I love reading Andy’s blogs because of the way he deeply thinks about the process of learning and how technology aids it. He has written about how we should re-think the textbook all together as a learning instrument amongst many other things. As someone who spends a lot of time thinking about how I like to learn things, his writings have offered a lot of perspective on how one should think about their learning practises.

What’s really exciting today is how much you can use AI to learn new things, and that unlocks this is AI’s ability to produce infinite representations of the same idea.

Humans usually have depended a lot on subject matter experts to explain things in various formats so that they can absorb them, it got easier with the internet. You could learn about the same topic from different people. You can watch a lecture by a professor on youtube or read the blogs or reddit comments from practitioners talking about the same from a completely different POV.

With AI, it becomes even easier, you can create almost infinite representations of the same piece of knowledege.

A lot of these thoughts bubbled up when I was trying to write some cron agents in my hermes setup to create curriculums for me to learn topics I am interested in.

I am a very prose-friendly learner, I prefer reading things to make sense of them. If you give me a bunch of pages that explain something like peeling layers of an onion, I am able to grasp things faster, if those pages have diagrams then it’s great, I was trying to setup flows where my agent would figure out this curriculum which is shaped very much to my ways of capturing knowledge, that’d be great.

NotebookLM is very good at this, you can give it papers and it can turn it into a nice deck of slides that can warm you up in a way that reading the paper won’t feel like walking into a dark tunnel.

I am very interested in turning this into a product that I can use to elaborate a desire to learn something and the agent goes out to build a curriculum, come up with exercises, block time on my calendar and keep track of my progress.

And this is what I feel is the actual end goal of all PKM systems, these were supposed to exist for two reasons – to learn so you can produce more. Whatever more is, more blogs better work, more creative output.

What’s the point of having complex looking knowledge graphs in obsidian if you are not using all that knowledege to do whatever you set out to do.

One of the things I love using AI for is to explain things in multiple formats or dimensions, if you’ve used NotebookLM you understand this well.

It’s very easy to distill any idea or explanation into a desired difficulty level

Which makes me kinda optimistic about AI in PKM.

Ideas can be sliced and diced into various formats

Edward Tufte’s book on the visual display of quantitative information. AI is adding so much multimodality to represent any piece of knowledge or information.

the speed at which ai can produce multiple representations of anything makes a very interesting case of products in the writing and education space.

Being able to visualise something a