The Great City Deep Meltdown.
2026-08-18
🔑 When a Billion-Dollar Brain Cannot Leave the Left Side of a Picture Alone.
It was a cold, miserable Tuesday in City Deep, and the universe appeared to have decided that one particular human being had not suffered quite enough yet.
That human being was me.
I was sitting in the cold, feeling as though somebody had driven a freight train through my body, fighting a full-blown case of flu and running on very little sleep.
My total technological fortune for the day?
A phone.
A graphic I wanted to edit.
And approximately 400 MB of mobile data.
Not 400 GB.
Not unlimited fibre.
Four hundred precious megabytes.
Every megabyte mattered.
And naturally, this was the exact moment when I decided to test the magnificent future of artificial intelligence.
🎨 The Request Was Ridiculously Simple.
The graphic already existed.
I did not want a new picture.
I did not want a reinterpretation.
I did not want artistic inspiration.
I certainly did not want a cybernetic Picasso having a nervous breakdown.
I wanted one tiny change.
My instruction was essentially:
Take this image. Leave ninety percent of it alone. Go to the right-hand side, change one small element, and change its colour.
That’s it.
A five-year-old with a box of crayons could understand the assignment.
A 1990s graphics program could probably have done it before I’d finished drinking my coffee.
Except I didn’t have coffee.
I had flu.
And 400 MB.
So I summoned the Giant Cloud AI.
☁️ Enter the Giant Cloud.
Somewhere across the ocean, inside a facility large enough to make my rented room look like a shoebox, thousands of computers were presumably humming away.
Liquid cooling.
Massive processors.
Gigawatts of electricity.
Billions of dollars of infrastructure.
Artificial neural networks containing more parameters than I have hairs left on my head.
The machine received my instruction.
It looked at my image.
It considered my request.
It contemplated the accumulated wisdom of humanity.
And then it apparently said:
“No. I think I’ll make a completely different picture.”
The result?
A brand-new image.
My original composition had effectively been fed into a digital blender.
The thing I had asked it to change was different.
The things I had asked it not to change were also different.
And somewhere in the process, I half expected a space dragon to appear.
📱 Fifteen Megabytes Later.
Data consumed: 15 MB.
Data remaining: 385 MB.
Blood pressure: Increasing.
I tried again.
I explained the instruction more carefully.
“Don’t redraw the picture.”
“Don’t change the layout.”
“Don’t touch the rest of the image.”
“Only change this element.”
“Only change the colour.”
The enormous neural network responded with another magnificent act of computational interpretation.
It changed things.
It invented things.
It redesigned things.
It apparently became emotionally attached to portions of the image that I had specifically told it to leave alone.
Another attempt.
Another 15 MB.
At this point, my patience had entered a state of permanent deletion.
🤖 The Absurdity of It.
This was the moment when the whole situation became hilarious.
We are constantly being told that artificial intelligence is going to transform civilisation.
AI will manage financial systems.
AI will revolutionise medicine.
AI will reshape education.
AI will change the workplace.
AI may eventually become smarter than us.
And yet I was sitting in City Deep, freezing my backside off, watching a machine with access to an astronomical amount of computational power fail to understand the digital equivalent of:
“Don’t colour outside the lines.”
That is where the joke becomes more interesting.
The problem wasn’t necessarily that the AI was unintelligent.
The problem was that I had asked a generative system to perform a job that was fundamentally deterministic.
I didn’t need creativity.
I needed obedience.
I didn’t need imagination.
I needed precision.
I didn’t need a machine to ask, “What might this picture look like?”
I needed a machine to understand:
“This pixel stays exactly where it is.”
📲 Meanwhile, the Little Android App.
And here is where things became particularly embarrassing for the Giant Cloud.
A tiny graphics application running on an ordinary Android phone could perform the operation I wanted.
Select the area.
Mask the area.
Change the colour.
Leave everything else alone.
No gigantic data centre.
No enormous neural network.
No billion-dollar research laboratory.
No philosophical debate about the future of humanity.
Just:
Select.
Change.
Done.
The little app wasn’t trying to be intelligent.
And that was precisely why it worked.
🔑 The Sixth Sense: We May Be Using AI Backwards.
And this is where my City Deep meltdown gave me a slightly more serious thought.
Perhaps the great AI revolution isn’t going to be about replacing every traditional tool with an AI model.
Perhaps the real revolution will be learning when not to use AI.
We have become so fascinated with enormous models that we sometimes forget the humble software tools that already solve problems perfectly.
If I need to calculate something, I don’t necessarily need a giant language model.
If I need to rename a thousand files, I don’t need artificial creativity.
If I need to change one colour in one section of an image, I don’t necessarily need generative image synthesis.
And if I tell a program to leave 90 percent of an image alone, I don’t want it to exercise its artistic freedom.
Sometimes the smartest system is the one that knows:
This isn’t an AI problem.
It’s a paint-bucket problem.
🧰 The PlebWare Principle.
That idea fits rather neatly into what PlebWare is trying to do.
Technology should not be impressive merely because it is complicated.
It should be useful because it works.
A giant cloud model can be extraordinarily powerful.
A tiny local program can also be extraordinarily useful.
One does not automatically make the other obsolete.
In fact, the future may belong to systems that combine both.
Use AI when reasoning, interpretation, generation or creativity are genuinely useful.
Use traditional software when precision, repeatability and control are what matter.
And when possible, let the user decide which one gets the job.
That is a much more interesting vision of artificial intelligence than:
“Put AI everywhere because AI.”
🥶 Meanwhile, Back in City Deep.
I still had the flu.
I was still freezing.
My phone was still consuming my precious mobile data.
And the image was still not behaving.
Eventually, common sense prevailed.
I closed the applications.
I saved what remained of my data.
I stopped arguing with the digital super-brain.
And I decided that civilisation could probably survive without my particular contribution to the artificial intelligence revolution for one afternoon.
I went home.
I got under approximately three blankets.
And I went to sleep.
The machines can take over the world another day.
😂 The Final Verdict.
Artificial intelligence may eventually become extraordinarily capable.
It may write books.
It may design machines.
It may discover medicines.
It may help humanity solve problems we currently consider impossible.
But until it learns one of the most fundamental principles of graphic design, I’m not particularly worried.
Leave the left side alone.
Seriously.
Leave.
The.
Left.
Side.
Alone.
Because if a machine with billions of dollars of infrastructure behind it cannot manage that simple instruction, humanity may still have a little time before the robots take over.
And frankly?
I’m keeping my paint bucket.
✍️ Author’s Note.
This was written after a real-world attempt to make a very small graphical modification using generative AI while working in City Deep with limited mobile data.
The frustration was genuine.
The comedy was inevitable.
And the lesson may actually be useful:
The best technology isn’t always the most powerful technology. It’s the technology that uses exactly the amount of power necessary to solve the problem.
Sometimes that means artificial intelligence.
Sometimes it means a paint bucket.
And sometimes it means going home and getting under three blankets.
Othello Cody Verrocchio
in conjunction with ChatGPT.