The more time I spend on social media reading posts by AI influencers, the more I see web developers (like myself) worrying that they’ll be out of a job in less than a year.
It reminds me of something I read a few years ago, the golden rule of software development:
No matter what the question is, the answer can almost always start with ‘It depends…’
Whenever a developer asks me if I’m worried that AI will replace me, I find myself saying, “Well, it depends…”
We’re standing on the threshold of a web development revolution
I think about the recent progress of AI the same way I think about the first steam-powered road vehicles (or what we commonly refer to today as cars).
Back in the mid-19th Century in the UK, people were concerned that compared to horse-drawn modes of transportation, these steam-powered vehicles would block up roadways, endanger public safety, and cause fatal accidents.
This led to a series of parliamentary acts that heavily restricted the use of road-going steam-powered vehicles. The most restrictive law was the Locomotives Act of 1865, which required vehicles to travel at a maximum of 4 mph (6.4 km/h) in the country and 2 mph (3.2 km/h) in the city, as well as requiring a man carrying a red flag to walk in front of road vehicles hauling multiple wagons. These acts effectively halted automobile development in the United Kingdom for most of the 19th century.
While fears regarding automobiles were justified — and eventually realized — cars were gradually accepted as they became more common and accessible. Simply put, automobiles made it easier for products to get from point A to point B. I think we can all agree that in some places, especially cities, too many cars on the road can be detrimental. We also can’t deny their impact on our society and how many aspects of our daily lives are made better by having a personal road-going vehicle.
AI usually gets it right
I’ve been fortunate enough to have used various forms of AI to support my work since early 2021. Back then, OpenAI had released GPT-3, and I was using it regularly as a Technical Writer for blog posts, tutorials, and documentation. Later that year, GitHub released CoPilot, and I was able to snag an open-source maintainers license, which I’ve used consistently for code generation. Since then, practically anything I’ve worked on has been AI-assisted. Generally, it’s been a positive experience.
I’ve successfully built everything from plugins to web-based games, and even a few personal utility desktop applications, all using AI. My most recent success was building the WordPress.com Content Calendar in approximately 2 hours, utilizing Perplexity for R&D and Cursor for the build.
But when it gets it wrong…
Every now and then, though, AI loses the plot.
I’ve read stories of people vibe coding SaaS apps, only to have their customer details leaked after the apps were hacked due to multiple security vulnerabilities. Recently, an AI agent deleted a company’s entire production database during a code freeze. I’m sure we’ve all heard stories of AI chatbots and agents confidently saying the code is fixed or working when, in reality, it is still broken.
Earlier this year, I used Cursor to build WP Debug, a small Electron desktop application that allows you to read and display the WordPress debug.log more cleanly. Someone opened a ticket to request adding the GitHub repository URL to the app’s About window. Because I had no Electron experience, it took me the better part of two full days just to implement this small request, something that an experienced Electron developer could probably have completed in an hour. Each time the Cursor agent confidently told me the About window had been implemented correctly, and each time I tested, it failed miserably.
Last week, I was using Perplexity to help me create a lesson on anonymizing production data. As I was reviewing the generated output, everything looked good until I got to the section on anonymization strategies. Perplexity had generated this:
WordPress provides several built-in sanitization functions that should be leveraged in custom anonymization scripts:
// Sanitize email addresses
$sanitized_email = sanitize_email( $raw_email );
// Sanitize text fields
$sanitized_text = sanitize_text_field( $raw_text );
// Sanitize file names
$sanitized_filename = sanitize_file_name( $raw_filename );
// Sanitize HTML content
$sanitized_html = wp_kses_post( $raw_html );If you know anything about WordPress development API functions, you know that sanitization is not the same as anonymization. If you want to see how wrong a vibe-coded project can go, check out my first attempt at building a browser-based 3D shooter game. Fair warning, only attempt to open the game in your browser if you have a lot of free RAM.
AI isn’t going to replace developers who adapt
Some years ago, when I was still a freelance developer, I had a client who built food blogs. She maintained a couple of single-purpose plugins that managed specific functionality, such as adding a recipe custom post type and various other tasks she required. I would hear from here when she needed something new added to any of these plugins. She knew what she wanted, but lacked the coding experience to build it; however, she excelled at providing specific requirements and thoroughly testing each addition. I can almost guarantee that today, she’d be able to manage this all herself without needing to hire a developer.
Just like those first steam-powered vehicles, AI-assisted web development is going to make what takes hours possible in minutes. Like any new technological advancement, there will be concerns about the use of AI to develop software, including its environmental impact, the security of the code it generates, and the impact it will have on the skills and abilities of human programmers. Its adoption is, however, inevitable, and to ensure that this AI-infused future remains a productive and positive environment, it relies on the developer community to evolve and adapt alongside it.
What you should be doing now
Like any new technology, there are two key takeaways. First and foremost, take heed of the hype, but learn to distill it down to the facts.
Social media is ablaze with folks who are trying to sell a future where the human element to AI is non-existent. It’s important to recognize these for what they are, the modern equivalent of those infomercial-driven late-night shopping channels from the early 2000s. Sometimes what’s advertised will end up being useful, but a lot of it will eventually end up on the digital trash heap.
AI, just like any other software, relies on humans to iterate and improve. Just like any other software, it’s prone to bugs that need fixing, new features that need documenting, and requires some training and education to operate effectively.
Second, learn how AI works, what it does well, and what it does poorly. Learn about prompting, context, tools, and all the things that make AI coding possible. That allows you to make valid decisions about when to use AI to streamline your processes, and when it makes sense to roll up your sleeves and do some manual work.
AI software experts recommend following good software development practices like clearly defining project scope, planning each step of the development process, writing tests as early as possible, constantly reviewing and validating AI-generated code, using robust security audits, and ensuring ongoing developer education to offset potential skill erosion.
At the end of the day, AI development tools are at their best when they are used as productivity aids — never as replacements for human expertise or oversight. The future of web development will be shaped by those who embrace these tools with a healthy dose of skepticism, understanding, and a willingness to learn.
This post was written by a human, assisted by Grammarly for grammar and spelling, inline wordsmithing using Google Gemini for clarity, and research via Perplexity.
