Withdrawal
What happens when humanity suddenly loses free access to ChatGPT, Claude, and Gemini
Do you still remember what it’s like to write an email to a client completely from scratch? Let alone in another language. Seriously, give it a try. Then compare it to the last email you sent.
Not so great, is it?
AI seeped into our daily lives so quietly that nobody noticed the exact moment we stopped doing ordinary tasks on our own.
We ask ChatGPT things we would have figured out ourselves just a few years ago. We use Claude to make important business and life decisions in much the same way we drive everywhere with Google Maps.
No idea where we actually are, as long as the arrow keeps pointing the way.
Now imagine it disappears. Imagine waking up one morning and free AI is simply gone.
Why?
Because the companies giving away intelligence for free today, go bankrupt.
Impossible?
OpenAI, if the numbers are to be believed, burns money at a historic pace. The “give it away now, figure out the economics later” model has an expiration date. And that date is approaching.
When the money runs out, free access will be the first thing to disappear. At that point there will only be two options: bankruptcy of AI giants, or a price for the service that most people simply won’t pay.
That’s when things get interesting. Because we’ll finally find out how many of us can still do anything without a prompt.
The free lunch someone else paid for
Free AI has been subsidized since day one. The investor picked up the bill.
Will Lockett has become my main guide to the financial reality of this sector (though not my only one). For months, he has been breaking the numbers apart and exposing a truth that is obvious once you see it, yet almost everyone ignores.
Subscriptions were sold below cost. The goal wasn’t profitability - it was our addiction.
OpenAI itself admitted that, if priced honestly, its $200 plan would cost well over $2,000.
Imagine renting someone a Ferrari for $10 a month. Of course they’ll fall in love with it. The problem begins the day you ask them to pay what it’s actually worth.
And that day is coming.
Copilot’s pricing has increased several-fold in some cases. Uber reportedly burned through its annual AI budget within months and publicly concluded that AI can be more expensive than human workers. Microsoft reportedly canceled a Claude Code license because it was too expensive. One company generated a monthly Anthropic bill approaching half a billion dollars.
A tool that once felt “free” suddenly became too expensive - even for an NBA star-player.
Beneath the surface, there’s a genuine cash bonfire. According to Lockett, OpenAI posted an operating margin of negative 122% in the first quarter of 2026. Anthropic looks slightly healthier, but largely thanks to accounting gymnastics.
Funding sources are drying up. Venture capital isn’t keeping pace. Big Tech is nearing the limits of its willingness to fund losses. Debt markets are becoming saturated with risky AI-related paper.
That leaves one exit. The stock market.
Bring in retail investors and pension funds so they can throw more fuel onto the fire. In this situation, an IPO becomes a search for someone willing to grab the hot potato before it cools down.
We’ve seen this movie before in additive manufacturing - just with far fewer zeros.
During the SPAC boom in 3D printing, valuations reached billions while revenues barely reached millions. The story ended with Desktop Metal collapsing.
Markets can keep rising even while companies are falling apart. The mechanism is exactly the same. Only the numbers are bigger.
So if that’s the case - it has to end the same.
The post-AI hangover
Let’s assume Lockett is right. Free AI disappears, or becomes so expensive that it might as well disappear.
What happens then?
We'll discover who spent the last two years outsourcing their thinking.
Today everyone is a software developer. Together with Claude.
Everyone is a content creator. Together with ChatGPT or Gemini.
As long as you have another browser tab open, you’re competent.
Close it, and you’re left with the uncomfortable truth about what you can actually do yourself.
Science is beginning to measure this, and the results aren’t encouraging.
MIT Media Lab connected EEG electrodes to 54 participants and asked them to write essays. One group used ChatGPT, another used Google, and the third relied solely on their own minds.
The AI group showed the weakest neural connectivity of all three. They remembered less of what they had “written.” Many couldn’t even quote sentences they had produced just minutes earlier.
The researchers called it „cognitive debt”.
Michael Gerlich studied 666 people and reached essentially the same conclusion from another direction. The more frequently people relied on AI, the weaker their critical thinking became.
The mechanism was „cognitive offloading” - outsourcing thinking itself.
Young people suffered the strongest effects, particularly those who had never learned to work without AI in the first place.
It’s brain fog that we’ve engineered ourselves. A muscle that isn’t used becomes weak. Everyone understands that in the gym. The brain works the same way.
Then there’s the second layer. Withdrawal.
Addiction clinics are already documenting AI withdrawal symptoms: irritability, anxiety, fractured concentration, and the constant urge to return to the chatbot just to ask one more question.
In a joint MIT/OpenAI study, participants agreed with statements like:
If I lost access to my chatbot, I wouldn’t be able to cope.
That’s language straight out of an addiction therapist’s office.
Now combine the two.
You take away the tool people have become dependent on, and at the very same moment expose the fact that they’ve forgotten how to work without it. Withdrawal and incompetence arrive together.
It will be a long hangover...
Months of relearning how to sit in front of a blank page with nothing except your own slowly warming-up brain.
Because for the past two years we’ve lived under the illusion that intelligence was free.
The currency we paid with was our ability to think
Personally, I’m not particularly worried. For one thing, I seem to have a trait I call “immunity to addiction.” Maybe I’ll write about it one day.
At some point I simply realize that something is bad for me, and I quit. Immediately. Permanently.
No more alcohol, no more nicotine, I’ve never used drugs, I’ve never gambled, no social media (never really on Facebook or Instagram, never on Reddit or TikTok; LinkedIn is mostly a publishing platform for me rather than a place to comment on everyone else’s life, and X is a sewer - that’s the single one “I needed to quit”).
So even if part of me could theoretically become addicted to something, I don’t see any difficulty in walking away from it in matter of hours.
Yes, I’m not gonna lie - I use AI extensively in my work. But I don’t depend on it.
The words you’re reading now are mine. They weren’t generated by ChatGPT, Claude, or Gemini.
Yes, I rely on AI to translate my writing - Polish is my native language. So the worst-case scenario is that my English becomes raw, with more stylistic and grammatical mistakes.
But the core ideas won’t change.
If you’re genuinely good at something, you’ll remain good at it.
If your primary skill is writing prompts for AI, you’ll need a hell of a lot of money to keep doing it. This might be a good time to start buying lottery tickets. Or take part in some basketball camp and hope for a lucrative contract. Just for you to know - NBA free agency just opened, so this could be your moment.
And if you can’t afford a real, full-price AI subscription then here’s my advice, as someone who genuinely wishes you well:
Read real books, not Chat responses. Write your own stuff, not prompt.
Go back to the intellectual gym. Train your mind again.
Before it’s too late.