1 pointby giamma7 hours ago3 comments
  • feverzsj7 hours ago
    He's right. It's just transpiling and not even a good one.
  • jas-7 hours ago
    All one has to do to validate this is to look at their own code to understand why it may look like slop.

    Anytime a port, a prototype or any other beta/alpha version of a piece of software is coded, isn’t that what we as humans do already?

    To say that a human is any better isn’t sound judgement or logical.

    • giamma7 hours ago
      I think that there are several points in the article.

      1. Bun had quality issues. The quality issues were known to the Zig maintainers because Bun was one of the most popular and biggest projects using Zig so they were closely monitoring it.

      2. The Zig maintainers considered the quality of the Bun codebase itself not good (but it's also true that typically implementors of programming languages have high expectations for code quality).

      3. Bun has been rewritten in Rust using Claude

      4. The fact that the test suite for Bun passes now that the code has been rewritten does not guarantee a better quality, because the coverage of such test suite was not enough in the first place to detect the quality issues when the code was manually written in Zig. So how can it be sufficient now to certify the code quality of the Rust rewrite?

    • nicce7 hours ago
      > To say that a human is any better isn’t sound judgement or logical.

      For the old code base, there were humans that understood the code. There was a thought process for many things before they were added. Now, all this is lost and nobody knows the current code. And they can’t invest time to understand it, as Claude might rewrite things again. Now, only way to contribute reasonably is to pay tokens.