If you haven't already, check/increase the GPU memory carve-out on your UEFI.

More details: https://rocm.docs.amd.com/en/docs-7.2.0/how-to/system-optimi...

that link actually recommends not doing it from UEFI and doing it via software

The code casino is real and I don't have an answer to it, only share the sentiment. It is frustrating and exhausting to go through those boom/bust cycles.

The workflow that works for me is: I do the thinking, the writing, then let the AI review it. I am still doing all of the creative work and thinking. The AI keeps me honest and its code reviews are very helpful. The loop becomes competitive and makes me pay more attention to detail; I try to nail things on the first attempt. This gives me a sense of accomplishment and improvement.

Whether this workflow has its days counted, I don't know. But I also don't care. If/when software development as we know it truly stops being a thing, then I'll sadly just have to move on and work in another field. I am not doing vibe coding or programming in natural language. Fuck that.

But I have also noticed that, as much as the AI can process things faster than I can (I also use it to help me navigate the existing code base, and that is also very helpful), it cannot answer the question of whether we should do something, or whether X is more desirable than Y, etc. Maybe for some people, thinking about those things and letting the AI do the rest is enough. For me, personally, I like to navigate all levels of the development.

"AI told me" -- yes, absolutely insufferable dudes. I just don't even bother stating a reply.

Because the discovery is how, not whether.

Curious, but how do the bots figure out the combinations? Or do you have links to the diffs from other sites? I assume the diff takes two files in query parameters or something.

I'm not 100% sure but I think links. There's a bunch on the history and revision pages. Yeah, the diff URL has two revision ID's as parameters.

I did try removing some of the links without success. I guess once they have them they just keep checking.

Why not block all odd v4 addresses while you're at it? I heard that that can reduce scraping volume by 50%!

That's harder to set up, and also unfair to people who have an odd IP address.

It's easier and better to just block 0.0.0.0/1 half of the time, and 128.0.0.0/1 for the other half of the time. Switch every day at noon.

Bot traffic will be cut by 50%, and humans are all treated equally! It's a total win!

And blocking ipv6 addresses isn't unfair to people who have an ipv6 address?

Yeah, I suppose you're right.

Just block it all.

Blocking Singapore reduces the AI load 90%.

I second this. My website exposes a cgit and 99% of the traffic now is AI scraping the sources, but the load is nowhere near DoS territory. And this is running on the cheapest VPS I could find.

Not saying I'm not annoyed by the scraping; I am looking to block them, but I'm also not going to put the site behind the gatekeeper. If anything, Cloudflare must love AI scraping now for the same reason AV companies love malware.

Now, if you are running a PHP stack...yeah, maybe that's the problem right there.

Is there actually any plausible theory why "AI" would repeatedly scrape the same sites? Are there that many competing, completely independent AI labs? Is it cheaper to repeatedly scrape than to buffer the scraped data locally? (I find it very hard to imagine that it's easier to deal with changing/disappearing content than it is to stand up such a cache.)

If you ask an agent to check sources / function definitions of open source packages it will wget / curl it

It's an AI generated scraper that scrapes nonstop.

> 99% of the traffic now is AI scraping the sources

I wonder if we should stop fighting this and instead create an API specifically for this purpose? Or, a central repository that you could send your data to and say to anyone wanting to scrape, "safe yourself some time and just get my data from this other place"

The thing though is that they are extremely idiotic. They are constantly, recurringly, scanning the same files, I suppose out of FOMO that a line might have changed. I don't know what a special API solves, especially because HTTP already has etags to save you from re-downloading the whole damn file over again. But these bots don't care. The extent to which they don't care is such that, after I temporarily took cgit down for kicks, they'd get 404s and still repeatedly ask for the sames files days on end.

The PHP stack isn't even the problem, it's having unauthenticated requests getting past the cache in the first place, something that most sites should be able to prevent.

Looks like arsTechnica mastered the clickbait. The title is misleading, the first sentence contradicts it, the claim is false (there is stock), and even if it weren't it's irrelevant without knowing inventory size. I can think of 0 positive things going for this article.

No, they weren't wrong -- it was listed as sold out almost immediately after. It's just out of date now already since it looks like it has been restocked again. Source? I had checked shortly after they said they restocked only to see it was sold out again.

Also, they addressed the inventory size in the fine article. Maybe the snark isn't warranted?

I used a tracker to be notified of when it become in stock. It went in and out of stock several times that day.

I agree, the title and article say exactly what happened

Why? You are still giving up privacy. There is 0 reason to be using Meta products, let alone pay for them.

Seems absolutely unhinged. I don't know who'd pay to doomscroll AI-generated slop and fake news. $49.99 for the top plan, lol.

But they know who.

All I can think of is people who need those accounts for professional reasons (i.e. public relations)

I had cynically done this sort of tokenmaxxing for a while as a burnt offering to the token-hungry non-leadership.

Eventually I got tired of it and got back to work.