I posted to HN, and then my website was flooded with AI BOT traffic

Entries describe a range of engineering experiments and reflections. One recounts posting a customizable feed reader, sprinklz.io, to Hacker News and discovering the site was overwhelmed not by human users but by aggressive automated crawlers. Top-level logging showed most hits never executed the single-page-app’s JavaScript; nearly half the traffic came from a service that polls sites for changes to produce AI-generated newsletters. Other traffic included Claude, OpenAI, Google crawlers and a few vulnerability scanners probing wordpress paths. Responses included blocking the abusive service, adding robots.txt, rate limiting, and a set of infrastructure and logging improvements aimed at handling much higher QPS and reducing latency. Sprinklz is presented as a free tool for filtering topics (e.g., AI/LLMs) from feeds.

Other pieces cover practical and philosophical work. One describes using LLMs to produce reusable JavaScript transformers that load a page in Firefox, run its JS, ask an LLM to generate a simplifier, inject that code, then forward simplified HTML to Emacs’ text browser (eww); the implementation is open-source as simpleweb. A cultural critique argues that worse online experiences stem from perverse incentives and attention-extraction, recommending offline time, protest, and breaking addictive patterns. A Clojure library update, Deft 0.2.0, adds namespaced-key and required-key support so protocols and records can declare and depend on fields without boilerplate getters/setters.