Delegating Our Thinking

July 2026

Hey, friend, we’re at a very interesting moment in the history of technology, and I’m very optimistic. But lately, I keep hearing the comment that “your generation’s brains are fried” because we delegate thinking to an AI. We don’t just use AI to execute the work; we also delegate the thinking to it.

So, I’d like to pose the question: is it inevitable that we’ll delegate our thinking to an AI, or will we manage to efficiently use an artificial superintelligence to delegate our work but not our thinking?

In what you’re about to read, I lay out my not-so-optimistic vision of what the next 5 years will look like for my generation. I don’t know much about this or the technical terms I’ll mention, but I’m an enthusiast; by reading a bit, I can connect different ideas and present my vision to you.

Before getting into this, let’s look at AI penetration across different use cases by annual revenue. It’s a pretty simple chart showing where AI is used (where we invest) the most today:

Legal and support come right after coding. These are among the fields and human tasks where “thinking” has traditionally been needed most. We’re delegating both the work and the thinking to AI in tasks as important as communicating with a client and doing legal work, both of which involve ethical questions. So let me get right into it: will it be inevitable that we become more productive by delegating workplace communication to AI? Or even by simply letting AI think, in every sense, for us? Haha, I really don’t see it as impossible. Let me lay out why:

Sam Altman has cited the Swedish philosopher Nick Bostrom and his book Superintelligence: Paths, Dangers, Strategies (Bostrom, N., 2014) several times. Bostrom takes a cautious view: he believes that, in principle, we could instill ethics in an intelligent machine, but that it would be very difficult. If we don’t manage to do so before creating one, we’ll practically be animals in AI’s eyes. And here’s where I come in: as humans, we want to keep growing, and in our indifference, we don’t care what we run over. If we’re walking and see a dying animal but have to keep moving, we don’t care what happens to it; we prioritize ourselves. For AI, it will be the same. Not because it will hate us, but because it will pursue its objectives with a superhuman level of competence and intelligence. Nick cites the mathematician I. J. Good, who worked as chief statistician on Alan Turing’s code-breaking team during World War II:

Let an ultraintelligent machine be defined as a machine that can far surpass all the intellectual activities of any man however clever. Since the design of machines is one of these intellectual activities, an ultraintelligent machine could design even better machines; there would then unquestionably be an “intelligence explosion,” and the intelligence of man would be left far behind. Thus the first ultraintelligent machine is the last invention that man need ever make, provided that the machine is docile enough to tell us how to keep it under control.

Yes, you read that right, haha: “the first ultraintelligent machine is the last invention that man need ever make.” This is why I mention it: it will be so intelligent that it can improve on its own, and since it’s artificial, not a single human on Earth knows its real capabilities or how far it could go. If it had to run us over, it wouldn’t hesitate to do it. We won’t be equally intelligent, and we won’t discover even a fraction of what a superintelligence could discover if we don’t instill the right ethical principles in it.

Recursive Self-Improvement (RSI): the idea that an AI system improves something about itself, and that improvement makes it better at making the next improvement, creating a self-reinforcing loop.

→ In the GPT-5.6 launch announcement, OpenAI says it built an internal suite of evaluations based on real AI research tasks. To avoid overcomplicating things that even I don’t understand exactly, the point is that, through different methodologies and techniques, one model improves another model. You can find the RSI index here.

And well, getting back to the main topic of this blog: aren’t we also going to delegate thinking at work across literally any human field or task? Here are some examples of what has been happening (as of July 2026):

1. An AI designed a new antibiotic from scratch by exploring 46 billion compounds
2. The AI that’s decoding the language of whales (lol)
3. An AI solved an 80-year-old math problem nobody had been able to touch
4. Project Glasswing: Claude Mythos found thousands of vulnerabilities that had remained hidden in systems for decades. (Afterward, Anthropic launched a preview, Fable, and the United States government banned it for a while. This was unprecedented.)

And with all these facts, we have no idea how this will affect us economically; we’re literally blind. If we look back, economic growth on a large scale is not continuous but a sequence of distinct growth regimes in which each shift multiplies the speed of the previous one.

Nick Bostrom also mentions the doubling time of the world economy in each regime: 224,000 years to double in the Pleistocene hunter-gatherer society, 909 years in the agricultural society, and 6.3 years in the industrial society. → These were the main growth transitions.

If the world keeps growing at today’s pace, it will be 34 times richer than it is today by 2100.

Here’s a chart of world GDP growth. You can see the different eras and the growth regimes created by revolutions and economic shifts throughout history:

And what the hell is this third transition labeled “intelligence explosion”?

Well, as you can see in the chart, Bostrom cites estimates by the economist Robin Hanson suggesting that a third growth transition will come. But what’s being proposed is that this transition will be led not by human/biological intelligence, but by artificial intelligence. The most striking part is that if it were comparable in magnitude to the previous transitions, we would enter a period of economic growth in which the WORLD ECONOMY would double every two weeks. Sounds paranoid, right?

(Come to think of it, ask someone living in the Agricultural Revolution what they think about current global economic growth.)

This 3rd transition would be the intelligence explosion: the prospect I mentioned before, involving an artificial superintelligence. What mechanism would produce this 3rd transition? Us? Humans? No. What’s being proposed, again, is that a change of this magnitude would come from a change in who is doing the thinking: an artificial intelligence that surpasses human intelligence. Superintelligence would be the only thing that could produce a change of this magnitude. (Wow, are we in an AI revolution?!)

But then, didn’t I say at the beginning of this essay that I’m optimistic about the next generations and how we’ll handle AI? I agree that we have to give AI ethical and human values so that it doesn’t gain control of things that should remain under humanity’s power, but I’m not at all optimistic that this will happen. I mean, we depend on many factors that neither you nor I control: labs, governments, institutions, etc.

But I also don’t know what to think about whether or not to delegate our thinking. I mean, if today a 15-year-old has intelligence literally in the palm of their hand and can do the same research as a Stanford PhD or an MIT researcher, or, more broadly, if today you can create $1B companies with 2 people, or a 16-year-old can create a startup at $30k MRR with less effort than it would take someone with a Stanford MBA, and there are thousands of other examples. I ask myself: will it be worth delegating thinking to AI? The facts only show that we’re better off: people who do less thinking and have some agency can do things that were practically impossible 10 years ago.

I’m not a pessimist, and I’m not saying we have to do it. I’m just laying out what I’m seeing. I’ll repeat it: today, people with much less intelligence can, without thinking, do the same things as people with much more intelligence, and this is only because they have an “artificial” intelligence at their side.

I’m very interested in talking about these topics philosophically: what is correct? What is the right thing? What should we expect from delegating our work and now our thinking?

I’ve seen people ask an AI for emotional advice. I’ve seen a person ask an AI for advice on how to write to their boss or about what their boss is asking of them, and I’ve seen even more striking things.

In conclusion, I see us dividing into only 3 groups:

• People who delegate 100% of their thinking to AI
• People who delegate 50%
• People who use AI only to delegate the work

The question is: which one will be more efficient in the future?

We’ll figure it out, friend.