@mischievoustomato @Junes @phnt @waifu @zero If you try hard enough you can bypass nearly anything, but is it worth it for what little you probably get in return? Probably not.

@mischievoustomato @Junes @phnt @waifu @zero they could, yes. They could also make their user agent something absurd, but they don’t. Why? Because it would look suspicious as fuck if they didn’t. Does that matter to some random git repository or open source wiki? Fuck no. Even the local LLM I’ve been messing with says that:

AI scrapers use normal browser user agents to:

• Evade anti-bot detection.
• Render dynamic content properly.
• Ensure compatibility with APIs and servers.
• Mimic real user behavior and reduce the risk of blocking.

However, this is just one part of a broader strategy (e.g., proxy rotation, request throttling, and behavioral mimicking) to avoid detection.

@sendpaws @Junes @phnt @waifu @zero @mischievoustomato you do realize that what you just sent and anubis are the same exact thing, right? They are both Proof-of-Work proxies. One of which just says 'based' instead of having a stupid anime girl on it.

Hard thinker, this AI model
Thanks Google

@mischievoustomato @Junes @phnt @waifu @zero The problem with that is the point of the crawlers using the regular UA is so they blend in, if they spell something wrong, they will stand out as an obvious bad actor.
That's why User Agent spoofers exist as anti-fingerprinting measures, it's more "private" to look like the most common browser.

the "normal" slurpers and other bad actors don't do that, a malicious one might

@waifu @Junes I do for git.froth.zone, yes. I kept running into downtime from AI scrapers slurping my shitty code and I saw it as an alternative not long after it was first unveiled. I could try disabling it to see if I keep getting slurped by AI trainers.
Malicious scrapers could bypass it by altering their user agent but that would only make them stand out even more and make it easier to just block them (eg. they spell Mozilla as Mozzarella would bypass Anubis by default but be blatantly obvious). Also none of the regular AI slurpers used by eg. OpenAI, Anthropic, Amazon, Alibaba, Google and Microsoft do that, they usually just use a standard browser user agent to hide mask their activity.

I put an annoyance over repeated downtime

@waifu It's me I got banned from telegram like 3 years ago and I wanted an account for some reason

@icst When it comes to super high performance stuff, not at all, afaik. Even when I was in uni pre-chatgpt most ML libraries really only used Python before it was popular and easy to use. Before Python, ML (especially the class I took on AI) was in Lisp.

Btw what I'm using rn is an open source Go program https://github.com/ollama/ollama making my 5070 Ti cry. It runs deepseek entirely on the GPU but the smaller openai model makes my CPU cry too

my CPU is also dying

making my GPU cry by asking about woodchucks

@SuperDicq Good point. Here goes nothing

@SuperDicq You're right, time to speedrun my dox

Running a local LLM exposed to the fediverse would probably be funny but you freaks would annihilate it with horrendous shit

@zonk I just got auto-rejected from a place in an hour, at least that is believable

@zonk wdym, hiring managers exclusively work on rejecting people at 03:00 on a Sunday, that's when everyone does it. At least that's what it seems like, when I get an email at 04:02 for a job I applied to the day before I definitely did not get auto-rejected from.

@sam i'm still getting rejections from internships i applied to in 2023 lmfao

@zonk They looked at your CV long and hard, please understand.

Rejection from a job I applied in April just landed in my inbox, 10/10 no notes

GNU/Windows operative system

@sneeden https://www.msys2.org/