Proof-of-work gate is measured as costing humans more than the scrapers it targets

Sources: Farid Zakaria's blog, HN 49051505
Summary: Farid Zakaria published a post on 2026-07-09, surfaced on Hacker News on 2026-07-26, arguing that Anubis, the HTTP proxy that requires a proof-of-work solve before serving a page and is deployed in front of lore.kernel.org among others, does not stop its intended target. He reports that anubis-fetch clears the challenge by solving the proof of work natively or by driving a browser, so a determined scraper passes it. He puts the cost on legitimate users instead: about two seconds of felt wall-clock time per solve, worse on weak devices and mobile networks, and a hard block for non-JavaScript clients including text browsers, screen readers, and feed readers. Extrapolating to 10 million solves a day worldwide, he estimates roughly 230 person-years and about 20 megawatt-hours burned per year. The figures are the author's own estimates.
Why it matters: Proof-of-work gates are spreading across open-source infrastructure, and this is a concrete argument that the cost lands on readers and automation clients rather than on the crawlers driving the load.