Fetching from the wire…
Policy2026-07-27 · source-backed
Reporting resurfacing on HN (739 points, 468 comments) and mainstream outlets documents bulk purchases of rare and out-of-print books, destructive scanning, then pulping. ISBNdb reportedly brokers anonymous orders from 1,000 up to a million titles, with buyers prioritizing pre-2022 publication dates specifically to avoid AI-generated text contaminating training corpora. The legal cover is Judge Alsup's June 2025 ruling in Bartz v. Anthropic, holding that digitizing legally purchased print books for LLM training is fair use; Anthropic's internal "Project Panama" document described the goal as "destructively scan all the books in the world." Booksellers note the asymmetry: the last surviving copies of an 18th-century text aren't replaceable once shredded.
Each link below shares sources, entities, or timing with this story.
The claim is bulk acquisition via intermediaries, destructive scanning, and pulping of originals to secure text predating the 2022 generative flood. Anthropic, Meta (81 terabytes of torrented ebooks) and Amazon are named. Anna's Archive 359 points, 266 comments, anchored on Ju...
Participants with AI assistance got 9% of answers right. The baseline group, no assistance, got 27%. Three times worse. And their self-reported confidence went from 30% to 76%. The research comes from University of Milano-Bicocca, École Normale Supérieure, and Sapienza, and th...
69% of all input tokens in production LLM traces are system prompts. Let that sink in for a second. Datadog's State of AI Engineering 2026 dropped yesterday, and it's the best empirical snapshot we have of how companies actually use LLMs. Not how they demo them. Not how they p...
The decision, covered by Simon Willison, frames the ban as strategic: maintainers invest time reviewing contributions to mentor developers into trusted long-term contributors. If an LLM wrote the code, that mentorship is wasted. The wrinkle: Bun (acquired by Anthropic) runs a...
The open-source coding agent space just got its first credible frontrunner. OpenCode, built by Anomaly, launched this week and immediately became the top technical story on Hacker News with 802 points and 359 comments — the kind of signal velocity that separates real developer...
Detailed analysis debunking viral claims about Anthropic losing $5,000 per subscriber. Breaks down token costs, caching strategies, and batch pricing. Deep HN discussion on LLM inference economics and whether Claude Code's pricing is sustainable. martinalderson.com | HN
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.