Why I Still Blog as a Developer in the Age of LLMs: Balancing AI Tools and Human Creativity
Large Language Models (LLMs) have reshaped content creation across all media, for better and for worse. The ability of these LLMs to generate content at the speed of modern computing raises questions about the authenticity and accuracy of everything on the internet. At least, it did for me.
“How do you define ‘real’?” –Morpheus (The Matrix, 2001)
How do I adapt? As a software developer, is it even worth maintaining this blog when an LLM can churn out any post on any topic in under 30 seconds, regardless of my expertise?
LLMs are efficient. There’s no denying that. As a late adopter, I find myself no longer looking to Google to search a question. Straight to Grok, I go. And when it tells me I asked a great question, I can’t help but feel a little warm and fuzzy inside.
Pushed aside
At the time of the release of LLMs like ChatGPT, I thought it was a neat little tool. But back then, LLMs weren’t running on up-to-date data. I was also skeptical of its accuracy. I pushed LLMs aside after a while and continued my work.
Other tools emerged, targeting software development with specialized functionality. Do I use these tools to “help” me code? I was torn. “Isn’t it cheating?”, is what I thought. I pushed these tools away in fear of being called out for “cheating” by my peers. How dare I use a tool to generate boilerplate code or spot a race condition in a code snippet, right?
This was a mistake. I should have used all of these tools from the beginning. My recent adoption of LLMs in my daily workflow has made me question the past few years of my career. Had my co-workers been using these tools to finish their tasks and bug fixes quicker thereby having an advantage over non-adopters (me)? Possibly.
My use case
Now, I look to the future. I may not use AI to produce mission critical things, but boilerplate code and quick scripts for one-off tasks? Sure.
“As long as you understand the code and what it’s doing, it’s okay.”
We’ve all heard this line before when software developers ask the question of “Should I use…?” even from the StackOverflow days. This is exactly what I’m doing but with a twist.
I assume all code from an LLM is mediocre at best. If my code is mediocre, others’ code is mediocre, your code is mediocre, and LLMs learn from all the mediocre code out there, then the code I get from an LLM is bound to be mediocre too. :)
I look over the code and get a high level view at what it’s doing. If anything feels or looks off, I put it through the wringer and fix it. I often find that LLM-generated code lacks proper encapsulation, so I almost always need to tweak the script for better maintainability. Or maybe I don’t tell it to encapsulate? Thoughts for the future…
LLMs make it so easy to learn. Gone are the days clicking through links on a Google search result set trying to find an answer. My learning pace has skyrocketed so much with LLMs that I had to subscribe after hitting the free plan’s limits. LLMs have made it easy to connect concepts at a high level while giving me the necessary research topics to dive deeper into.
Blogging
It doesn’t stop there for me. The very outline for this blog post was generated by a LLM. I work better when I have something my mind can spring ideas off of. Do I follow the outline rigidly? Absolutely not. I use it for structure and ideas only.
Tutorial blog posts used to be really popular and a great learning resource. Are those done now? But if they’re done, then what will the LLMs use to learn from? So I guess they’re not done but where do we draw the line?
This blog post could have been written by a LLM in less than a minute. So what’s the point? I’ve been asking myself this lately. I think I have the answer…
The point of it all is the human experience. LLMs can be cold and feel disconnected. As introverted as I am, I still like human connection. I think it’s important blog posts have some amount of soul. I type like I talk and talk like I type. I’d be kidding myself if I thought an LLM couldn’t mimic my tone.
The future
So what’s the future of this small blog and my little space on the internet? I still don’t know.
I’m not keen on doing tutorial posts anymore. But, they may be a learning experience. That would make it worth it. Perhaps future posts will be more opinionated. That would force me to be a little vulnerable. I’m not sure how I feel about that. Maybe I’ll cover things like how Apple lost all its mojo in their products, or how Arch Linux and Neovim are the way to live life.
Regardless, I hope to utilize LLMs as an assistant to blogging and not a crutch! Keep the human experience!