Computer programming, and friction as a learning activity

R.D.Laing apparently said that we are all entitled to our thoughts. So, like you, i also have a lot of thoughts. This post is about some self-reflective thoughts i have been having, and finally i have i think come closer to what i have been feeling about the way AI a.k.a. our agents have taken over software development.

I recently had a chance to reflect on my relationship with software development as a thing I do and as a career. I will refer to both of them as computer programming for the rest of this post.

I will leave out all the career trajectory, progression, junior, senior, principal, senior, architect, promotions, etc out of this post.

Introduction

So, computer programming. I have been programming in one form or another since 1995. My first program that i can remember that really clicked it for me was to simulate a physics experiment around convex/concave lenses, and refraction using GW Basic. Then, I picked up a few languages. But overall, it was essentially, an act of manual typing by hand, reading from books, learning from books, or a teacher, and then writing by hand literally on paper, and then on a computer. One of my favorite stories for myself is i read a book on HTML (HTML complete) on one night, and i knew all i needed to know about html. I loved the beautiy of learnings like those. Similarly, I wrote lots of TSRs in C, and wrote Kernel modeules, compiled Linux from scratch for adding bluetooth support! What fun.

Things have mostly been the same for me and I have kept doing it since i have enjoyed the process of problem solving. A specific type of problem solving. Features don’t interest me, they bore me, soon-ish. But, give me a bug, and then i will get so stuck on it that everything else goes out of the window. That was what prompted me to switch from software engineering roles to DevOps/SRE roles because I would then be responsible for writing tools/frameworks for spotting those bugs. But that became boring too after a while. However, i could see myself doing that for a longer period of time than working on features. Sure, i have not worked or didn’t have it in me to work for all the advanced software companies out there, so I don’t know if i would have liked to work on features for those companies. Either way, i do feel that features essentially will become boring for me at some point, cause it is really the happy path of things.

Now, what kind of a programmer have i been? I am the kind of programmer who loves the problem solving aspects of the craft. That it is a tool to solve problems. Hence, if you ask me to solve leet code, i will suck, i have sucked for as long as i have programmed. I couldn’t code a quicksort for you, or a merge sort or those large number arrays, etc basically, i cannot and haven’t made it through a lot of programming interviews. Hence, code quality - the design patterns, or writing amazing code, has never been my focus. I love unit testing, and i would definitely want to practice TDD if i wanted to. I prefer typed languages like C, Java, and Go over Python or typescript (not a typed language, something that is trying to transpile to a untyped language) - the latter two are all i code in these days, but i am flexible. And so i have worked with a few people who have been so nerdy about their type requirements that i have either hated them or felt very inferior to them, often a mix of both.

Overall, to summarize, it’s fair to say, problem solving has been the reason i have been programming since i first learned about it. Types, code quality, good design, they are all nice to haves, but i am not gonna debate anyone over your way or my way. It’s the act of writing code, building up an understanding of a system - however small it is, and then knowing how it works, the next time i come back to it, that’s what I love about computer progrmaming. It is also an act of mindfulness for me, i zone in when i am coding by hand. The noise of the brain goes away. I am developing an understanding of what i am writing. I am feeling involved. I am part of the solution designing.

That leads me nicely to my relationship with a few things:

1. Third party libraries
2. Auto completion and configuring tools to make programming easier
3. AI based Inline code assistance and auto generation
4. Agentic development

Concurrently, interleaved with my software engineering “career”, i have also been writing articles in various publications (since early 2006/7), giving talks at conferences and writing books as another track to develop my understanding. As I have shared in another post on this blog, that has been how I have been able to develop a understanding of a lot of things. The friction involved in an often implicit and assumed understanding and translating that to a form for someone else to make sense of, has resulted in a lot of pleasure for me - the revised, often fuller understanding of things that came out of any of those exercises gave me a lot of pleasure. And I only realize it when I stop doing it. For example, this blog itself - there was a two year gap where I didn’t post anything here.

Third party libraries

I started my programming with GW Basic. We will ignore it for the rest of the post as it is not relevant.

C, C++ and Java were the first languages I coded in for the first few years of my programming life. I learned them all and wrote various programs. I coded in Java to contribute to NetBeans IDE. I wrote C/C++ during my time at Red Hat contributing to lshw and wrote some Linux kernel modules, tried to write to MySQL storage engines during my time at Sun Microsystems, etc. The key aspect that is relevant in this context was, all my programs were all in most cases, using the standard libraries - glibc, libstdc++, JDK and perhaps a few very standard third party libraries, may be stl.

However, around the 2005/06’s I started hearing about the Python programming language. I jumped into it. Working on various projects to learn a new language. It is with Python, that i first started noticing that third party libraries were a big requirement. In fact, as the years progressed, i could see various shiny open source projects coming up, libraries for Python. And that’s where I first i think felt like why do I need to use this another random library to do the work I need to do. Despite Python saying batteries included and I actually liked what it already had, the wider world appears to be gravitating towards third party libraries.

And that was when I first felt this pang of, “What am I gonna do?” - just put these things together and build the application?

And then fast forward a few years later, I discovered Go and loved it for its standard library. And as my book will show, or the projects i created for Manning, the usage of third party libraries are minimal - by a conscious choice. My goal is to teach things, and how they work, not to build applications.

I will not do this section justice if I don’t talk about abstractions in this context.

So, when i felt that pang of “what am i gonna do?” in the context of third party libraries. What I am realling feeling is that, “don’t abstract things away from me”. However, by using a high level programming language, I am accepting an abstraction - whether over the JVM, or the assembler, or the Python bytecode compiler. I feel Ok about that choice. As much as I love working for my results, there is a limit to my love for working hard and also achieving outcomes.

So, accepting abstractions are i believe a necessary choice to live in this world and for sure in the world of computer programming. At some point, i have desired to write my own programming language etc, but I got bored with the idea of inventing my own syntax. I also tried to write my own operating system back in the day, but never made it past booting a desktop from a floppy disk.

Autocompletion, snippets etc

The next “issue” for me or the wall I ran into as I worked with larger projects was the problem of customising my editor or IDE to do fancy things or learning all the keyboard shortcuts. I used the word fancy because that’s what they appear to me. What they really were, were ways to make the work of coding less cumbersome.

Some famous person said that if they had a day to cut down a tree, they would spend 70% or 80% of the time to sharpen their axe. Well my brain is like, forget sharpening, i want to start cutting the tree.

And so I continued writing my code, using the basic code completion and occassionally doing some minor “fancy” things, like when using Emacs C-x C-p would insert, import pdb; pdb.set_trace() when working with Python code. My favorite way to debug till is adding print statements, and i like gdb, pdb etc but really when I am working with applications such as servers, i don’t know printing is the way to go.

Besides the friction and the time needed to customize my IDE/editor - because I am simply not going to copy someone else’s configuration - I need to know what’s happening, I also was protecting myself from the distraction of suddently 10s of options coming up while I am coding, right in the middle of my thought process, I don’t want any interruption - from my editor. I am in the zone, i might be struggling, but i want to power through the struggle. Hence, the era of auto completion, snippets, customising editors in the perfect manner, I didn’t join that either. Of course, I made false starts, dotfiles etc, copied other people’s .vimrc or .nvim and watched with admiration and jealousy with a mix of inferiroirity complex the efforts other people went to.

AI and Inline code suggestions

Now, considering my relationship with not necessarily intelligent auto completion, i didn’t fare well with the AI driven inline code suggestion and completion. I did it for a bit, and then once again I realized, i want no disruption to my workflow. So, I turned that off and started using manual chats/inline chats to generate code and then manually add them. That meant somewhere in there, i was still in the loop.

In the last major project i worked on where I was building out the chat backend for my startup’s products, I sparred the design manually, listing out the API endpoints and once I was sure of what was going to happen, I generated the blueprint without any interference from me. However, once the skeleton was generated, I then manually coded the implementation themselves, including the DB schema, the integration with a third party service provider we were relying on. I felt good about it, since the overarching design had me involved, in my head i know exactly what the call flows are like, and thus, crucially, the breaking points. Where can things fail, and if they do, where do I start looking.

The exception to this me being in the loop workflow is frontend design/CSS code. I will try and not interfere or even try to understand the different coloring/positioning/opacity, only relying on the output. So, this is where again I make a careful choice of - what do I need to know? - I have chosen that I don’t need to know any of the amazing engineering that’s going on in the frontend.

Agentic software development

So, finally, I am here. I have finally decided that for some parts of projects, I am going to hand over things. For example, for the chat backend project, now that the basic infrastructure/design/blueprint is all in place and things work, any incremental work i now do i have stopped coding by hand.

Part of it is i know the critical paths, so now the incremental work being done now is not something I will not know anything about when I need to look at it.

The other part is, I am choosing at every point, “is there anything I would like to learn here?”. If no, hand it over. For example, when i code for my research work, I am coding by hand, but when I implement some software, that i might be building on top of my research, i am handing it over.

A bit of it is the ability to claim to myself - that i know what’s going on, and a bit of it is, this is my project, i want to work on it.

Another overarching question that i keep asking myself when i make a choice is - what is my goal here? is it the outcome i am interested in here or is it the process that motivates me?

When building a startup, i have finally realized (finally!) that it’s the outcome, and some projects, or some parts of projects, it’s the process, but some parts, it’s the outcome. Perhaps it’s a dance, a dance of flexibility.

When working on my research, it’s the process first and foremost, the rigor, the friction between what I know, what i don’t know, and what I need to know to go where I want to go - that’s where my joy lies.

Writing books and creating teaching materials

I have alluded to my joy in the friction involved when I am writing code. I am also channeling well known people when I say that, to really learn something, I have always tried to teach somebody that. Somehow I read it after I had already started doing it. My initial motivation for publishing articles, and then books, and giving talks or creating training materials was to be known - become famous on the internet. Become somebody.

That i have realized, perhaps, since it hasn’t happened yet or it is not what I likely have been doing it, or wanted to do it in the first place.

I feel the real reason I like it, and have liked it, is because the friction introduced in creating any kind of material for an audience is how I challenge, validate and clarify my own understanding. I looked at some trends in my writings over decades and the talks I gave, they all try to demystify something, or bring two areas together, or small deep dives into things that randomly walked into and then I wanted to shed more light on them.

So, once again, it appears that the friction of coding, and anything to do with coding is what I seek in coding. The friction is where my learnings have happened and I know how valuable it is to me.

Hence, i have this itch of working on a new book again, i have no idea what i want to write about, except for a few ideas. However, I know what I am seeking, it’s the wrangling of the ideas, the wrangling of the unknowns and somehow make it all legible for someone else to consume.

Conclusion

The turbulence started with me being angry about anyone telling me to use AI to write my applications.

Then, I wanted to give my own anger a moral - societal voice - so i made it (in my head, of course) about other people who are juniors starting in the industry. That they paid money to become software developers and now suddenly everything has changed. I even created a nice story too for myself - i was like one of these software developers 15 years back, etc. And I wanted to do something for them. So I wanted to write a book to help them.

I felt justified and righteous. What an amazing mix!

And then finally, I chose to come back to myself - this self that has witnessed and only knows about what’s happening for me. I wanted to write this post to help me clarify my own emotions.

I feel quite surely at this point, as much as all of that is true, and I don’t know if i will do anything at all to solve that particular problem to help other people, but it is/was really me mourning the learning involved in software development/computer programming as I experienced it. And yet, the reality is all these resourceful software engineers, the ones who are right now in the middle of it all - they will discover themselves and what they like to do - are they builders? Business persons? Or do they love the problem solving aspects of it? It’s their journey.

Now, what is my relationship going to be with software development/computer programming when the practicalities of life knock on my door? I will find out. I am less righteous about it and bringing the whole problem home - what am I gonna do about it? Perhaps that’s a better way to tackle this problem (a very personal problem) than from anger.