Hey friends, today I am releasing my LLM-related Python scripts for working with youtube, podcasts, and articles.
These CLI-based scripts have boosted my consumption and production process so much.
Here’s the link to the github repository: Nootroscripts.
Check out the comments in the code. It should help figure out the usage. The code is solid in terms of practical usage but not by any means a “proper” engineering example. I left most of my working notes in it, of references, technical decisions, and other just-in-cases.
Developing these has helped me wrap my fingers around LLMs. I’ll delve into the technical aspects of it in the next posts if anyone’s interested.
But today I’ll talk about the original purpose and spirit of these scripts to just set the context.
In the next couple of posts in the upcoming month I will also share about the different ways I have been using them, the details of the different scripts and any notable components of each, planned features and functionalities, and some lessons and insights learned while developing and using these scripts, related to prompting and LLMs.
I have been battle-testing and using these scripts on a daily basis for close to 3 months. Then a good friend encouraged me to write about them and publish the code after hearing about me gushing about how I love my knowledge workflow now. I have indeed been wanting to work, learn, and think in public more, so I took up the challenge and spent some time cleaning the code and writing this post up.
But first: who is this for?
You will find these scripts useful if any of these are true
- you work with lots of information from different sources (aka, knowledge worker)
- you work with words
- you want to learn to read/think/write better
- you are a Personal Knowledge Management enthusiast
- you are interested in practical use of LLMs
- you are a content creator/producer who wants to boost your workflow
- you are a wordcel who wants to hone your craft
My why
The problems I was trying to solve when developing these scripts:
- am overwhelmed by things sitting in YouTube watch later, articles opened and bookmarked on browser, tutorials, substack, all the things to read, constantly snoozing it into TODOs… mem it, readwise
- feeling like I have too many things to learn, and keep up with
- with not enough attention span or patience to go through them all, even at 1.75x speed.
- when I listen to audio content, I still need to take notes in text so I can easily search for it and revisit
- want to be able to navigate the audio-video content more easily, jump around, drill down. opting out of fillers and irrelevant bits. but more importantly: revisit relevant meaty bits as much as I need
- cross-platforms discovery of ideas and insights available in audio-video content is still lackluster. these are things you can’t discover through conventional search engines or fabricating LLMs that can’t reference or vet for their sources. these aggregations currently mostly exist through trusted and like-minded curators (that you need to discover yourself) and are hosted by walled-gardens content-platforms
- I want to build actual “second brain” that goes beyond memorizing aka storing. but one that can “think”, i.e. synthesize across different materials and articulate the result in words.
- I want to get a list of definitions and concepts discussed. I listen to ideas in different disciplines ranging form business, communication, philosophy, technology, engineering, economy, science, fitness, psychology, and creative process. having a list of definitions discussed help me remember and build the model of how they interact across disciplines
- want to process and revisit all the existing corpus (journal, learning notes, blog drafts, things I’ve clipped)
- I need help in articulating my genre and making sense of what are the different topics and rabbit holes I have followed. I’m sure there are big buckets of underlying thread behind all my interests and work.
- clickbaits are wasting my time. so why not save some time, get LLM to read it for me and answer the question posed as the title, and still scratch the itch of my curiosity
- when I use instagram, I enjoy people watching. I am interested in keeping track of social phenomenon, how people are reacting to them, getting a sense of what’s popular, and seeing patterns in the topics
- help extract and drill down into nuances, absorb things in different set of words, not just summarizing
- help me produce content that draws from all these rich resources
Can you spend less time “learning”?
I initially thought, if I can just summarise these long articles, long podcasts, long videos, I can learn in at least 25% of the time.
But that’s not what I found.
Yes, learning exclusively from text is efficient, but it is not effective.
I found that the most effective process for me is actually not either hearing or reading. It’s both.
I need to hear and read the data different forms, different length, and different articulation in oder to best absorb the information and turn it into knowledge that I can then practice and turn into wisdom. I need to engage with the material through as many senses. This is not surprising to people in education.
Text is flat. Text has no intonation. Text demands more of our focused attention.
Some people, myself included, forget what they hear more easily. I remember things I see (therefore, read), better.
But I absorb information more easily through audio. There is just so much more information conveyed through spoken words than written words. They are the same words, but they land differently.
In a nutshell, the sweet spot for me is: audio for absorption, text for reminding, action for mastery.
That’s all for today. Let me know on Twitter, Instagram, or Github if you have any questions, requests, or ideas.
Have fun.
