A few days back, my son was stuck on a school assignment, and before I could even open my mouth to explain it, he had already typed the question into ChatGPT and had a full answer staring back at him. It made me pause for a second — when was the last time I actually worked through something myself, instead of just asking a chatbot?
That one small moment took me down a rabbit hole. Because if you think about it, this is not a new problem. We have been doing this to ourselves since the Stone Age.
We Have Always Traded Something for Power
I have always found this strange — of all the animals out there, we are probably the weakest. We cannot outrun a rabbit. We do not have the claws of a tiger. We cannot even see in the dark like most predators can. And yet, somehow, we sit right at the top of the food chain. Nothing hunts us. We hunt everything else.
Thousands of years ago, humans made a quiet deal with the world to get there. We were not the fastest animal on the ground, like the cheetah. We did not have thick fur or sharp claws, like the bear. So we picked up a sharp stone, learned to control fire, and made the first spear.
Every tool made us more powerful. But every tool also took something away from us.
When we started farming, we lost the endurance our ancestors had as hunters, always moving from place to place. When we learned to write things down, elders worried that people would stop training their memory. When calculators and GPS came along, we quietly gave up doing maths in our head and finding our way using our own sense of direction.
This has been the pattern for thousands of years — we hand over a task to a tool, we get used to the tool, and then there is no going back. Nobody gives up the calculator to do long division by hand again.
Now, in 2026, we have reached a new stage of this same old story. This time, it is not our hands or our legs we are handing over. It is our thinking.
Our Brain Is Slowly Becoming an Index
Look at how we deal with knowledge today.
For centuries, the human brain did everything by itself. We remembered facts, learned grammar rules, held formulas in our head, and recalled historical dates when needed.
Today, our brain is turning into something else — more like an index in a book than the book itself.
We are no longer the ones storing everything. We are the ones who know where to look, what question to ask, and how to make sense of the answer. We bring the intent and the judgment. The heavy lifting — the actual “thinking” — increasingly happens somewhere else, on a machine.
This is where it gets interesting. Every tool we made before this — a stone axe, a bicycle, fire — belonged to us the moment we made it. Once you sharpened that stone, it was yours to keep and use whenever you wanted, at no extra cost.
AI is different. It does not sit in your hand. It sits somewhere far away, in a massive data center, running on machines worth billions of dollars, using huge amounts of electricity. And every single time you use it, someone is quietly keeping count.
The Meter Is Always Running
At first, this new “brain” feels free. Anyone with a phone can open a chatbot and start asking questions, for free or for a small monthly fee.
But behind that simple chat box, there is a meter running. Every question you ask, every document you upload, every long conversation — all of it is being measured in small units called tokens. Think of it like a taxi meter that keeps ticking, even when you are just thinking out loud.
For a person casually asking a question here and there, this cost is nothing — a few cents at most. But the moment you start using AI to run entire workflows — say, an assistant that reads hundred emails, analyses them, and drafts replies automatically — that meter starts spinning fast. Very fast.
And here is the catch: we cannot simply choose to walk away from this, the same way our ancestors could not go back to living without fire once they got used to it. Once a society builds itself around a tool, stepping back means falling behind. So the real question isn’t “should we use AI.” It is “how do we use it without going broke, or without losing our own ability to think.”
Because really, there are two costs hiding in this story. One is the quiet cost to our own brain — what we lose when we let a machine do our thinking for us. The other is the very real cost in dollars and electricity, sitting in a data center somewhere, ticking away every time we ask a question. We need to worry about both.
Learning to Think Sharply, Not Just a Lot
I am now kind of convinced — we cannot avoid AI anymore. We need to master it. If we don’t, we lose relevance.
If we can’t avoid using AI, the smart move is to change how we use it.
The person who does well in this new world won’t be the one writing long, messy prompts and hoping for the best. It will be the one who asks short, sharp questions — the same way a good manager gives one clear instruction instead of five confusing ones. One precise sentence often beats three rambling paragraphs.
Interestingly, the question you type is usually not the expensive part. When AI goes searching the web for you, it quietly pulls in whole web pages, not just the answer to your question — and that adds up fast. The answer it gives back also costs more per word than what you typed in. So a short question can still lead to a costly search, or a long, detailed answer that adds up quickly. Your one line prompt is rarely the problem — what happens after it is.
But the one skill that matters more than ever is knowing when something feels off. AI can sound confident even when it’s wrong. So the real edge isn’t knowing more facts — it’s the instinct to question what you’re told, and to solve a problem yourself now and then, just so that instinct never goes rusty. A bit like astronauts exercising daily in zero gravity — not because they need the strength there, but so their body doesn’t forget how to function once they’re back on Earth.
But here is the thing — even if every one of us becomes a sharper thinker and a better prompt-writer, that alone will not fix the bigger problem. Individual discipline can only go so far. The real change has to happen in how these AI systems are built in the first place.
This Has to Work for Everyone, Not Just a Few
Fixing how we think is only one half of this story. The other half is bigger than any one person — it is about how these AI systems themselves are built.
If this technology stays expensive to run, only large companies and wealthy individuals will be able to afford using it at full scale. Everyone else will be left with the leftovers. That is not progress. That is just a new kind of inequality.
The good news is, this problem might be solvable — and not necessarily by throwing more money at bigger data centers. But I don’t claim to have the final answer here. I just have a few questions worth putting out there:
- Why send every small request to a giant, all-knowing cloud model? Do we really need that much horsepower just to draft a simple email, when a smaller model running right on our phone or laptop could do the job just as well, for free?
- Why should an AI solve the exact same common question from scratch, millions of times a day, for millions of different people? Shouldn’t there be some kind of shared memory — a way to reuse a good answer that’s already been worked out, instead of starting over every single time?
- On the other hand, when an AI is doing one specific task for you, why should it drag along everything you’ve ever said to it? Couldn’t it just carry the small, relevant slice it actually needs for that task, instead of burning power re-reading things that don’t matter?
- And maybe the bigger question — is the future of how we think safe in the hands of three or four big companies? Or should the core of this technology be open-source, so anyone can look inside it, question it, and build on it?
- Could there be something sitting quietly between us and the AI — a kind of translator that cleans up a messy prompt or trims a bloated search result before it even reaches the model, so we don’t have to think about the cost at all?
- And just as important — if we are paying for this, shouldn’t we get an honest, itemised bill? Most of us have no idea whether it’s our question, the search, or the answer that’s driving the cost. A little transparency could teach all of us to use AI more wisely.
I don’t know if these are the right answers. But they feel like the right questions to be asking.
A Story from 1959
This reminds me of something Volvo did back in 1959.
They had just invented the three-point seatbelt — the kind we all use in our cars today, the one that has saved countless lives. It was a breakthrough, and Volvo could have kept it to themselves and made a fortune licensing it to every other car company on the planet.
Instead, they gave the patent away. For free. To every competitor. So that anyone, in any car, anywhere in the world, could be a little safer.
Notice what they actually gave away — not just a safety feature, but a simple, efficient design that any car maker could build into their own vehicles without much extra cost. That is the real lesson here. It was not generosity alone that made it powerful. It was generosity combined with something efficient and easy to adopt. That is the real lesson buried in the AI cost conversation. Progress was never about how big or how expensive something is. It’s about how many people actually get to benefit from it.
If we build AI the way Volvo built that seatbelt — smart, efficient, and shared — then the ability to think, build, and create with this technology won’t just belong to the biggest companies with the deepest pockets. It will belong to a village teacher, a small-town researcher, and anyone else with a good idea and a phone in their hand.
That, to me, feels like the right way for this story to go.
anoop.blog 