For the construction of my AI youtube advisor project, I was in the process of building the brain of the project. This brain would give the AI the conceptual information about how to behave on youtube to be successful.
In this context, I wanted to find the truth about how youtube works, what matters most, how to be successful. So I researched this a bit, on youtube mainly, and I found some good videos that were explaining how youtube functions. After some thinking, I realized that what I see wasn't shown to me because it was reflecting the truth, no the videos that I see are the ones that the algorithm shows me because the videos are well made and attract a lot of views. This means that they probably did something good in respect to the algorithm of youtube, but it doesn't mean at all that they embody the truth. What you see on youtube is 'bro-science' not real science.
Find the source code
I decided for the purpose of my project to go further and not listen to second sources on youtube. I wanted to really find how youtube works. I asked exactly this question to an LLM and the answer that I got was that I should look at the white papers and documents of the engineers that actually build and maintain youtube! This was an eye opener. However, I didn't want to go through these dry documents myself and collect the data. So I asked AI to create a prompt for a deep research on the topic and how it would present the data.
It generated a document that is very powerful, not for me, but to feed to an AI to actually understand youtube and it's algorithm. When I read the document, I hardly understand it, but AI does and on this basis, it has a real understanding of the source code, the algorithm of youtube.
In this video, I show you how I proceeded and which prompts I used to create this document.
With
, I'm running the 
project.
Please consider supporting our Witness nodes: