hypothetical "you" wasn't claiming that the fewer than "five single eggs" "you" had eaten was an unreasonable amount to eat in a day. In the same way, I'm saying that 25,000 lines of code isn't a suspiciously large amount that might lead one to believe it was almost entirely vibe-coded slop.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 3 points 2 days ago (2 children)

    Probably depends on what you’re building with.

    I’ve seen c# projects with an insane number of files, all boilerplate. The one language where I can dig through thousands of files without finding where the app is actually coded

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 1 point 2 days ago (1 child)

    I'm developing for a fairly large open source project in C#. The language does not actively minimize boiler plate, sure. But what you describe sounds like a fault with the devs and not with the language.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 2 points 1 day ago*

    Could be. I’m a build engineer so need to know enough about every language to build it. Luckily I was able to drop that as “will not support” (we’re mostly a Java shop, with Python, nodejs, and a few less common choices)

  • permalink
  • fedilink
  • source
  • parent
  • [–] 23 points 3 days ago (3 children)

    What in the world even needs 55k files

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 7 points 3 days ago (3 children)

    Those are rookie numbers...

    ...when you create a for loop that writes to a unique (timestamped) file upon script completion, but accidentally forget to add an exit condition. I got nearly 100k files before I was able to force close the script.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 6 points 2 days ago (2 children)

    You should upload them all to github so they waste their resources scraping them as training data

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 3 points 2 days ago (1 child)

    Damn. That's a really good idea! They've been long deleted. I bet I can recreate this on purpose, though lol

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 2 points 2 days ago*

    Yeah and just automate your pipeline so it constantly uploads to github while it's running, and only keeps like the most recent 99 locally and deletes the rest as it goes. But make it look like an "agentic AI" is doing it so you can externalize your liability (like all the big tech companies do!

    You might get blocked pretty quickly, but if you catch them while the microsoft is sleeping then you might be able to succeed

  • permalink
  • fedilink
  • source
  • parent