They said artificial intelligence would save companies millions. They promised smarter code, faster development, and a revolution in how software gets built. The suits in corner offices believed every word.
I believed it too, until I didn’t have a choice.
The Day the Music Stopped
Thirty years. That’s how long I wrote code for one of the biggest names in finance. Three decades of building systems that handled billions of dollars in transactions. Systems that actually worked.
Then one Tuesday morning, I received a call from my HR manager. The speech was rehearsed. Budget constraints. Restructuring. Nothing personal. I was being replaced by a team of four programmers in India who would cost less than my salary alone.
The decision came from executives who probably couldn’t write a single line of code if their stock options depended on it. But they could read a spreadsheet. And on that spreadsheet, four workers for the price of one looked like genius-level strategy.
After three decades of keeping their systems running, I was a line item to be optimized.
After months of looking for another job, I was forced into early retirement. No one wanted to hire someone of my age and salary requirements.
Testing the Replacement
With so much time on my hands and nowhere to be on Monday mornings, I started thinking about what everyone kept saying. Artificial intelligence is the real future anyway. Forget offshore teams. The robots are coming for all of us. That’s what the think pieces said. That’s what the breathless tech reporters promised.
So I decided to run an experiment.
I built a mini financial application, something like Quicken but simpler. A clean little program that tracks income, expenses, and generates basic reports. The kind of thing I could write in my sleep after 30 years. But here’s the twist. I planted bugs. Not obvious ones. Real bugs. The subtle kind that slip through when you’re tired or distracted. The kind that cause real problems in production.
Then I went to the machines everyone says will replace us all. ChatGPT, Claude Code, and Grok. The supposed future of programming. I gave each one the same task: find the bugs and suggest fixes.
ChatGPT: Creating Problems While Solving Nothing
ChatGPT went first. I fed it my code and asked it to identify issues. It came back confident. It always comes back confident. Found several “potential problems” it wanted to fix. None of them were the actual bugs.
When I pointed out where the real bugs were, it apologized and offered solutions. Those solutions created three new bugs while fixing exactly zero of mine. The code it generated looked plausible at first glance. That’s the dangerous part. It reads like it should work. But run it and watch things fall apart in creative new ways.
This is the tool that’s supposed to replace senior developers with decades of experience?
Claude Code: Same Song, Different Algorithm
Claude Code performed basically the same dance. Missed the bugs entirely. When I showed it exactly where the problems were, it generated fixes that introduced more issues than they solved. Different bugs than ChatGPT created, but bugs nonetheless.
Both of these systems have been trained on millions of lines of code. Both can explain programming concepts in clear language. Both can generate boilerplate code that looks professional. Neither could actually debug a real program with real problems.
The artificial intelligence that’s supposed to eliminate programming jobs can’t actually do the job.
Grok: The Most Human Experience of All
Grok was the most interesting failure. And the most frustrating. I showed it the bugs directly. No hide and seek. Here’s the problem, I said. Please fix it.
It came back insisting it had fixed everything. The code was identical. Same bugs. Same problems.
I pointed this out. It apologized and provided a “corrected” version. Same exact code. Word for word. Character for character. But absolutely certain this time it had solved everything. 100 percent confident.
This back-and-forth went on for 20 minutes. Me pointing out nothing had changed. It insisting the fix was perfect. Me showing proof the bugs still existed. It providing the identical code again with complete certainty.
The experience reminded me of something. Not a machine at all, actually. It reminded me of arguing with an H-1B worker who had been given a script and refused to deviate from it no matter what reality showed. The same circular logic. The same false confidence. The same inability to actually solve the problem while insisting the problem was solved.
What This Actually Means
The people who run things want you terrified of artificial intelligence. They want you thinking your job is obsolete. They want you believing the machines are so smart that human workers are just expensive legacy systems waiting to be decommissioned.
Why? Because scared workers accept lower wages. Scared workers don’t push back when their departments get gutted. Scared workers compete with each other instead of questioning why executives who can’t even do the work get to decide who’s valuable.
Here’s what my experiment proved: We are nowhere close to artificial intelligence replacing real programming work. Not even in the same galaxy. These tools can’t find bugs they aren’t specifically told about. They can’t fix problems without creating new ones. They can’t reason through complex issues or understand context the way a human developer does.
Could that change someday? Maybe. But someday is not today. Someday is not next year. Probably not the year after that either.
The Real Threat Isn’t Artificial
The actual threat to your job isn’t artificial intelligence. It’s human decision makers who care more about quarterly earnings than quality. Who would rather pay four people poorly than pay one person fairly. Who treat decades of institutional knowledge as a disposable commodity.
They’ll use AI as the excuse. They’ll use offshore teams as the excuse. They’ll use whatever excuse makes them look smart and forward-thinking while they pocket the difference between your salary and your replacement’s cost.
The robots aren’t coming for your job. But the people writing the checks absolutely are.
Trust Nothing They Tell You
For three decades, I kept systems running that handled serious money. Real consequences if things broke. Then I became a budget problem. Not because I couldn’t do the work. Because someone somewhere decided the spreadsheet mattered more than the reality.
Now those same types of people want you panicking about artificial intelligence. They want you thinking you’re obsolete. They want you desperate and willing to accept less.
Test things yourself. Build something. Break it on purpose. Ask the artificial intelligence everyone fears to fix it. See what happens. Don’t take anyone’s word for it. Not mine. Not theirs. Especially not theirs.
The evidence is pretty clear once you look. We’re a long way off from AI taking over the world. But we’re right on schedule for people in power taking everything they can while telling you it’s inevitable progress.
