So we can be suspicious that there is some truth, one way or another that they could have reused prompt/data generated by the user session.

But that was before, now they pack more bloatware than samsung devices.

Then, one day you are in a hurry to do something, like buy a ticket for a train that leaves in 3 minutes, or find the number of the train that you are supposed to take in the e-ticket.

So, you open the app and you encounter the following error "A new version is available, please update to continue using the app". No way to discard that and do whatever you have to do in a hurry. So you think, let's update that fuckin app. Ok, internet is not working well, let's connect to the train station public wifi. Oh they have a broken captive portal that is not loading or redireting you to the login page... ok let's find the captive portal connection ip/url.

All good, internet is working, now we go to the app store, starts the update, wait a minute for download, and...

bipbipbip not enough remaining storage space to install the update. You have to understand them, they needed to pack inside the app, all the PDF with picture of their ecofriendly policy of planting trees each time you swear while using the app, and sadly that increase the app by a few mega bytes...

So then, at this point you missed your train and you hate more than anything else in the world, the *** product manager at company XXX that decided to 1) force using an app, 2) decided to force an update that you can't skip.

React for example.

And we could easily guess that there wouldn't have been so much open source models, and grand public experiments and free tools if llama models were not release to the general public.

In that case you can be satisfied of yourself while your user base and promoters is plummeting.

It is a great thing that you still have a large user base that will still complain of issues they have so you can improve or ever on that if it is your wish.

And nothing as strong in the end that the fidelity of an user that was complaining about an issue and now feel that his voice was heard!

It really change the url the moment you click on it, but it is so fast and subtle that you see nothing. But you can see it more clearly by doing a right click, like when you want to "copy the link". Just doing right click, you will see in the status bar the link url changed and it will stay to the google redirect server as target. That sucks hard!

https://www.euronews.com/business/2026/05/01/contentious-eu-...

https://www.euronews.com/my-europe/2025/05/07/metsola-threat...

https://www.consilium.europa.eu/en/press/press-releases/2026...

> But most of the routes scrapers are hitting are e.g. computing diffs between arbitrary pairs of commits, or other computed-on-the-fly views into history.

Looks like to be wrong based on the article: Today, git.kernel.org receives about 6M daily requests demanding to see random commits.

Also, in "how bad is it", there is a bar graph of the tasks requested. And only diff can't be cached, and it is clearly not the most important one.

> batched waves from laundered residentail proxy IP addresses

Again, I don't see anywhere anything stating that there was "grouped" requests that are directly related, except in something not related to AI: The thing that usually takes us down are not scraper bots, but poorly designed CI systems that try to do something stupid like shallow-clone stable.git from 20 different nodes, all at the same time. (Shallow clones are awful. Run your own damn mirror if you're going to do something nasty like that.)

Just "waves" of requests coming from the same "residential proxy IP". But that makes sense because they might come from a same service without being a single "source" trying to scrap for training.

Let's say that I'm chatgpt, or Claude, and I have a lot of user related requests to do things based on the commits there, each one will be a different task, but they will all go through the same pipe, that would be a proxy with the residential IP if the provider notice major websites like this one blocking requests for bots.

For a shopping website or user generated content website, I might understand the terrible load of crawlers that are trying to "steal" the data.

But for the kernel, what's the purpose? Are you that "no human" are seeing your page or its content? Maybe we should investigate more the usage being this "bots".

I don't buy the explanation that there are millions LLM that are constantly trained on redownloaded data from kernel.org. What would be my better guess is that it is not training, but users are actually accessing this content through chatbot and co. Like when you ask why your sound is suddenly not working anymore after an update or why your wifi driver is constantly disconnected after leaving sleep, it might be possible that the "LLM agent" is requesting the commit contents to "understand" or refer or explain them. Is it a bad thing if it helps users?

But actually, regarding this article, I'm quite amazed that with all the advances of the linux kernel, and server softwares, and that the C10k challenge is solved since a long time, still such a basic traffic is such an issue.

> At any one time, across 5 geo-distributed nodes, there are 14 CPU cores doing nothing but rendering git commits as html. 14 cpu looks nothing to me. It's like you have 1 iphone and 1 raspberry pi active in a corner of a room. Counting in "seconds" of activities, easily shows meaningless huge numbers. Do you want to know how many breaths I take per year? 8 to 9 millions!

Most certainly, the usage of this shitty Anubis has ruined the climate million times more only with the wasted cpu resources of legit users...

But moreover, by definition the git commits are not supposed to change, ever, so can someone explain to me why the fuck do kernel.org "re-render" the commit to html each time someone is accessing it instead of using a cache or a static version of the html of this commit?

> oh, several BILLION valid URLs you can scrape, only to get 922 duplicates of the same 1.48 million commits Again, reading that, my immediate thinking is that it is a shame that such talented people would not be able to have a proper optimization, so that getting the 922 duplicates are just costing a fraction millisecond more after the first person retrieve the first page.

Looking into ideas, I was a little bit disappointed to not find community website of people sharing ideas, experiences and results around such hacking.

You can't mix non async code with async. And you can easily do in some way and block your application without you noticing.

And in the end, async is not even really async, just cooperative execution.

I'm a big fan of Python since very low version. And I don't have too much difficulty using async/asyncio now I'm used to it, but in all honesty I don't think that it is really great in the grand scheme of things.