[duozerk]

So you did use LLMs to write at least part of the software. I imagine you feel no shame, but it would be nice to at least mention it on the github page. It's a security risk.

As for your question, I don't know about the person you're replying to, but for me any software where part of the source was provided by a LLM is a no-go.

They're credible text generators, without any understanding of, well, anything really. Using them to generate source code, and then using it, is sheer insanity.

One might suggest it means I soon won't be able to use any software; fortunately the entire fever dream that is the ongoing "AI" bubble will soon stop, so I'm hoping that won't be the case.

[satvikpendem]

They literally state that they used LLMs to build it in the second sentence of their initial comment so not sure why you frame it as something they weren't upfront about.

As for it being a bubble that will stop completely, that ship has long since sailed and I assume you're inadvertently using LLM generated code somewhere in your software stack already, due to news reports saying certain companies are already using LLMs in their codebase.

[yoavm]

I wish I could speed up time just to see how this comment would age. While I personally prefer living in a world without LLMs, I do suspect you're going to end up without any software.

[duozerk]

A more reasonable response than my admittedly slightly aggressive comment deserved.

Indeed, we'll see.

[dylan604]

I'm imagining some apocalyptic world Mad Max style where there are underground groups hand writing code to avoid the detection of the AI. Unfortunately, so few people are able to do it any more and the code is so bug ridden that their attempts at regaining control over the AI often ends in embarrassing results. Those left in the fight often find themselves wondering why everyone just rolled over for the machines, what, because it made their lives easier??

Maybe it's a scene from a show I've seen already??

[bigfishrunning]

I suspect we'll all end up without any software, once we've successfully gotten rid of anyone who can evaluate the output of an LLM