DEV Community: The Self-Taught Dev

To Uni, or Not to Uni, That Is My Question

The Self-Taught Dev — Wed, 20 Mar 2024 09:00:00 +0000

I didn’t go to university. It took me over a decade of professional life to discover something that I wanted to study. I’m now on my second career as a self-taught software developer.

I have proven to myself that I can make it as a developer without formal education. But I am left wanting more. It’s not about money: I don’t think six years of a part-time degree would produce a monetary increase that I couldn’t achieve through professional experience and grit in the meantime.

This dilemma has played on my mind for a while now. I want knowledge. I want to identify unknown unknowns. I am attracted to the structure that education provides. But I don’t want the financial burden of university — I especially don’t want to pay for modules I have essentially obtained through personal study and experience.

An idea began to grow: could I do this on my own?

A quick search on eBay confirmed I could obtain second-hand materials for Open University modules and other university-level textbooks are available for purchase online too. Large Language Models such as ChatGPT are potentially good-enough for my needs as a tutor/professor.

Just to be clear, AI cannot currently replace the expertise and wisdom of an experienced university professor. Perhaps one day. But not today. AI is, however, at a point where it can provide additional insight through investigative conversations, and confirm understanding through targeted Q&A sessions. I created my own “GPT”. I instructed it to behave like a CS professor and guide me to answers rather than regurgitate facts. I can even give it photos of pages in books and it will generate questions to test my comprehension! As I learn more about configuring GPTs, I will be able to improve its effectiveness.

Next, I just needed a curriculum. I used Open University as my baseline to linearise topics. I compared their various Computing & IT pathways and chose the most relevant modules for my career plans. Most of these modules have reading material available second-hand. I then used resources such as teachyourselfcs.com to identify additional materials to study. I spent some time deciding how deep I would go into each topic and disregarded a few recommendations that I felt would be overkill for my needs. I created a three-stage curriculum.

Then I began to diverge from the typical university pathway. I began to include books considered by most to be must-reads for anyone working with code. And I fit them around the structure I had already defined.

But there are so many non-technical skills that compliment a career in software: leadership, finance and business to name a few. So I created a fourth stage to capture these additional skills. This stage is unstructured and designed to be studied as and when I desire. If I need a break from data structures, I’ll study business or leadership. Why not?

I published the whole curriculum on my website and called it The Self-Taught Degree. Each ‘module’ has a tick box next to it so I can track my progress. I imagine it’ll take me years!

Knowledge for its own sake is futile. It needs a higher purpose. Yes, it will make me a more competent developer, but there’s also an opportunity here. Considering my ‘Self-Taught Degree’ will lack the status of a formal qualification, I need to prove my competence another way.

So as I progress through each module, I shall do what any good student does: I’ll take notes, converse with “Professor GPT”, summarise content and rewrite it in my own words to confirm understanding. And then I will publish it all online. My website has been designed to become a hub of knowledge as I go through this process. Being a resource for both myself and others on their own self-taught journey.

What a world we live in. Access to education has been a barrier for the individual progress of humans since time began. Those days are long gone. Now it’s a matter of grit.

Watch this space.

This article was originally published on my blog.

Will Artificial Intelligence Inherit My Software Dev Job?

The Self-Taught Dev — Thu, 08 Feb 2024 01:36:18 +0000

During the summer of 2018, as is common in the military, I was assigned a new role that required rapid skill development to be done by last Friday. As is not so common in the military, that skill was coding; and I was instantly hooked.

Work wasn't "work" anymore: it was adventure; it was exploration; it was discovery. I fondly remember the penny-drop moment when I wrote some code that looped and printed a statement ten-thousand times in an instant! The potential was exhilarating.

The Advent of AI in Coding

There were two significant moments when I realised change was on the doorstep:

A demo in 2022 of an early version of Github Copilot. You typed what you wanted the code to do and then watched as it magically generated before your eyes. I remember thinking that we had automated away the fun part of coding. Ironically, I have grown to love Github Copilot and often use it as an autocomplete for my thought process 🤫 But it does still fall over with new features of languages/frameworks—AI is only as up-to-date as its training data.
The first time I held a technical conversation with ChatGPT and the weight of what that meant. It helped me understand a new concept I was grappling with. I was able to validate my comprehension (or lack thereof) by repeating the concept back in my own words and having ChatGPT confirm or correct my understanding.

Contemplating the "Near" Future

This all leads to the question on my mind: if Artificial Intelligence can be so good at technical creation in this early stage, will it eventually inherit my job of actually writing code?

I don't actually know the answer to this. I'm not sure anyone does...yet. If the past has taught us anything it's that the future is hard to predict. But I think it's important to contemplate potential outcomes.

In the next few years, Artificial Intelligence could take on a role comparable to an aircraft's autopilot system: it'll do most of the heavy lifting but won't be trusted to do it unconditionally. Pilots in aircraft often do very little actual flying. They usually control take-off and landing. Then for the rest of the flight they monitor, make adjustments if required and—most importantly—are fully trained to take control in case of emergencies.

A day in my future dev-life might look like this:

Tell AI to generate feature X to solve problem Y
Look over generated code—and its related unit tests, of course—to verify it works as intended. If not, adjust and repeat step 1 ⚠️ potential infinite loop for a stubborn mind ⚠️
If all is well, merge code into codebase
If all is not well, oil programming hinges and put fingers to keyboard like I did in the good old days. Maybe ask ChatGPT to help...ahem

The human will have moved from a person who creates, to a person who guides: mastering the art of leveraging AI to generate vast quantities of code that integrates into the existing codebase. Teams could end up moving at a pace we can only dream of currently.

More Than Just a Coder

However, it’s so easy to get caught up in what AI can do that I sometimes can’t see the forest for the trees. The best coders write very little code. Some of the most important skills a developer can have are uniquely human.

For a lot of companies, the developers have to engage with end-users, elicit their problems, empathise with their situations and discover creative, bespoke solutions to their needs. There are times when customers don’t fully know what they need, and it requires discernment to separate the wheat from the chaff.

Developers communicate complex technical concepts and their business value to non-technical stakeholders; this may serve as the critical determining factor for the allocation of funds.

Consider solutions and implementations that are the result of two humans chatting casually about what they’re doing—the old “watercooler effect”. These situations are founded on the coming together of two different beings, often with drastically different outlooks on life. My music-teacher wife often provides a perspective on a problem that I hadn’t even considered.

What about leadership? Mentoring junior developers and onboarding new members of a team are obvious examples of leadership. But the developer who consistently includes general refactorings into their workload understands the value of quiet leadership—the kind that might get overlooked. They do it anyway because technical debt can strangle a project.

The role of a developer is rooted in technical skills, but it's the soft skills that bolster a team and keep business goals and technology aligned. AI is currently great at the former, but not so much the latter.

Ethical Considerations and Innovation

But what about further into the future? When AI can be creative intentionally, instead of on the back of hallucinations. When it can generate production-ready code consistently and communicate clearly to stakeholders. Could devs be out of a job?

These game developers are already pushing the limits of what AI can do. And my gut tells me that some companies will push it to the extreme. If there is a buck to be made someone will want to be the first to prove it. Let's look at some possible consequences if this happened across the industry.

Programming languages have "Core Devs": teams of people who maintain, update and improve the programming languages themselves. Their most important job is implementing security patches. They also fix bugs, improve usability etc. If AI inherits developer jobs, maybe it will inherit the job of a core dev too. Being responsible for the very code that it generates. Can AI be held responsible?

Maybe AI will do away with the need for programming languages and default to writing machine code. Arguably, all languages are designed to be human-readable and add overhead as a result. If a human doesn't need to read it, then why bother? You'll get more efficient code.

I hope this never happens—no matter the autonomy given to a system, we should always be able to verify what it's doing and why.

How will innovation fare? Innovation is usually driven by determined people who are convinced they know how and why something needs to change. They often face large amounts of push back until the world either accepts the inevitable or gets onboard with the idea. Could AI innovations be as impactful as the lightbulb, the telephone or the internet?

Maybe human innovation will flourish! There might be a world ahead of us where AI takes on so much of the mundane that humans are freed from the repetitive and empowered to innovate full-time. My job as a dev would be one of AI-empowered innovation.

The future is a mystery. Questions abound more than answers. This may always be the case. How AI will change my role as a software developer is up for debate. I hope the changes feel like improvements to those it affects.

Either way, I'm excited to see what happens.

Refactoring My Learning

The Self-Taught Dev — Tue, 09 Jun 2020 22:33:10 +0000

This article is a to-do list, of sorts, to myself. I've shared it here just in case someone else finds it useful too.

The Requirement

I've been developing for nearly two years now. I found myself in a learn-from-scratch-on-the-job situation which has one Major Pro and one Major Con. The Major Pro being all of the real world experience on a daily basis such as website deployment, legacy code, unforeseen UX issues, the list goes on. The Major Con is that there seems to be very little time you can dedicate to learning. So in what little time I do get to study I really want to maximise the return.

Some of the books I've read and online courses I've taken have really stuck with me but a lot no longer seem to have any place in my memory, even though the content was brilliant. This feels like such wasted time and aggravates me to no end! I needed to refactor the way I approached learning. So I spent a few hours digging through resources on the net (listed at the end) and I feel confident in my new approach. It's time to take responsibility for my own learning.

The Solution

My solution comes in four parts: Coding Books and Online Courses, Projects, Textbooks and Effective Note Taking.

Coding Books and Online Courses

When I say 'Coding Books' I mean books on HTML, CSS, JS, PHP etc. Books that teach you the language itself and common ways to use it. I'm also referring to online courses such as the ones on Udemy, YouTube, PluralSight, etc. These resources focus on teaching you how to DO something, rather than HOW or WHY something works.

First things first: I do not need to remember everything! There are too many languages out there. Each with a different syntax. So unless I suddenly become Dustin Hoffman's Rain Man I simply don't need to hold myself to such standards.

The key is to understand how the language works. How the pieces fit together to fix a certain problem.

So what steps will I for take coding books and online courses:

Regurgitate: Repeat the code examples and ensure the code works as intended.
Comprehend: Make sure I understand every word in the example. This step may require re-reading/watching the resource or even referring to official documentation. This step can give a feeling of real progress that comes from understanding what each piece of a puzzle is doing and how they all fit together.
Deviate: Play around with the code. Take something away. Add Something. What happens? Is it what I expected?
Innovate: Use what I've learnt and build something new. It doesn't have to be anything fancy. A small function or feature to ensure I understand what I'm learning.

Making mistakes is a key learning opportunity throughout all of these steps. Every mistake made is a chance to reduce ignorance. Embrace the failure, knuckle down and figure out what went wrong.

A Note on Snippets

A snippet is an extract of code that you know you will need to use over and over in the future. You can take that piece of code, plonk it inside a text file and store it for future reference. This is especially useful to prevent you from repeating the same Google search over and over each time you 'need that bit of code again'.

These books and videos provide lots of brilliant, reusable snippets. Record them. It will reduce the pressure of having to remember everything.

If you want to store your snippets in the cloud, there are some great snippet managers out there. I'm currently trying out 3Cols.

Projects

Learning resources often have a project built in to be followed step-by-step. These are great for learning how the code works, the author has designed it that way. But what is also important is running riot on a project of my own that solves a real problem in my real life. I will use the top-down approach for these projects: choose an end result and then work out the details.

Decide on a project. Make sure it is something I personally want or need. This way I will gain experience and a useful tool at the same time. Win Win!
Upon deciding which language/s I'm going to use, skim over a book or online course on that language and list the major concepts. This could be responsive design, prepared statements, ajax etc.
Break down my project into small chunks, each chunk requiring me to utilise a few of those concepts.
Have at it! (Get coding.)

This top-down approach is much more 'real life': you know what the end result needs to be and you must determine and execute the steps to bring it to fruition!

Textbooks

This section is not for how-to-code books. It's for high-level textbooks. I have a few Computer Science textbooks I plan on studying to supplement my developer knowledge. So this is my plan for tackling those concept heavy monsters.

These five steps are to be repeated for each major chapter in the textbook. The first four steps are about getting exposure to the topic. It is not about reading into the details at all! This is preparatory work to do before sinking your teeth into the content, which is step five. It builds a foundation on which to construct the rest of your understanding.

Flip through each page in the chapter. Ignore any specifics. How long is the chapter? Are there many diagrams?
Is there a quiz at the end of the chapter? If there is, read over it. Authors use these quizzes to test you on what they, as the experts, believe is the most important information to remember from the chapter.
Read all the bold text. This means headings, sub-headings, key words and key phrases.
Read the first and last sentence of each paragraph. Granted, this step assumes the book to be well written. If it is, these sentences should be the introduction and conclusion of each paragraph.
- By now you should have a good foundation as to what the chapter is about in general. The kinds of words and phrases that are important and the concepts that are most crucial.
Now it's time to get truly stuck in; time to pursue comprehension. Study the chapter fully. Take notes. Take my time. Ensure I understand. Thanks to the first four steps, my notes contain almost nothing that isn't relevant.

Effective Note Taking

This section's guidance is very generalised. I will use this method of note taking whether I'm learning a coding language, computer science or any other topic that peeks my interest.

Use the outline method, which I am using for this section and which looks like this:
- A Main Topic.
  - Sub-Topic.
    - Details.
    - Details.
  - Another Sub-Topic.
    - Details.
- Another Main Topic. etc.
Record all of the questions I have about the topic. This is extremely important as it highlights knowledge gaps as I learn.
Add to these notes by doing my own research.
- Reference up-to-date sources if my main resource is a bit out of date.
- Find answers to all of my questions.
- Be curious.
Create summary sheets of specific topics.
- If my topic is development related, I could turn it into a blog post of some sort as my way of summarising my learning. Similar to what I'm doing now.
- Focus on concepts, leave the details to the language documentation and my snippet library.
- Making my summary sheet visually pleasing will help (right now I like using markdown) but don't get lost down the rabbit hole of trying to obtain visual perfection.
Brute-force my long-term memory with flashcards. This is a whole topic on its own but I'll try to summarise it:
- Flashcards should be utilised once I fully understand exactly what I'm learning. This is when I will work to embed it deep into long-term memory.
- Create my own flashcards (preferred over using someone else's) and ensure they follow the "one card one question" rule.
- Focus on concepts. Memorising every syntactical rule of code won't make me a better coder the same way memorising a dictionary won't turn me into William Shakespeare!
- Anki is an amazing piece of free software for creating flashcards and using Spaced Repetition to memorise them.
  - Anki is available on a huge variety of platforms, including mobile. YouTube is full of useful Anki tips and tutorials.

Resources Used For This Plan

Tips On Learning How To Code

How To ABSORB TEXTBOOKS Like A Sponge

How to Take Notes | Science-Based Strategies to Earn Perfect Grades

OWN YOUR LEARNING

4 steps to learn programming faster and better