Netflix thinks it knows what movie I’ll want to watch next. Amazon thinks it can guess when I’ll need a refill on dishwasher soap. Wayfair is convinced I need new blue bar stools. 

Most of these predictions miss the mark. Nevertheless AI is no doubt a fascinating and powerful technology. And if I had to put my money on predictive power, I’d have to put it on Facebook. Instagram’s ability to translate what I say, browse, and basically think into mostly relevant sponsored posts is honestly pretty scary to me. And yet, I’m impressed as well. 

But what if I took back control here? … and built my own database of preferences, and trained an algorithm on what I say I actually like…

What if I could unleash this personalized algorithm on say nytimes.com’s home page and ask it to tell me the most timely, relevant articles to read? Too many news aggregators have tried unsuccessfully to tell me what I’ll be interested in. They ask me maybe 5 to 8 preference questions and then bucket me in some audience cluster. It’s pseudo-personalization at best. 

I’ve been dreaming up some variation of this experiment for ten years now. The idea came to me as ‘The Little Black Book’ when I was working as an eighteen year-old at a summer camp, and I think at almost 28 it’s high time I put this idea into the world. 

Now it’s 2025: we have the power with AI chatbots. If you feed your chat bot enough of your preferred information sources, it can be trained to search your personal database and respond off of it, rather than replying from a more wide-reaching scrub of the internet.

Curious to experiment on your own? this article provides a step-by-step guide to “training Claude with your own Data(RAG chatbot).”

Leave a comment