At this point, anyone looking into AI and coding has encountered the many posts where developers point out mistakes that originated with AI-assisted code. Heck, there’s even “vibe coding” a term I’ve immediately come to despise.

At the heart of all the criticism is the very real fear that “AI is coming for your job” and the generic “No, it’s not” responses from people who have little insight into what the future holds. For instance, I saw this post on my LinkedIn feed recently.

The poster, a software engineer, is claiming that a potential client who expects more out of a software engineer now that AI tools can code really quickly is wrong to want human coders to also work more quickly.

They think it is wildly unrealistic to expect a human to get smarter and faster at churning out what is mostly very generic and patternable code. I’ve spent about 12 years being a software engineer and most of the work I did had very real and common patterns. I also saw many buggy code released, and I personally caused many of those bugs over the years.

So what gives? Why are these developers being so defensive and cagey?

Professional Gatekeeping

Mostly the author of the post is simply keeping up appearances with other “coding experts” who of course are all humans and never make mistakes. They fear that salaries for coding will slide, and so they want to point you away from trying new tools, and instead keep interfacing with code the way they like: not at all, only through a human developer.

Throughout my 12 years of coding, I’ve experienced this phenomenon all the time. That’s why I tend to “shapeshift” as a dev and look into many different programming communities based around all kinds of frameworks and languages. In fact, I’d say I probably went to as many disparate events as an industrious tech recruiter would, but no “expert” ever needs to.

In fact, “experts” who are curious get called out all the time by gatekeepers since the curious must spend some time not in the same realm as the expert who just likes control and has very little curiousity to find out if they are wrong.

We Need A Scapegoat

They also need a scapegoat for their anxiety. The poster is clearly angry that someone is insisting they are misquoting how much time it should take to make the codebase production ready. Rather than deal with the fact that they are now “slow at coding”, they turn to tired tropes and maxims repeated at conference claiming that AI-assisted code never includes those concerns.

When, in fact, my experience is that AI coding assistants focus more on testing and security than 90% of the people I’ve worked with over the years. I generally don’t even have to prompt for tests, but if I add that to a readme or coding guidelines file, the AI assistant always remembers the guidelines.

No human I’ve even met remembers what they were told at the beginning of the job, and that’s why I prefer to work with an AI assistant to compliment my work with human developers. It’s not that humans are failing in this regard, it’s just that we can’t keep up with the processing speed of a modern computer so we should not try.

Instead, we should offload tasks to the AI so we can preserver more brain power to review code and think about how not to have bugs or security issues. Everyone who refuses to use AI will inevitably have worse code than the early adopters.

How do I know that, you might ask? Well, let’s take an example from the past shall we.

A look at the past

Get the girl to check the numbers.

John Glenn famously said this line in 1962 before orbiting the earth based on calculations from digital computers. At the time, human computers were still employed and doing a lot of calculations but the field was changing.

Glenn, like many others at the time, didn’t trust the machines and would point out any mistakes they made as proof that humans should still be doing the calculations and more trusted than the machines.

However, any math teacher can tell you this is false due to how many mistakes they seem their students make even with the help of digital calculators in modern times. Hell, even doctors are getting schooled now by machines predicting diseases sooner than they can with greater accuracy.

Despite the early criticism, there are practically no human computers these days and when you mention the term “computer” everyone thinks of a machine. But do we all need to worry we will become as extinct as the human computer?

Don’t Worry, Expand Yourself

I don’t think we have any reason to worry. In fact, I think we have reason to hold great hope in our hearts and minds. The switch from human computers to digital computers did not slow people down from computing things.

No, people sped up computing things. In fact, whole new fields of study emerged due to the productivity increases. Human computers had the ability to transition to statisticians and ask more involved questions about the data they were computing.

Think about it: would you want to hand calculate thousands of figures taking weeks to complete? Or would you rather have a machine calculate those figures in minutes while you explore relationships within the calculations?

Hence, this is why I treat code as a “side effect” now and not the source of truth. The source of truth has always been in the natural language we use to describe behavior and not the code that deterministically carries out the semantic intent.

And reasoning with language is far easier to do than sort through a codebase written by humans who probably never have the whole context of the codebase in their minds when making a change. Humans who fail to write tests time and time again. Who would rather overengineer code than conduct user research.

No, I don’t think we have anything to worry from these professional coding gatekeepers. Just like human computers, their anxiety will fade into the background and we will forget all about their protests in a new age where humans are far more capable than they are today.

Here’s to “vibe coding”! Keep doing it everybody, it’s working.