No, ChatGPT does not replace a real living programmer today. Developers are safe for now.
If that is the answer you were looking for, job done!
How many site owners, though, are going to nuke their own website (or bank account) before they realize?
What is the problem?
I keep seeing people selling guides and courses about how you can use ChatGPT to enhance your website functionality without any programming knowledge.
Here on the right is just one example (click the image to embiggen).
My reaction is always to think how many times I have seen actual developers introduce vulnerabilities and code conflicts by mistake, how easy would it be for one of these course students to wreck their site?
Maybe not today, or next week, but months after they activate their shiny new plugin?
End the struggle, frustration, and overwhelm.
Get the help and advice you need to take your business where you want it to be.
Putting Theory to Test
Today I tested my theory out.
I hand-coded a WordPress plugin in regular PHP with no additional libraries by following the developer integration documentation at LifterLMS and Aweber.
The plugin signs up a newly registered free member resources user for my free newsletter. Straightforward stuff.
When I had a working proof of concept, I asked ChatGPT to create a similar thing for me.
To be honest the result was impressive … but still WRONG, and worse, insecure.
And the process was slow. I had to keep refreshing and use the “Continue Generating” button a lot, as well as give the tool a lot of guideance about what it should include, and essential changes it needed to make.
So it wasn’t particularly quick, and it was producing broken code.
When it was wrong, it was not always wrong in an obvious, easy to correct way.
No QA process would understand the entirety of what wasn’t working, so it would have needed detailed debugging.
Even as a developer, I couldn’t just look at the code that was generated and say “here is the problem” unless I had already immersed myself in the docs.
What is it Good For?
Don’t get me wrong, as I say it was impressive, but mostly for boilerplate or code-snippets that you would find in the official docs or Stackoverflow. That’s not suprising, these things are really autocorrect on steroids.
In fact, I think a lot of the actual working code was probably cripped from a Github repo or forum, and that is where the issues of security and being incorrect crept in.
My advice is to keep using ChatGPT etc, they are useful tools, but trust it about as much as a brand new intern coder.
Have someone with domain experience check, validate, QA, and put into staging.
Don’t EVER just upload, hit publish then hope for the best.