@mjbommar That sucks, but my employer has datacenters all over the world, 2^16 SHA2 operations is literally too small to even measure? For people on limited devices however... 😬

@sjvn I understand the intention, the article is explaining why I think the solution has the problem inverted?

@eruizc_dev I'm only aware of some anecdotal reports, like this one for example. social.anoxinon.de/@Codeberg/1150…

@taviso Anti-clanker clankers just like @uwu_underground

@taviso You must submit proof that you own a pair of cat ears to prove that you’re a kernel developer

@taviso no-anime costs $50/mo

@taviso It's not a catgirl. It's a doggirl.

en.wikipedia.org/wiki/Anubis

@taviso Keep up the amazing work, Tavis!

@taviso I recently set up elixir to index some kernels that aren’t online. It’s still indexing them a couple days later. So I can kinda understand why an anti-bot layer is needed to keep resource use practical. I don’t want to sound like an asshole but you could always run it yourself?

@taviso Read the docs

@taviso tldr @OpenAI are ruining the internet for everyone

@taviso Protip: Change your useragent to curl

Anubis is literally the worst thing that's happened to the interent in ages

@taviso also, fun fact, you can get rid of the anime catgirl by just using non-browser user agents. would be a real shame if someone built a browser extension that does that automatically when detecting anubis 🤭 addons.mozilla.org/en-US/firefox/…

@taviso I think a few big/notable open-source projects use Anubis because ... it seems to work? I think the interesting question is why - it's just enough of a metaphorical "locked door" such that AI crawlers would look *too sleazy* if they just "kicked it down"? For now?

@taviso Huawei bots can already bypass this

@taviso "There's a solid strategy behind it: give people what they want with a flaw that makes it unviable for corpos. Then sell the cure."
news.ycombinator.com/item?id=449288…

@taviso Why don't you ask the guys that sign your checks? If you're going to scrape and steal everything easily accessed online, nothings going to be easy to access online

I found a vulnerability for you. What happens when you run out of data and you start recursively training?

@taviso I’d argue it’s “because we can’t have nice things” and if you allow bots to scrape your site they will take down your infrastructure and hosting costs will go up. But still weird that it’s an anime girl, you couldn’t find a different mirror?

@taviso Cool blog post, just finished reading thru it. I liked the links to relevant info like the register and wiki pages.

@taviso About the article:
could you redo your calculations for the cost(s), by *assuming* the AI bots are (1) independent, (2) don't have a cookie file, and tries to access ~300k pages + ~3*300k images (a site I sysadmin for)

@taviso Perhaps Anubis works because the most aggressive bots are not browsers, they just pretend to be one, and do not execute JS. Browser-based bots presumably pass Anubis, but they don't hammer servers as much, so it's a win (for now)

@taviso why does half of software online have anime girls in there 😭

i overall dont like any sort of companions in apps/websites because they look cringe but this?

this is just my personal preference though of not wanting to see this

@taviso You seam to have found yourself in a hairy situation.

Why is it as if it was written on the wall in 1999?

The cats changed slightly, but: cats!

@taviso That's Anubis, and it's a jackal, not a cat.

@taviso i think this is cat and mouse behavior. it works because it’s not famous enough, once it block enough sites all crawlers will bypass it with minimal effort, as proven by your calculations.

@taviso that's not anime

@taviso it says it right there at the bottom