July 25, 2026
Bot blocker or human blocker?
Who does Anubis actually stop?
Turns out the bot blocker may be slowing humans more than the bots
TLDR: A developer says a web blocker meant to slow AI scraping was easy for AI to bypass, while ordinary people still get stuck waiting and burning battery. Commenters split between calling it a blow against the open web and defending it as a useful speed bump against mass data harvesting.
A developer tried to feed a public Linux discussion into an AI helper and ran smack into Anubis, a website gatekeeper that makes visitors do a tiny bit of computer work before the page loads. Dramatic twist: the AI apparently worked around it anyway. The author even built a tool, anubis-fetch, to solve the challenge or just open a real browser if needed, then asked the question that set off the comments section: if the bots get through, who exactly is being stopped?
That question lit a fire. One camp basically yelled, "this is just another annoying roadblock for normal people". The loudest critics said it feels like a tax on humans: more waiting, more battery drain, worse access for phones, text-only browsers, screen readers, and old devices. One commenter went full doom-post, saying it "accelerates the death of the open web". Another piled on with disbelief that the system lets a reusable cookie do the heavy lifting, calling that the exact opposite of what they expected.
But the thread was not a total pile-on. Defenders pushed back hard, saying the author is aiming at the wrong target. In their view, Anubis was never supposed to stop one person with an AI toy; it was built to make giant scraping operations more expensive and less attractive. And then came the comedy relief: one commenter casually claimed their simple Python script still worked fine by tossing in a basic browser-like label, which is exactly the kind of reply that makes security debates on the internet feel like a sitcom.
Key Points
- •The article centers on a developer encountering Anubis proof-of-work protection while trying to have an LLM read a lore.kernel.org discussion related to a Linux kernel patch.
- •The author says an automated tool, anubis-fetch, can solve Anubis challenges natively or use Chromium as a fallback, enabling access to protected pages.
- •The article states that cached challenge cookies make the proof-of-work cost low for scrapers over time, while human users face repeated delays and battery use on fresh visits.
- •The article says non-JavaScript clients, including text browsers, screen readers, and RSS readers, are excluded by the access mechanism.
- •The article provides rough calculations estimating annual human time and energy consumed by proof-of-work challenges at different worldwide daily solve volumes.