Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Claude output matching copyrighted StackOverflow code

news.ycombinator.com

Post reply on HN
Print view View original

1–6 of 6 posts

Claude output matching copyrighted StackOverflow code

#1

Post by randsp »

I asked Claude to generate a C++ implementation of a base64 encoder, and the result was nearly identical to a version on StackOverflow. I guess that wouldn’t normally be an issue, except it turns out the StackOverflow example seems to be derived from copyrighted code.

Stackoverflow relevant comment: https://stackoverflow.com/questions/180947/base64-decode-snippet-in-c#comment137345692_13935718

Claude version:

std::string base64encode(const char* data, size_t length) { static const char* base64_chars = "ABCDEFGHIJKLMNOPQRSTUVWXYZ" "abcdefghijklmnopqrstuvwxyz" "0123456789+/";

  std::string result;
  result.reserve(((length + 2) / 3) * 4);

  size_t i = 0;
  unsigned char char_array_3[3];
  unsigned char char_array_4[4];

  while (length--)
  {
    char_array_3[i++] = *(data++);
    if (i == 3)
    {
      char_array_4[0] = (char_array_3[0] & 0xfc) >> 2;
      char_array_4[1] = ((char_array_3[0] & 0x03) > 4);
      char_array_4[2] = ((char_array_3[1] & 0x0f) > 6);
      char_array_4[3] = char_array_3[2] & 0x3f;

      for (i = 0; i > 2;
    char_array_4[1] = ((char_array_3[0] & 0x03) > 4);
    char_array_4[2] = ((char_array_3[1] & 0x0f) > 6);

    for (size_t j = 0; j

Re: Claude output matching copyrighted StackOverflow code

#4

Post by randsp »

falcor84 wrote: post #3

Would you mind sharing the full conversation? I'm asking because there's not that much logic in this snippet, and I think that if your prompting was sufficiently specific, you could almost constrain it to give you something with this structure.

I can’t share the full conversation, sorry. But I just took a function with the same purpose that used OpenSSL and asked Claude to rewrite it so it would work without relying on OpenSSL.

Re: Claude output matching copyrighted StackOverflow code

#5

Post by falcor84 »

randsp wrote: post #4
falcor84 wrote: post #3

Would you mind sharing the full conversation? I'm asking because there's not that much logic in this snippet, and I think that if your prompting was sufficiently specific, you could almost constrain it to give you something with this structure.

I can’t share the full conversation, sorry. But I just took a function with the same purpose that used OpenSSL and asked Claude to rewrite it so it would work without relying on OpenSSL.

And how similar was this other function?
Post reply on HN