For the best reading experience with links to related notes in my second brain, I highly suggest you read the full edition in my uploaded second brain here.

Welcome to Weekly Wonderings, a newsletter exploring the journey towards Digital Self-Actualization: realizing our meaningful potential through the synthesis of digital tools, life management systems, and humanistic philosophies.
​

Here's a quick galactic overview of this edition 🌌:

• Leverage your second brain’s unique knowledge with AI 🤖
• Celebrating my first stream 🥳
• I’m Attending BASB Summit 🧠🏔️
• The science behind self-actualization ✨
• The secret power of being a generalist 🌐
  ​

Leverage your second brain's unique knowledge with AI 🤖

Reminded by the idea on how [[AI will not make PKM obsolete]], I believe it is going to be indispensable for the future of knowledge work.

There’s an infinite amount of information online, so a second brain helps you intentionally select only the best and most relevant ideas to use and refer to in your life.

When mastering a skill or area, it gives you a place to externalize nuanced knowledge to nurture your unique preferences that you’ve been slowly collecting from the corners of the internet and in conversations.

The problem with large language models like [[ChatGPT]] is that they require lots of prompt engineering and context to give more than just general responses.

But with existing AI models like [[Claude]] which have insane limits of 150,000 word context windows (that’s 3 copies worth of The Great Gatsby!) where you can upload any form of files (markdown files, PDFs, transcripts), then the possibility of personalized agents is soon approaching.

Especially with GPT-o1’s release highlighting the rational potential of these models, I’m eager to see how we can create specific agents for well-defined tasks.

Instead of having to do all manual intellectual labor, how much can we outsource it to AI as virtual assistants?

This is what I’m currently exploring with [[Claude]] and their projects feature.

Fortunately, it’s already possible.

I was surprised to see how someone I saw in a local Vancouver event went viral as they used it to create a fusor in their bedroom.

Using Claude Projects, he turned the specific data he was collecting into a highly contextualized database he could pull information from.

Unsure of what parts to use? Just ask and it will naturally query the conversations you had.

Unsure of the structural layout? Just ask for it to create a diagram, and it will quickly code up a detailed one in React based on the

Want highly digestible and contextualized expert information? Just copy paste the whole page of documentation (or entire forum discussions about the topic) and use the model’s conversational abilities to break it down into concepts you already know by meeting you where you’re at.

If you’re interested he shared a bit of behind-the-scenes in this podcast interview, to which I’ve copied his prompt in [[AI prompt to break down a goal]])

For my personal implementation, my first thought came when thinking about ideation for YouTube titles. I can feed it existing that performed well, my personal notes like [[YouTube Titles]], and ask it to suggest titles that work well with my video idea or existing script. With less time worrying about packaging, I can continue focusing on creating 😋

So far I’m thinking my process would look like this: - Create an agent - Manually look for these notes myself through keyword search and browsing links - Use [[Smart Connections Plugin]] to find semantically similar notes that could be useful for context that I may have forgotten about - Feed these all into [[Claude]] as I continue working on a project related to the information (in the case of YouTube titles, when creating a new YouTube videos)

If you have any ideas or experiences with this, I’d love to have an open discussion in my Discord community :)