Another amazing example of "Everything is selection".* This experiment is really neat, so let's go through it. Facebook users are exposed to more like-minded than cross-cutting news sources. In yellow, you see how this looks for people who were enrolled in this study and in

Interesting summary of the drawbacks of building your own GPT-esque bot. It brushes over the fact that these things aren't all-or-nothing (you can augment GPT with your own data in a vector database) but even "just" RAG is a big data pipelining problem. medium.com/@aldendorosari…

Many are liking this because it casts their poor impulse control and hierarchy blindness as a social good rather than just a problem. Well, that's why I liked it anyway.

It's becoming a more and more common solution to inject recent data into short term memory (context window) as a way to fill in gaps in Large Language Model knowledge and paper over hallucination. Nice approach here, using Wikipedia as a database

Jason has long been my go-to source for breaking down Google and Meta legal developments, particularly when it comes to things like tracking, looks like he'll be my go-to for OpenAI and LLM copyright law developments too!

It has been SO useful to get an idea of how historic GA4 data might have looked, before it was live on site has been wildly useful. Here Dave is sharing a way to do it, free! Check it out!

Great post for SEO's looking for an introduction to LLMs through the language of how SEO's look at information gathering, retrieval, etc by Robin Lord. moz.com/blog/know-abou…

Want an easier way to handle GA4 BigQuery? We're launching Pipeline! Enterprise level GA4 data in 15 minutes, with no SQL or data engineering. Session level dimensions, multi touch attribution, Looker Studio connector. All ready to go.