> For me personally, I’d trust Yegge’s Beads about as much as I’d trust most other OSS projects, not because the LLM is good, but because of those tens of thousands of eyeballs. I’d trust something an LLM produced just for me significantly less, because I’m the first person to run the thing.
I thought this was an interesting and valid point!
I'm old enough to remember when this wasn't the case. Compilers back in the 80s and even early 90s had bugs. Heck, even Intel processors had bugs!
The bigger problem though with treating LLMs as a compiler is they're non-deterministic. The same prompt can generate two different outputs at two different times. One may be correct, the other may have subtle bugs. In the end, we should be focusing on what we should have been focusing on all along: what evidence do I have that this code is correct?
> For me personally, I’d trust Yegge’s Beads about as much as I’d trust most other OSS projects, not because the LLM is good, but because of those tens of thousands of eyeballs. I’d trust something an LLM produced just for me significantly less, because I’m the first person to run the thing.
I thought this was an interesting and valid point!
> I trust the output of my compiler
I'm old enough to remember when this wasn't the case. Compilers back in the 80s and even early 90s had bugs. Heck, even Intel processors had bugs!
The bigger problem though with treating LLMs as a compiler is they're non-deterministic. The same prompt can generate two different outputs at two different times. One may be correct, the other may have subtle bugs. In the end, we should be focusing on what we should have been focusing on all along: what evidence do I have that this code is correct?
> These usages of LLMs require us to trust the output of the LLM without having a human read and understand it.
Would be nice if it was true. Whereas in fact anyone can put out parrot-generated code without trusting anything except that it saved time and effort.