AI for AD(H)D - Groot-9

I have a focus problem. My friends are joking about that. I have to admit. My GitHub is 425 repos that are minor patches. That’s addiction. I can’t focus on what is important.

Yesterday I got fixed on creating a super minimal Groot-1 model from scratch. When I woke up, it is usually the time when I am extremely excited to do things. Now is the time. But I don’t know a lot. And my focus will be gone in the next hour.

So how to teach person who has no focus? Skip boring parts, do not even mention them. Give a piece of knowledge that short-circuits what the person knows and the next thing it needs.

How to find out what person knows? Create a map of all human knowledge as micro/nano lessons. Then observe person posts, talks, commits to color that map with “yes”, “no” and “unknown” paints.

I’ve heard that the math tool called Bayes statistics or Bayes inference can help with calculating what a person needs given his goal and current questions.


DLAI course need to be more interactive and more personal for this to work. But it also should give users control over the data that they share with 3rd party services. By control I mean ability to download and use the data about themselves to help them.

I am really excited to put this theory into practice, but it saddens me to death knowing that it is impossible. The superposition between AD(H)D and job requirements to be focused on specific tasks and projects does not exist.

The only thing I want right now is to make a minimal neural net, manually, with complete understanding up to a single hex in a hexdump, that outputs nothing but “I’m Groot”, studying different layers and their effects, how input can influence output (or don’t), how their math works (to some degree), and I am tired going through a tons of verbose and probably misleading LLM texts.