Using LLM to fix real software bugs

An experiment asks real developers to fix real bugs in real software, each day randomly asked to use a bullshit generator (LLM) or not to use it. They generally expected that their work would go faster using the LLM, but instead it slowed them down.

Side issue:

I am disappointed that this article promotes the amoral "open source" philosophy instead of the moral philosophy of the free software movement. But that doesn't invalidate the experiment itself.

</li>