Discussion
When you "break" the anti-AI software I have on my website that does not illustrate that the software doesn't work well enough: It shows that you are someone who explicitly circumvents people's stated values, preferences and borders. You are an abuser.
It's a test about how you treat other people, whether you respect their dignity. And you failed.
@tante What kind of software are you running/what does it do?
(Genuinely interested, I am not a coder.)
@monoxyd I run iocaine https://iocaine.madhouse-project.org/ .
It's quite good at detecting bots and serves them efficiently generated texts based on Marx's "Das Kapital" and historic Luddite writings. It tries to trap those bots in tarpits made of "nonsense" text to
a) keep them away from actual content (therefore saving resources)
b) keeps them busy so at least that one bot is not harming some other website while I keep it busy
c) poison their data set
It's not always easy to determine, just how effective these processes are but it does keep those bots away from hammering very expensive endpoints (with database queries) on my server.
(if you search for "tante.cc" on bing.com for example you will see texts that do not correspond to the actual website text because the bing crawler was poisoned)
@monoxyd in the last 2 months I have generated about 257 GB of content through iocaine that crawlers sucked in instead of my website. That is massively more than actual readers produced as traffic.
This is also a message for people who say "no LLM generated code policies are useless because how will you detect that my code is LLM generated"
Access and consent are two different things. Ethical people understand the gap.
@tante
In the end most pages will probably have to go the troy hunt way and have people pay at least a symbolic dollar for an account and track abuse based on that...
@tante I wonder if that is even legal in Germany… They are circumventing a protection mechanism!
@tante Is your "anti-AI software" some half-asssed coinminer? Because if so you are an abuser and should think carefully before throwing stones.
@tante out of curiosity, what notable examples of this are there (aside from the thing when a bluesky user submitted an llm-generated patch to open-slopware)
@hexaheximal @davidgerard just linked a tool some Kernel dev generated to beat "Anubis", which is one of those kinds of tools
@tante @hexaheximal @davidgerard
what? how is anubis beatable? dont they have to use at least a headless browser? so it is heavy for scrapers because of the challenge?
@utf_7 @tante @davidgerard from a technical perspective, it is actually trivially easy to defeat anubis. it works by calculating a hash proof-of-work, which a standalone program can do too. and, of course, once the challenge is completed it stores a challenge proof as a cookie. the result is an anti-bot system that is really just a javascript checker with fancy marketing
Install bonfire.cafe
Get the full app experience
