how to choose what to do

Posted: April 2nd, 2025 1:41 AM

The digital revolution is placing software and the hardware that supports it as the system that underpins all of modern society. There are countless opportunities to create technological innovation. At a minimum, there are apt opportunities to join a group who is attempting to bring about innovation. Originality is not a lost art despite a large population and growing number of STEM enthusiasts each year.

The idea is that as more people enter these fields as a whole, specialize, and form groups with like-minded individuals, then the overall surface area of frontier equally grows. That frontier may manifest as bleeding-edge AI technology or decades-old networking hardware. The antique internet appliances have a large frontier because the tech has not required significant changes, and thus it's surface area has remained stagnant.

A domain of expertise that is either poised for change or has made its way into casual nomenclature is a breeding ground for potential innovation. These domains today would be networking, AI, ML, data compression, data storage, and databases. These are a few of the domains that are either mentioned among manual labor groups as often as they are common language in Silicon Valley, or they will need to be drastically improved in order to support future technology that may be closer than we think.

Data and networking deal with the movement, manipulation, and storage of the lifeblood of the internet. Without data in the form of variables, loops, objects, and classes, there is no software. Without data in the form of articles, research papers, social media blogs, or videos, there is no pre-training data for a majority of LLMs in the world today. Without packets moving across TCP/IP and UDP, there is no functional internet.

So how then can one go about entering these fields and begin to master the necessary concepts to create something new? Operating from a framework of first principles helps to go from foundational concepts to marathon pace. Scott Young posits that it is the best use of time to pick a project, learn the absolute basic syntax of the programming language associated with the project, and then get as far as you can googling along the way until you can finish the project. At that point, you are best served by learning the theory behind what you just did. This is likely the most effective path because getting into the motion of learning basics, applying them quickly, finding answers to difficult questions, remediating issues, and seeing near-immediate outcomes exposes one to the feel-good feedback loop that is common throughout any of these technical domains. Each of these domains involves some layer or level of programming, so starting here can at least get you in the feel good feedback loop.

I teased the idea of first principles; a concept I am trying to use as a framework of operation in most things I do. The idea of first principles to me is that all processes should be broken down into their most basic ideas, primitives, and steps. As a mental exercise, we can take the topic of building a website:

Create a personal website -> Display text to a user who visits a url -> Put HTML and CSS on a Virtual Server running a Web Server -> Create a file for a page, configure the webserver to listen on certain ports for traffic.

This is the framework for breaking down what sounds like a complex task into individual processes and ideas. We can then take the ideas or processes and break those down further as we begin carrying out the tasks associated with accomplishing the overall goal.

In practice, this has proven to be an effective way to really learn about the domain. Coming to the conclusion that the best use of my time now is to study programming drove me to originally select the domain of AI. It seemed that not only was the field the most interesting of all, but also was comprised of individuals who I enjoy learning from. Extremely obsessive intellectuals; the kind of person I want to be when I grow up.

There is a lot to learn, an equally vast number of individuals to learn from, and AI is the most interesting topic in the world today. So I bought some books off Humble Bundle and began to read. However, in not having gone through the complete first principles exercise, I quickly realized that although the material was exciting, I spent too much time googling what the algorithms were. So I pivoted to a book on algorithms. Then I needed to pivot to a book on data structures. Then I had to switch to a book about algebra. Finally, I landed on a book about math from natural numbers, proofs, to basic linear algebra. Well, almost finally, I am now starting at the true beginning of this journey with George Polya's How to Solve It.

Throughout each step down in that mental ladder, I would also start a video lecture series and a podcast to increase the mediums of which I am gathering information. At each turn, I would have to go to a new corner of the internet to find the corresponding blogs, articles, papers, videos, podcasts, and lectures. This is what is required to reach a level of competence and performance in any field. Study and application over a sustained period of time results in compound interest returns. Studying is the logarithmic function, and application is the exponential. Together, you zero out somewhere around the middle quartile until you have enough applications of what you have learned to supersede the logarithm's diminishing returns.

One key observation I have had during this period of study is that while learning a technical concept, you are better off learning the concept in its own right rather than attempting to draw comparisons or use analogies to similar things you may already know. Metaphors, comparisons, or analogies will not help you as much as learning the material for what it is all on its own.

When learning Russian, I began by studying Cyrillic and comparing the alphabet sounds to what I already knew. This is somewhat helpful to understand the approximate noise you should attempt to make in order to mimic the sound. However, at a point, you are better off learning the sounds and how they join together in the way a Russian would say them. Mimic what the native speaker says rather than what you think it would sound like if an American tried to sound it out.

The same applies for learning AI, algorithms, data structures, or linear algebra. I have noticed that learning the material without using analogies forms new intra-material connections in my brain. It's almost tangible; the new neural pathways being paved into the hippocampus and across the cerebral cortex.

So, pick what is interesting and spoken about among the general public. Select a domain that genuinely interests you or that you can at minimum get enjoyment from participating in. Data, AI, and networking are not going anywhere as long as the internet is alive. Specializations within those fields would prove promising as well.

lesson

It was rather frustrating at the beginning of my attempts to study. I found it hard to listen to lectures, videos, or podcasts on the different topics. There was no one who would make an attempt to speak in layman terms. It took my emotional response from irritated to angry. It felt like everything was a foreign language and I had no possibility of learning that language. I then realized that the lexicon of academics is acquired by attempting to become an academic.

These writings then serve two purposes. 1) To help me at least feel like I'm becoming an academic and 2) To act as a medium between my brain and the world.

You are always better off rising to the level of others than demanding they come down to you.