In the late 1960s, a Stanford psychologist named Walter Mischel put preschoolers in a room with a marshmallow. The rules were simple: eat it now, or wait fifteen minutes and get two.
Some kids ate immediately. Others waited.
Mischel tracked them for decades. Turned out that the ones who waited had better SAT scores, lower body mass indexes and better stress management.
Delayed gratification, the experiment suggested, was a predictor of success.
The experiment (which was later replicated, with even more interesting findings) became a staple of self-help literature. Discipline defines destiny. The ability to resist now in favor of later separates winners from losers.
And then came AI.
“ChatGPT, find me flights to Lisbon under 200 euros.”
“Claude, code a script that processes these CSV files.”
“Gemini, summarize these three hours of meetings into action items.”
These aren’t hypotheticals, this is a regular Tuesday morning for millions of people.
Tasks that required effort—sometimes hours of it—now take seconds. The search, the comparison, the learning curve, the context switching, the debugging? All absorbed by something that never gets tired.
I catch myself doing it more and more. Something that would have taken me an afternoon to research now takes a prompt and thirty seconds.
The marshmallow doesn’t exist anymore. There’s no waiting anymore. You get both marshmallows now.
And this is where I think it gets really interesting.
For the first time in human history, we have a technology that changes the relationship between effort and outcome. Not like tractors replaced manual farming. Not like calculators replaced mental math. Those were just tools, amplifiers.
This is different. This is the compression of cognitive labor itself.
Think about what we actually learned during those hours of searching for flights. We built a mental map of airline routes. We developed intuition for price fluctuations. The friction forced us to evaluate whether the trip was worth it at all.
Now that friction is gone. The thinking happens elsewhere.
What happens to a generation that grows up without that friction?
I don’t think anything apocalyptic will happen. But I do think something very relevant – generational level relevant – is just around the corner.
Here’s what I’m watching for:
1. Society will split on patience
Some people will become remarkably impatient with anything that can’t be delegated to AI. If a task takes more than a few minutes and AI could do it, they’ll feel it as wasted time.
Others will go the opposite direction. They’ll deliberately choose slowness. They’ll see patience as something worth protecting.
Right now, patience is still considered a universal virtue. In ten years, it might be a lifestyle choice. Something you opt into, like meditation or digital detox.
2. Doing things the hard way will become a status symbol
When mass production made goods cheap, handmade became expensive. Artisanal products carry a premium precisely because they’re inefficient.
The same thing will happen with cognitive work.
Hand-coded websites. Manually researched travel itineraries. Essays written without AI assistance. What I call bio-content, provably human generated content.
The process itself will become the product.
We already see early signs. And I think this will only grow.
3. Knowing what to ask becomes the new skill
The marshmallow experiment didn’t test what you did with the extra marshmallow. It only tested whether you could wait.
Maybe that’s the new test. Not whether you can do the work, but whether you know what work to request. Whether you can orchestrate AI tools effectively. Whether you can evaluate the output.
Prompting well, directing AI, knowing when to trust it and when to verify—these are becoming real competencies. In some fields, they already matter more than the underlying technical skills.
4. The capacity for difficulty might weaken
This is the one that concerns me most.
There’s a specific capacity that develops when you stay with something difficult. Not because you have to, but because that’s how capability builds. The willingness to be confused. The patience to debug for hours. The tolerance for not knowing.
If every hard thing can be outsourced, what happens to that capacity?
I’m not sure we know yet. But attention without regular exercise tends to weaken. Muscles you don’t use atrophy. I suspect the same is true for the ability to persist through difficulty.
I’ve been coding since 1987. I’ve built companies, written thousands of blog posts, run ultramarathons.
Most of my skills were built through repetitive, often frustrating effort. Hours of debugging. Days of research. Months of building physical resilience that only 0.00001% of the people on this planet can reach.
My children will never experience the world the same way. Their cognitive friction will be much lower – if any at all.
Is that a problem?
I genuinely don’t know.
Maybe the friction I remember fondly was just waste. Maybe the real skill was always something else—creativity, connection, judgment—and the grunt work was just the price we paid because we had no alternative.
Or maybe delayed gratification wasn’t just a predictor of success. Maybe it was the training itself.
We’re running the marshmallow experiment in real time: an entire generation raised with AI as cognitive infrastructure.
We’ll know the results in about twenty years, maybe sooner.
Until then, I’ll keep asking Claude to help me code things faster. And I’ll keep doing some things the hard way, just to make sure I still know what it feels like.
dragos@dragosroua.com (Dragos Roua)
Source link
