zlacker

>>gregor+(OP)
> I thoughts LLMs were glorified Markov chain generators that didn't actually understand code and couldn't produce anything novel.

so he's been convinced by it shitting out yet another javascript oauth library?

this experiment proves nothing re: novelty

>>blibbl+wg1
While implementing the OAuth standard itself is not novel, many of the specific design details in this implementation are. I gave it a rather unusual API spec, an unusual storage schema, and an unusual end-to-end encryption scheme. It was totally able to understand these requests, even reasoning about the motivation behind them, and implement what I wanted. That's what convinced me.

BTW, the vast majority of JS OAuth libraries are implementing the client side of OAuth. Provider-side implementations are relatively rare, as historically it's mostly only big-name services that ever get to the point of being a OAuth providers, and they tend to build it all in-house and not release code.

>>kenton+gj1
I think you're easily convinced.