Wikipedia tells us that _low-background steel_ is steel produced before the detonation of the first nuclear bombs. Yep, you guessed it, **this is a talk about Large Language Models**. LLM outputs have quickly spread like radionuclides, threatening everything from the scientific record to the existence of the Internet as we know it. In this talk I'll discuss _practical small web approaches_ that we can use to build a new Internet that doesn't suck quite so badly. There will also be memes ;-)

Have you noticed how the **good stuff** on the Internet is increasingly hidden behind bot checks, subscriptions and paywalls? And that it's getting harder and harder to find things online due to LLM pollution? Welcome to the club! You are in the right place.

In this talk I'll highlight some of the most egregious examples, consider how we can best preserve _low background information_ for future generations, and how we can use small web techniques like **self-hosted blogs and static site generators** to bootstrap a new infosphere that doesn't rely on a handful of _hyperscale operators_.

I'm particularly interested in how we can _federate and syndicate search_, learning from protocols and standards like RSS and ActivityPub. As part of the talk I'll give you some practical tools and approaches to try. If you find this interesting, consider joining us in the [SearchClub](https://matrix.to/#/#searchclub:matrix.org). **Let's have fun building the new Internet together!**

Licensed to the public under https://creativecommons.org/licenses/by/4.0/

about this event: https://program.why2025.org/why2025/talk/FHLCMR/