Just as developers have long been susceptible to the ‘Makefile Effect’ when it comes to configuration files, the rise of generative AI tools brings a new risk of compounding a lack of understanding. Like copy-pasting Makefiles, using AI-generated code without fully following how it works can lead to unintended consequences.
Though it absolutely helps us move faster in building The Thing™️, it’s worth noting: Many of these configuration files are the result of taking a working version and copying and pasting them into our project, tweaking a few things until it works, and then deploying it.
As it currently stands, we may not be copying and pasting pre-existing files, but generative AI may be close enough (if not a step further): It produces what we need and, if it doesn’t work, we can just tell it to keep tweaking the script based on whatever error is returned until we have something that works.
It’s obviously not limited to configuration files, either. Let’s include functions, classes, libraries, or full programs.
Again, the advantage this gives us now versus just a few years ago is great but failure to understand what’s being produced has compounding effects.
To that end, prompt the LLM to explain what each line or block or function is actually doing and then consider adding comments in your own words to explain it. This way, future you, or someone else, will have that much more context available (versus needing to feed the code back into an LLM for explanation) whenever the code is revisited.
Perhaps this will help to resist the makefile affect as well as a lack of understanding as to whatever code is being produced and ultimately maintained.
To say that 2024 has been a year would be an understatement. Though I’m talking about things that have happened offline, the same can be said for the WordPress economy at large, too.
On a regrettable level, the degree at which I’ve written has decreased more this year than likely any other year since I’ve been writing. Some of this can be attributed to stage of life, some can be attributed to work, and some of this can be attributed to the rise of AI in our industry.
AI taking a bite out of WordPress (or something like that).
Though the goals of this question are not mutually exclusive, I think getting an answer fast often outweighs the “I’m looking for an answer but it was neat to also read about someone else’s situation while searching for it.” And this is why ChatGPT has “wrecked” some of the content a bunch of us typically write.
But, as stated, it’s been over a year since this was written. And since I work in R&D in my current role, we’ve done – and continue to do – a lot of work with the various systems, applications, utilities, and so on.
Given that, I – like many of you – have recalibrated my perspective on how this changes the work we do.
ChatGPT Didn’t Wreck Our Type of Content
Improved Productivity
First, it’s undeniable that when used properly, AI assistants can vastly improve productivity. I run both Copilot and Cody in my editor as I’m consistently evaluating which one performs best for a given use case. At the time of this writing, I’m partial to Cody though I also know Copilot is going to support multiple LLMs in the coming months (or weeks?).
So, sure, AI assistants have changed the way the work in our day-to-day but, as the months have passed, I’m no longer convinced it’s “wrecked” our type of content so much as it’s “drastically altered” how we explain – for lack of a better word – our content.
One of which is more neutral than the other.
Large Context Windows but Lacking Context
Secondly, for as much as I typically work with ChatGPT, Gemini, and/or Claude (is there a clever acronym for all of these, yet?) on a daily basis, I find myself continuing to enjoy well-written content either in newsletters (see The WP Minute, The Repository, or Within WordPress) or blogs (see Brian Coords, what Mike is doing over with Ollie, and so on). Though I’m but one person, each of these properties or people continue to publish even though LLMs are available for any of us to use.
And that brings me to the final point: There are reasons AI hasn’t completely wrecked the type of content I – and others – have often published:
AI hallucinates. Recommendations provided within a given LLM are presented with an authoritative sense regardless of if the recommendations even use hooks, function names, or language features that don’t exist.
Lack of context. LLMs do not have the context as to how a given developer arrived at a solution and why one was chosen over another. Sure, you can ask for a variety of solutions and tradeoffs but there are times in which it’s still faster to read from someone who’s had the same experience, shared it, and provided contextual information as to how and why they arrived at a solution.
Aggressive Autocomplete. I’m a fan of using coding assistants within my IDE. As I said, the level of productivity and speed of solving problems has definitely increased, but that doesn’t mean its attempts to autocomplete a piece of functionality are always helpful. It still takes a critical eye to review what’s being proposed and determine whether or not it’s worth integrating.
There are likely more and your experience likely varies – but I suspect aren’t much different – from mine.
The Why Behind the How
The reason I share all of this is because one of the fundamental things that is missed when working solely with AI is the value that human beings bring to the table when sharing the why behind the how.
This is not me taking a position on whether or not AI will, can, should, or whatever other argument is the current hot topic replace humans. Instead, it’s me saying that although I appreciate the value AI has brought to our industry and I recognize it alters the need for certain types of content, I no longer think it completely negates or replaces the type of content about I – and others – used to write.
Sure, our approach may need to be tweaked but there’s still plenty of ways to regularly share what we’re working on, how to solve a certain problem, and why one solution was chosen versus another.
Finishing 2024, Into 2025
Given that 2024 is coming to a close in the coming weeks and that we seem to have accepted the role AI plays in the day-to-day work of software development, perhaps I can start writing somewhat regularly once again.
There’s no shortage of things I’ve built, learned, saved, and archived. And while others have continued to publish their stuff, I’ve missed doing the same. Perhaps the coming weeks – and coming year – is a time in which those of us who so frequently wrote about development can find our way to back to doing exactly that.
Site analytics are funny things regardless of how you use them (that is, through marketing, engagement, content, and so on). I say this because analytics give us information about:
how long people are reading our content (per article, even),
how many people are reading what we write,
how much people are reading what we write,
how often people are returning to read what we write,
And all of this coalesces into informing the things about which we write and how we write about it. At least, this is my experience.
The day to day experience of understanding analytics through the use of two mice.
For technical writers specifically, this should give us pause on if there’s not a slightly wider range of related topics about which we can write that continue to contribute to the field in which we work.
Technical Articles in the Era of AI
As far as analytics, SEO, AI, and all of the other related technologies to blogging are concerned, I still hold to the mantra I’ve had for over a decade:
Write what you want on a given topic and don’t over think it.
This has proven the most useful and has transcended whatever changes have happened within the industry.
I primarily write because I enjoy the process, but it’s afforded opportunities that wouldn’t otherwise be available (the least of which isn’t developing solid friendships with people I’ve met in conferences or online).
Even still, I – like anyone else who’s maintained a site for a reasonable amount of time – still pay attention to some level of information analytics provide.
On Analytics
I rarely do a legitimate deep dive on the analytics of my site. Generally, I like to see:
the number of visitors over time,
how much time they are spending on the site or on each article,
and the bounce rate.
I’ve developed this habit in part because I’ve been writing technical content for so long it’s of some interest to see if people are [still] paying attention to it. But, as the advent of AI (see this post) has hit the mainstream of this industry, there’s been a change in how we all look up technical content.
And where analytics may have been useful for a very long time, there’s now another dimension to the field.
On AI
Given blogs and technical articles have helped train the LLMs we’ve so quickly adopted, it raises questions.
Is it useful to continue writing technical articles?
If the content we wrote helped train the LLMs then are new articles also continue to add to the data set the LLMs are using?
If more and more developers – myself included – are going to an LLM to help solve problems first (versus a search engine), how much less valuable are other sites and blogs becoming?
I still think there is value in treating a blog like a public notebook of sorts even if it’s just for personal reasons.
If not, then what else is there for technical writers to publish?
I don’t think there’s any shortage of content engineers have to write because so much of our job is more than just development, architecture, and so on.
The amount of things tangentially related or even adjacent to our work provide plenty of content that’s useful for other people to read (take articles like this and or this, for example).
As the technical aspect of our jobs may be enhanced – or substitute whatever word you’d like – by AI, there is still a human factor.
Clearly stated: As long as a human experience exists as something unique, it has potential to be an article that cannot be wholly generated through the statistical probability of words assembled through generative AI. (Though I’d be foolish to say that it can be a challenge to discern the difference between what a person has written and what has been generated.)
Writing technical articles does not have to be published into the void.
Though it may be easier to refer to ChatGPT for a technical question rather than a blog, that doesn’t mean a developer has nothing about which to write in relation to the field. For example, just as I could write about my day-to-day in working from home as a father of three an trying to maintain a schedule for reading, writing, exercising, music, work, and continued growth in what I do for a living, so can any one else. (Or so should everyone else?)
Software Developers Should Expand Topics
Ultimately, the way our work is altered through the advent of AI is undeniable. And though it may mean there are some changes in how we get our work done, it also informs how we can continue to contribute content related to what we do in our day-to-day. (This is something I used to do way back when, too.)
In other words, looking up how to properly sanitize data before it enters a database is going to be something the current – and the next – generation will ask an LLM. But looking up how to be productive as a remote engineer living in rural Georgia in a family of five, three of which are kids, is something AI cannot answer.
And perhaps that’s an area in which we could easily – and should – expand our content
Periodically, I review the content I’ve written over the last decade or so and am surprised to at some of the things I wrote about in the past (like My Day-to-Day). I also find it interesting that I stopped doing so. Then again, I likely exhausted that particular topic. At least for that time.
Specifically, I’m surprised that I used to write about such things despite the topics not really being relevant to what I consider my core content.
Personally, a lot has happened in the last, say, roughly five years alone – between changing jobs, growing the family moving, pursuing additional hobbies, and more – one of the things that’s taken a back seat is writing. Then again, though, isn’t that how it goes?
We have a finite number hours on how to spend our time and as that time gets allocated to other things, something gets squeezed out. And that’s what has happened with writing.
This gentleman fears the amount of time that’s passed. The perpetual ticking that surrounds him isn’t helping.
For a while, I felt guilty about it. Partially because writing daily was something that I enjoyed doing and that I did habitually. Partially because it had become such an habit that when I didn’t do it, I felt as if I was dropping the ball on something.
And though there’s truth in some of that (such as I miss writing every day), that doesn’t mean I’d trade out some of the things I’m doing that occupy that time now. Some of it’s related to my day-to-day work, some of it’s related to my family, and some of it’s related to other hobbies.
Over the holidays (and as I’m trying this, I realize I didn’t write a short Christmas post for this year which is likely the first time since I can remember not doing that), I had time to think about a lot different things some of which included both how I want to spend my time and how I currently spend my time.
Though I’m not one for setting resolutions, I’m for settings goals. And I was planning different goals for myself over the coming year, I couldn’t help but reflect a bit on this site.
Apparently, this is how Meta imagines me doing exactly what I just described. I’m drinking something out of a pepper container.
Sure, the goals would be fun to share (and maybe I will in a future post – I always enjoy what other people are planning!), I found myself thinking a little bit about software development, WordPress, the WordPress economy, where things have been, and where things are headed.
But writing about WordPress in the age of AI especially as developer is proving its own set of challenges. Of all types of people, though, shouldn’t we be here to meet it?
And with the rise of popularity in AI, the more-or-less standardization of the Block Editor, and the upcoming changes to the administration area UI, there’s a lot that can be discussed and there will likely be a lot about which to write (either via commentary or tutorials on how to achieve something).
When thinking through that, though, I found myself remembering all of things about which I used to write that weren’t always dedicated to programming but were still dedicated to what I, as a remote developer working in software developer in WordPress, was doing.
Why did I stop doing that? And what’s to stop me from doing that again?
I used to write differently about things. How did I get here?
Just as I do think tools such as ChatGPT has wrecked some of the content I (and others) have historically written, it’s by no means a call to inaction – or a call to stop writing. It’s just a call to adjust and keep moving forward.
Though I don’t know if I’ll ever write daily again, I do think there’s plenty I can share that extends beyond:
Here’s what you may want to do in WordPress using PHP or JavaScript
Here’s how you can do it.
So at the end of 2024, we’ll see how I’ve done. Here’s to a greater variety of content all the while still keeping the focus on the type of content about which I’ve historically written.
Last year, I wrote the first type of article that I’ve written in a very long time (if ever) given the amount of time that I’ve been writing. The Most Useful (or Popular) Articles from 2022.
There was generally derived from analytics data but also I used light engagement metrics via X/Twitter, LinkedIn, and even email to determine what were considered the most useful (or popular) posts.
This dude is completely scared of the fact that the calendar is nearing the end of another year.
And given that we’re nearing the end of 2023, I thought I’d do the same this year. So, in keeping with the previous trend, here are the most useful (or popular) articles from 2023.
2023: Most Useful Articles
The Architecture Astronauts of WordPress. Solving the problem is what’s useful. How you solve it is useful, too, but the architecture can be secondary to how it’s solved.
How To Import Programmatically Defined ACF Data. In this post, I’m going to cover how to import a programmatically defined ACF group. And in the next post, I’ll talk about solutions for migrating data into its own table.
ChatGPT Wrecked Our Type of Content. When you’re someone who enjoys writing about technical concepts for yourself and others, the ultimate goal isn’t to show someone how to quickly do something. More so, it’s to show how to approach problem solving in the abstract and to understand the tradeoffs of various solutions.
The Term ‘WordPress Adjacent’ Should Be More Precise. First because I believe it’s important for us to be able to clearly define about what we’re talking and how it relates to WordPress. And secondly, so that our audience can follow it.
You Rarely Get Greenfield Projects (Or Do You?). If courses teach you’ll be building more greenfield projects than not, I’d be hesitant to consider it a well-rounded education in software development.
Thoughts on the Upcoming WordPress Admin Redesign. There’s likely going to be some friction – as is the case to come with any major updates – but I’m optimistic that we’ve learned a lot from how Gutenberg was initially developed and rolled out and I hope to see others work even more closely with the administration experience to make the transition easier both for developers and users.
For the last few years, I’ve claimed that I want to write more than the year before and get back to how much I was writing in a few years prior. Truth is, I don’t know if this is possible given how much has changed since this. Work is different, life is different, and the way my day-to-day is structured is different.
It’s all great, but it’s different.
Regardless, I still urge everyone with a blog to continue doing the same and then syndicating out to the web. It helps surfacing your content in search engines, on social media – be it X/Twitter, LinkedIn, Facebook, whatever, – and on RSS (which is still my personal favorite).
Father Time wrapping up another blog post for the end of another year.
Anyway, these are the posts for this year. On to 2024.
Kinsta offers premium managed WordPress hosting for everyone, small or large. Powered by Google Platform and strengthened by Cloudflare, they take performance to the next level. All hosting plans include 24/7/365 support from their team of WordPress engineers. Get startedwith a free migration today.
If you are looking for WordPress guides and unbiased reviews, I recommend checking out WPKube. They also have an exclusive deals section.
Just Getting Started with WordPress? I write a lot about WordPress development but if you're just getting started, I recommend checking out WPBeginner. They have free training videos, glossary, and more.
