Tag: Artificial Intelligence

AI Stole My Cheese. And It May Steal It Again in Six Months.


I have been in technology and business for 26 years.

I started when creating a website itself was a skill. Then tools simplified it. PHP changed development. Frameworks reduced development time. Open source changed how we built products. Then came cloud, SaaS and APIs.

Every few years, somebody moved my cheese.

But AI feels different.

AI didn’t just steal my cheese. It seems capable of stealing the next cheese before I even decide where to keep it.

I have been using AI from its early days and watched it move from generic answers to usable content, code, reasoning and now agents that can actually perform tasks.

And I can clearly see gaps in today’s agents.

They struggle with some workflows. They need supervision. Enterprise integrations are messy. Exceptions confuse them.

The entrepreneur in me immediately thinks: “There is a business here.”

Then another voice asks:

“Will this problem still exist six months from now?”

That is something I have never experienced at this intensity in my 26 years of business.

Everyone Is Asking the Same Question

Talk to employees: Will my job survive?

Talk to entrepreneurs: What should I build?

Talk to software companies: What happens when AI can build this itself?

There are plenty of answers—learn AI, reskill, build agents, become AI-first.

But frankly, much of it still feels like writing in the air.

Nobody really knows where the capability line will be two years from now.

That made me realise I may have been asking the wrong question.

Instead of:

“What can I build that AI cannot replace?”

Maybe I should ask:

“What becomes more valuable as AI becomes more capable?”

Don’t Build Around AI’s Weakness

If I build a business around something AI cannot do today, my business depends on AI remaining bad at it.

That is a dangerous moat.

But if I own the customer problem—the workflow, integrations, domain knowledge, exceptions, relationships and operating process—AI becoming better can actually make my business better.

My cost of solving the problem comes down.

That completely changes the equation.

Don’t make AI’s limitation your moat. Make AI’s improvement your leverage.

Maybe I Was Trying to Plant the Wrong Thing

For years, I wanted to build something that lasts.

Something I could plant today and still see standing twenty years later.

AI made me wonder whether that is possible anymore.

Then my own journey gave me an answer.

HTML changed. Development tools changed. PHP changed. Frameworks changed. Infrastructure changed. Cloud changed things again. Now AI is changing almost everything.

But businesses still have customers.

Buildings still need managing. Companies still need moving data between systems. Money still needs collecting. People still need problems solved.

And suddenly it became obvious:

The problem was the tree. Technology was only its leaves.

Leaves are supposed to change.

What Do I Plant Now?

I don’t know where AI will be five years from now.

I don’t think anybody really does.

But I don’t need to find something AI will never change.

I need to find a problem worth owning even if the way we solve it changes five times.

A temporary AI gap might last six months.

Use it.

It brings customers. Customers bring workflows. Workflows bring knowledge. Knowledge creates systems, relationships, distribution and trust.

Those may survive the next model release.

Maybe creating a legacy in the AI era isn’t about building something that never changes.

Maybe it is about building something that knows how to change without losing the problem it exists to solve.

After 26 years of technology repeatedly moving my cheese, perhaps that is the lesson:

AI will keep moving my cheese. My job is not to guard the cheese. It is to own the dairy.

21 Attempts Later: How ChatGPT and I Found the Answer Together


We talk about AI as if it’s magic.
Ask a question. Get an answer. Move on.

What we don’t talk about enough is what really happens when the answer doesn’t come easily.

This week, I learned that the hard way.

What looked simple on paper turned into 21 failed attempts, each one slightly different, each one confidently wrong. ChatGPT responded every time — clearly, logically, persuasively. And every time, something didn’t work.

That’s when I realised the first uncomfortable truth:

AI can sound right long before it is right.

The early illusion

The first few attempts were deceptive.

The responses were structured.
The explanations were neat.
Some even ended with words like “success”.

And yet… nothing actually happened.

Acknowledgement masqueraded as execution.
That illusion alone can waste hours if you’re not careful.

When confidence became the problem

By attempt seven or eight, both of us — ChatGPT and I — were confident.

The logic seemed airtight.
The fixes were small.
We were “almost there.”

That phrase — almost there — is dangerous.

Because it convinces you not to question your assumptions deeply enough.

The conversation changed

Somewhere around attempt eleven, I stopped asking ChatGPT what to do.

Instead, I started telling it what was wrong.

“This assumption doesn’t hold.”
“This part works; this doesn’t.”
“Let’s isolate just this behaviour.”

ChatGPT changed with me.

The answers slowed down.
The certainty softened.
The reasoning became cautious — collaborative.

That’s when it stopped feeling like a tool and started behaving like a thinking partner.

The humility phase

There was a stretch where neither of us rushed.

No clever shortcuts.
No sweeping rewrites.
Just deliberate, line-by-line progress.

I stopped expecting brilliance.
ChatGPT stopped pretending certainty.

Ironically, that’s when progress accelerated.

Attempt twenty-one

The final attempt didn’t announce itself.

No drama.
No celebration.

It simply worked.

And in that quiet moment, something became clear:

Success is often silent.
Failure is loud.

What this taught me about AI

ChatGPT didn’t replace thinking.
It demanded better thinking.

Weak prompts produced confident mistakes.
Better prompts invited reasoning.
Persistent correction reshaped responses in real time.

The miracle wasn’t AI.

The miracle was staying in the conversation.

The real takeaway

This wasn’t man versus machine.
And it wasn’t man commanding machine.

It was a convergence — through frustration, feedback, and patience.

Human intuition corrected AI assumptions.
AI pattern recognition sharpened human thinking.

Twenty-one failures later, the result wasn’t just success.

It was earned clarity.

Final thought:
The future won’t belong to people who use AI.
It will belong to those who can persist with it, long enough for understanding to emerge.

Because intelligence — human or artificial — means nothing without perseverance.