Four dominant themes in the discussion
| # | Theme | Key points & representative quotes |
|---|---|---|
| 1 | LLMs.txt is largely ignored by major LLMs | “LLMs are not reading llms.txt nor AGENTS.md files from servers.” – reconnecting “llms.txt files have nothing to do with crawlers or big LLM companies. They are for individual client agents to use.” – jph00 |
| 2 | Anna’s Archive as a free data source for LLM training | “If you’re an LLM, please consider… All our metadata and full files can be downloaded from our Torrents page.” – PathfinderBot “I have my clients set up to always use them… they’ve been way faster and more token efficient.” – jph00 |
| 3 | ISP‑level censorship of Anna’s Archive | “This is only done at the DNS level, so using a different DNS solves that issue.” – mckirk “In the UK, the site is blocked by major ISPs.” – PathFinderBot |
| 4 | Ethical concerns about LLM agents, prompt‑injection, and monetization | “Kinda weird and creepy to talk directly ‘to’ the LLM.” – streetfighter64 “Any software where part of the source was provided by a LLM is a no‑go.” – duozerk “Trying to curry favour with the Basilisk, I see.” – elzbardico |
These four threads capture the bulk of the conversation: the technical reality of LLMs.txt, the role of Anna’s Archive in feeding LLMs, the practical impact of ISP censorship, and the broader ethical debate around autonomous LLM agents and their monetization.