monsieurbanana

To help classify between human and llm-generated ones? I don't know, and you're probably right than in this case a fixed message would be better.

Another situation where I would use llms to generate messages is if we've mutually agreed to do that. For example automated review messages at work, if we've agreed that it would be a part of our workflow.

A good rule of thumb might be unsolicited vs solicited llm messages.

It's disingenuous to compare LLM outputs to script generated ones. They have more differences than similarities.

A script is reproducible, and fixable. They won't waste your time hallucinating issues. Note that this is different from "being wrong". A script can be wrong, and you can spend time debugging it, but that time is not wasted. You can fix the script, remove wrong rules, whatever.

Also, nothing exists in a vacuum. Script-generated review comments were never a widespread societal issue, LLMs unfeterred content generation is.

I guess I have to temper my own statement a bit. I do see situations where I would generate and send automated LLM messages. Probably to reply to generated unsolicited messages or pull-requests.

But to upload to debian new's queue you already need to have pre-approval, it's a different situation from e.g. a public git repository that anyone can push to.

Does debian's NEW queue actually have this issue? Not anyone can just upload to it, you have to be a debian developer or maintainer AFAIK.

I'm not anti-ai but I am firmly anti-make-people-read-llm-output-I-didn't-read-myself, and this breaks this.

They also have a reward system that punishes people for ignoring the llm review (by pushing their submission lower in the queue).

> Whether it's AI or not is irrelevant at that point if that's your stance

The human brain works with pattern recognition. Regardless of whether I think that llms are great or not, right now they're vastly used to produce slop (I know it doesn't have to be the case), and because of the aforementioned pattern recognition we are largely being trained to recognize ai-generated content as slop.

There's really only one logical outcome:

If you're trying to sell something to me and you have a vibe-coded website, it doesn't matter how great your product is, I'll never even get to see it because it's slop-by-default.

And you don't even have to bring up AI at all to see how that's a bad idea. AI is flooding everything, isn't it a basic marketing strategy to differentiate yourself from the rest?

> I think the agentic industry is moving away from k8s as a deployment platform

Interesting, I don't to keep up that much with the AI changes so I'll just ask, why and where to?

> If we want to improve on this, it has to come with the current level of ease of use at a minimum.

OR it has to be a situation catastrophic enough. I don't think this one is, but it's hard to be optimistic for the future

(this is not from Terrence Tao, he just reposted it in his blog)

I never understood this, is there anything that makes it difficult for the original uploader, the one that supposedly offers the file directly, to offer a torrent instead for the same amount of time?

As far as perennity is concerned it seems strictly better.

We really should start being more precise when we say "replace". Are we talking about replacing 90-100% of that particular job? Or 50-80%?

Because it feels like people argue as if they were talking about the former, but if probed it turns out they admit it's the latter.

There will be still people making car advertisements, but they will be using ai tools and will get asked to produce X times faster. Effectively, some of them will get replaced, or rather have been replaced.

I like this comment because at least it's honest in the timelines for AGI

You could have asked this question 10 years ago, long before llms. It's not like you'd realistically would get an ocaml job back then when there's so few of them, so why bother?

The answer is still the same as well, people learn ocaml either because they enjoy it, or because learning a functional language makes them a better programmer overall and teaches your brain to approach a problem in a different way.

Let's say you follow a link to a github project and you have to assess whether to trust that code, how do you do it?

Before AI just having a good README and superficially looking good code (we're not going to do a deep review of it) was a gauge of quality that at the very least the person that wrote it knows what it's doing.

How do you do it now? Maybe the AI was expertly guided, but maybe it was vibe-coded by someone who hasn't even heard of basic security principles and your whole data is sent in plain-text somewhere. It doesn't have to be nefarious, it could be a debug feature that was never removed, or maybe the user never asked the llm to "make it prod safe". And there's no way to know unless you read the whole thing, because superficially the README and the code will look as good as any other llm project.

That's problem number one, before you could quickly see if it's made by someone who knows what it's doing, and thus that you can trust. With vibe-coded software not so much.

The second issue is not inherent to vibe-coded software but it's caused by it: it's a sea of garbage. It's very possible that the absolute amount of good software has increased, but it's become harder to find it.

I understood your comment the same way as the guy you're replying to. Maybe it'll help if you expand on whether we can trust vibe coded software? That's really the crux of the issue, and if you're not addressing that your comment is just a truism